The rwx cycle: read, write, execute.

Jiabin Huang research advice offers a variety of tips for researchers.

First, perform a literature review.

Boiling the Ocean method: There are various "awesome-xyz" premade lists on GitHub that you can use to find papers. Consult past conferences to see where the field is heading. Cluster these papers into categories to understand "why x hasn't been solved yet".

Availability heuristic: Since many papers exist on X, you can just ask Grok to search for threads. Alternatively, look online at your favourite theorists' websites, as well as non-traditional researcher websites, and see if there are any recent writings that haven't been implemented yet. For example, I have implemented [undisclosed] suggestions from gwern.net for a research project, which isn't widely seen in academic literature.

Then, feed the context into your favourite research agent, in this case, I use Feynman developed by my mutual Advait and have suggested the following prompt contributions.

Feynman prompts sorted into different levels of granularity:

1. [insert link] look around! As a leading expert in neuroscience, explain this lab's research canon and how it situates in the contemporary work ranked by originality, then point to individual significant papers of interest.

2. You are a research methodology expert. For each paper, identify the major vs minor contributions buried in this data, rank them by originality, and show with precision and accuracy where each one challenges or extends existing literature. Use bold text very sparingly for notetaking in Logseq. Provide a glossary for a technical 20 year old. For each formula, provide intuition for reconstructing it step by step, with patterns understandable to a technical 20 year old. Do not use bold text.

3. Still acting as the research methodology expert, explain from the primitives level how they designed the pipeline, the quality and type of evidence and reasoning used, and consistency.

4. Michelin-tabling question (if you're meeting with the authors) As a leading neuroscientist, you find yourself at a fancy lunch with the authors. Read the discussion or looking forward section while thinking granularly. What are the 3 quirky follow-up questions that you would discuss with the team? What are the 3 key implications of this argument that you would encourage the authors to think about?

• If you have around a week for a project, you should start running experiments ideally on day 1 or 2 to leave time for adjustments. Notice and resist the urge to execute all experiments on day 4.
• 50% of the time should be spent on running experiments, logistics, or building.
• Code should be constantly running.

Representations of the World in the Human Brain by Jack Gallant, an experimentalist.

According to your "why x hasn't been solved yet" buckets, continue narrowing each bucket down to testable and plausible hypothesis lineages. Come up with a list of hypotheses (anywhere from 1 to even 40) ranked by upside risk and time taken to test.

As soon as we have the hypothesis, we can scope the experiment. Tetraslam's playground has a great style guide for agents. You should also write a markdown style guide.

Sketch a diagram of how information flows through the key components, starting out simple. Pick your tools accordingly for each stage. Ask agents to contrast tools by community size, features, cognitive complexity, and pick the most relevant one for your use case.

It is a delicate balance to understand the experiment you are running vs optimizing for speed. You will feel cognitive strain. Adjust mental impedance accordingly. Remember: Too much speed can lead to accumulated technical debt exponentially further down the line. (See A Philosophy of Software Design by John Ousterhout.)

Put all the observations and graphs into a page and share it with your collaborators. If an experiment didn't work, why? Brainstorm the most probable explanations. Review related literature again, adjust the direction, and proceed to test.

Models need to be evaluated. Consult vals.ai

• Throughout this entire time, you should be accumulating data points in all sections of the paper in parallel, rather than starting from scratch.
• Figures drive the narrative. See: https://x.com/jbhuang0604/status/1437443017510621185, https://x.com/jbhuang0604/status/1665738070002483201, and https://github.com/jbhuang0604/awesome-tips/blob/main/paper-writing.md
• Write these first: Introduction, Related Work, Methods
  • Discussion: Feel free to word vomit, and ask Feynman, or equivalent, to clean it up later.
• Write these while you are running experiments:
  • Discussion, Appendix
• Write these last:
  • Key contributions (Abstract), Results
• Ask Feynman, or equivalent, to compile the paper from any medium, such as markdown, into TeX
• Preferably, import the paper into Overleaf so your collaborators can edit it too, as opposed to just having a PDF.

In most studies, the resting state data is first collected through fMRI and MRI.

At the MRI scanning facility…

When I lied in the MRI machine, I heard 5 different octaves which performed a T1 echo sequence. My brain was not visibly different from most brains by rough inspection. I wondered about the white matter to gray matter ratios, especially in trans individuals.

study #2's setup

study #2: preparing the hair and ultrasound transducer

A high density EEG system was used in the 3rd study. Researchers used a 128 electrode cap and injected saline water into the electrodes with a pipette.

The baseline data collection involved a task that minimizes blinking with intermittent free-blinking breaks. This was subjectively painful. The research team was evidently speedrunning the study and was under immense pressure; The PI was considering skipping the lab meeting, offering me a chai latte.

Then there was the gold cup electrodes for recording during the task; My scalp was prepped with abrasive scrub to remove deadskin and gel for the gold. Three electrodes, blue, red, orange, were placed on the back of my head (ground, reference, signal) with several other electrodes. The gold electrodes are more conductive but are paired with an amplifier from the 1960s i.e. Grass technologies, which kept the noise in the data. This is contrasted with the other 128-electrode system that used an EGI amplifier from the 2010s with noise filtering.

Memory tasks were given in a screen usually involving a fixation cross in the center, and then a mirrored setup: The screen was divided into left, right, top and bottom regions where the stimuli could occur. Only one side of my brain was stimulated.

It felt like being lightly zapped on the forehead. There was no temperature increase for more than 2 deg. Even though the transducer may be on for 40 mins, it is only operating at 10% duty cycle. The ultrasound was audible as it pulsed at designated frequencies: 5 Hz, 50 Hz and 500 Hz.

There were sham sessions which stimulated unrelated regions milimeters shallower/adjacent to targeted regions. For example in study #2, the focused ultrasound targeted the insula first in a later-revealed sham session, which may explain a reaction time speed up.

My reaction time became noticeably & marginally faster and working memory increased to my own peak performance point.

However, the memory task in study #2 was moderate difficulty. I had came up with a heuristic before stimulation by mentally drawing paths and noting dots around the monitor to get the right answer. After stimulation, I feel cognitively effortless and ditched the heuristic, and likely performed still the same in terms of measured results. In study #2's 2nd session, the pulvinar (correct region) got targeted. this is a coordinator nuclei in the thalamus for visual attention, processing and filtering. Subjectively, I registered no noticeable effect.

Wiggling. The ultrasound had to pass through milimetres to hit study #3's target region: the LGN, just as large as the beam's focus, buried deep inside the brain. While I accidentally wiggled in my chair, I felt a phantom cold sensation in my throat. I suspect my vagus nerve was stimulated by accident through the incident beam. "Can focused ultrasound intentionally cause cognitive impairment?", @FemboyEE asked. This is practically demonstrated to be quite likely using present day set ups.

Manual transducer placement was the bottleneck in this study, taking 15 mins on average. Robotic arms are developed for TMS coils e.g. Axilum robotics, but could not generalize to the coregisteration step. Narrow robotics applications emerge frequently. 6 DOF arms like the LeRobot SO-101s are everywhere, one can vibecode a hello world to move all the servos in 10 mins, its a race to generalize as of 2026.

Algernon-Gordon effects remain unstudied. The postdocs weren't sure about how the focused ultrasound stimulation decays as a function of time, since they're focused on the what.

Institutional arbitrage. The study coordinator mentioned skipping their masters and directly going to a PhD program. Thus they needed to publish papers via a coordinator role. As it is currently May, I think the team was under immense pressure to finish the study before major deadlines. The publish-or-perish pressure had caused me, an outsider, physical pain.

Mistakes are more information than correct results. Through analyzing types of errors made at each stage in study #2 (directing + suppressing attention and manipulating information) among at least 30 participants, researchers could infer the ultrasound effects.

Computing history has moved from large, centralized mainframes to distributed PCs back to the centralized cloud, followed by distributed local models so the models within the same enterprise can communicate quickly with each other. There appears to be a current swing in computing to specialization, evident in various GPUs, ASICs and aforementioned local models. At the Founders Inc.'s Physical AI hackathon this past weekend, people toasted bread and set tables with robotic arms. One team found it easy to drop bread into the toaster, but harder to "push down the toasting button" to refine the toaster pipeline. In robotics, the pendulum seems to be swinging towards generalization.

"Software ate the world". AI eats software. Hardware increases in relevance.

In an assembly line, whether in the cafeteria or in software, having the best people at the start is important. Shift left.

I noticed I had been combinatorially testing code. Combinatorial testing could not grow at Moore's law's rates.

RAM shortage expected to last until late 2027, reflected in the current silicon stock market trend. Fabs are being built/infrastructure is built from ground-up. Cerebras IPO! Cerebras is coming to IEEE @ Stanford May 18th.

1. Many companies have policies sponsoring only top-tier hackathons with famous people or large prize pools. If you do want to organize a hackathon, start from having interesting people as the base layer. This goes for seeding social networks in general.

2. A Unitree partner was kicked out of the hackathon venue, since they weren't sponsoring any APIs, prize money more than displaying a robotic dog. An employee was treated as a cog. Unitree's robotic dog wasn't technologically groundbreaking, however, the co-organizing company in this hackathon is just remixing the LeRobot arm. If everyone is repackaging existing open-source systems, the social norm shifts towards gatekeeping rather than creation. This is like how the DJ culture is a bit cold since people're remixing music that is publically accessible, compared to the ML research scene where there are thinking moats.

3. Technical skills are often the entry ticket, systems-thinking skills are differentiators. In the NeurotechX Hackathon at Frontier Tower, Sharena Rice shared a systems account of failure modes in neurotech. Individuals with deep expertise get stuck in neurotech with no authority or decision-making patterns. My friend and I discussed that after ~Jul 2025, people purely focusing on technical skills are kind of having a challenging time.

4. One of the hackathon participants tracked neurotech funding via IBIS World, cold emailed accordingly and secured an internship.

Permutation City. First impressions: Its 2050 homogenous-earth timeline seem to closely echo late 2020s San Francisco in a heterogenously developing world. It gets AI, VR systems that we have now, but not yet any mind upload. Also, the characters are major procrastinators and heavily use coercion.

Cybersentics book club — Sentics was the proto-cognitive science in the 1950s. There was 7 universal emotions across cultures, measured with pressure of the tip of the finger when a stimuli is shown. It resembled a crude version of EEG waveforms. Composer's typical musical forms correlated to sentics. There are many well-thought out discussion questions by the host, Anastasia Chernysheva. I would recommend going to these book clubs.

One sleepy morning, I appeared for a transcranial ultrasound study at the Lucas Centre at Stanford, where I was sent into an MRI machine, asked to focus on a visual stimulus and click buttons accordingly. Despite my sleepiness, the force is extremely strong with this one (metaphorical and literal), and I instantly focused!

Fig 1. A flyer recruiting for a focused ultrasound study at Lucas Center.

Historically, MRI has dominated the medical imaging of the brain. Ultrasound was traditionally considered inadequate for neuroimaging, due to a series of reflection issues making the image blurry, but algorithmic improvements have caught up to speed these days to remain hopeful.

Fig 2. A resolution enhancement technique from Guasch et al. (2020), contrasting unimproved ultrasound (c) to improved FWI ultrasound (d). Source: Nature Digital Medicine [8].

Each of MRI and ultrasound exploits different tissue properties and wave physics. Ultrasound stimulates the brain via mechanical pressure waves and is favoured for its capabilities to reach deep structures with a handheld device at a reasonable cost. One simply applies conductive gel on the subject’s head and sets the transducer array on top. Then, pulses of ultrasound are delivered, transmitting via longitudinal pressure waves into the tissue.

Fig 3. An ultrasound transducer array on the subject’s head. Source: Deffieux et al. [6].

17.2.32.3 MRI basics

MRI is an imaging tool based on the water nuclei within the human body and their responses to magnetic fields. Initially, within the water nuclei, the protons are randomly oriented. A radio-frequency (RF) pulse, which is a burst of electromagnetic energy, disturbs the water nuclei. The nuclei have excess energy and will, after a certain period, spontaneously lose the energy by emitting another RF signal as they relax. These signals are measured by the computer [2].

During the time when protons have gained energy, they rotate about their axis, known as precession. The precession varies in frequency, known as the Larmor frequency:

f = γB

where γ is the Larmor constant (magnetogyric ratio) and B is the magnetic field strength. The Larmor frequency is sent back to the computer.

Fig 4. Diagram of a singular proton precessing. Source: UPenn MRI Basics [2].

How does the computer distinguish between multiple different protons with the same Larmor frequency? One answer is to produce different frequencies by varying the magnetic fields. Inside an MRI machine are three coils that coordinate to do this:

First, the main coil produces a uniform magnetic field for the scanner as a backdrop.

Fig 5. The main coil inside an MRI scanner. Source: MRI coils explanation [3].

Second, along the z-axis, several subcoils introduce a linear gradient of intensifying magnetic field. As the magnetic field becomes stronger, the Larmor frequency of protons also increases. Thus, the frequency becomes a function of position along the z-axis, also known as the technique of Motion Encoding Gradients (MEG).

Fig 6. Gradient coils showing a linear gradient. Source: MRI coils explanation [3].

Finally, the RF coil delivers specific frequencies, perturbing only the protons whose natural precession frequency matches the RF. Each snapshot is a 2D picture which is stacked into a 3D structure of the brain.

Fig 7. The RF coil inside MRI. Source: Source: MRI coils explanation [3].

As the ultrasound stimulates, MRI watches the brain activity. However, the biological tissue behaves like an overdamped harmonic system: it shifts positions during ultrasound, and a bit after the sound turns off, relaxes back to the original spot. Simultaneously stimulating and recording with MRI is not actually the optimal approach. The displacement as a function of time can be modelled as:

D_rise(t) = (F/k)(1 − e^(−t/τ_rise))

D_fall(t) = D(T_off) · e^(−t/τ_fall)

K is a proportionality constant resembling the k in Hooke’s law, while τ_rise and τ_fall resemble the capacitor discharge constant. For brain tissue, the rise time is roughly 5.97 milliseconds with a shear wave speed of 1.56 m/s [4].

Brain tissue is heterogeneous in stiffness. (Side note: one modern challenge in neurotechnology is to match the stiffness of the brain with the implant so that it doesn’t get squished.) Young’s modulus provides a measure of strain [7]:

E = stress / strain = (F · L₀) / (A · ΔL)

If an area is destroyed by heat (a common scenario when ablating brain tumours with ultrasound) ΔL decreases, effectively increasing overall stiffness.

Considering the delayed movement of protons in the magnetic field in response to the ultrasound, a phase shift is accumulated. The phase is the time integral of the motion:

ϕ = γ ∫(T₀ to T_enc) G(t) D(t) dt

Since researchers noticed the damping, they invented a method by shifting the focused ultrasound pulse to a few milliseconds before the MRI gradient begins, described in the MR-ARFI paper [4], which was used in the Lucas Centre’s study.

Now, if the wet tissue causes problems by being a damped harmonic system, then the skull also produces problems by being too stiff and smearing the signals. The skull has a higher speed of sound (˜3500 m/s) compared to soft tissue (˜1500 m/s). Similarly, MRI itself only has millimetre imaging resolution, while the displacement caused by ultrasound is on the order of a few microns—too tiny to see.

Skull aberrations are difficult to correct for. There are two recent solutions. In the original MR-ARFI paper [4], the approach is to poke the brain and measure in a loop until the tissue displacement is maximised. Another approach borrows from a mature technique originally for modelling hydrocarbons in the Earth, viewing the human head as a geological structure. This is known as Full Waveform Inversion (FWI) [5].

17.4.14.1 The FWI Algorithm

FWI is an iterative optimisation process to find a computational model that reproduces experimental data.

Step 1: Forward model. It starts with a default model of the brain tissue and propagates the wave forward. It computes the acoustic wave equation—a manifestation of Newton’s second law in continuous media—discretised as:

Au = s

where A is a large, sparse matrix representing the discretised wave equation, u is the pressure wavefield (a solution at every point in space and time), and s is the source vector.

Step 2: Cost function. It compares the predicted data with experimental data produced by the transducers to create a cost function that it seeks to minimise:

f = ½ (p − d)ᵀ (p − d)

Here, p and d represent the predicted and observed data, respectively, as vectors.

Step 3: Gradient descent via chain rule. To minimise the cost, it uses gradient descent. The gradient of the cost function is computed with the chain rule:

∂f/∂mᵢ = [∂p/∂mᵢ]ᵀ (p − d)

The intuition is to proceed with backpropagation:

∂f/∂mᵢ = (∂f/∂p) · (∂p/∂u) · (∂u/∂mᵢ)

We have respectively: the loss derivative, the data operator, and forward sensitivity. However, computing ∂u/∂mᵢ directly is impossible: u is a huge continuous vector representing the forward wave, and m(x) is a huge continuous field rather than discrete parameters as one typically encounters in machine learning. For a 3D model of submillimetre resolution, we would need to solve 100 million individual parameters m simultaneously.

Step 4: The adjoint method. Fortunately, one can bypass computing du/dm and rather only compute its effects on a vector. Imagine playing the wave from the source back to the target, stepping backwards in time in the wave equation by taking advantage of Aᵀ, the time-reversed propagation operator. This is the adjoint method.

Differentiating Au = s with the product rule applied to A(m)u(m) and noting s is a constant:

A (∂u/∂mᵢ) + (∂A/∂mᵢ) u = 0

Isolating the wavefield gradient (assuming A is invertible, i.e. the wave equation has a unique solution):

∂u/∂mᵢ = −A⁻¹ (∂A/∂mᵢ) u

Substituting back and using the property of transposes:

∂f/∂mᵢ = −uᵀ (∂Aᵀ/∂mᵢ) A⁻ᵀ Rᵀ (p − d)

Finally, the algorithm matches the backwards wavefield with the forward wavefield, requiring in total only 2 solves per iteration compared to 10⁸ solves. By using this efficiency trick, the researchers chunked 32 hours of computing into a process that could run under 10 minutes on a small GPU array [5]. Ultrasound becomes much more focused and a powerful tool.

Fig 8. Flowchart of the Full Waveform Inversion algorithm with credits to claude

In San Francisco, focused ultrasound has been having a moment since the end of 2024, with the rise of the transcranial focused ultrasound society (tFUS society) and collaborations with local non-profits, including Foresight Institute, Frontier Tower, Stanford University, and the Neurotech at Berkeley. What a time to be alive!

Infamously at hackathons, certain participants remain productive deep into the second night, while others gradually descend into deadly errors.

When studying sleep deprivation, the effective altruist diaspora is particularly interesting for three reasons. First, a productivity self-experimentation culture already exists: Guzey’s experimentation being a prime example, where he discovered sustaining on 4 hours per day of sleep over 12 days did not significantly impair his cognitive performance measured via the psychomotor vigilance task, SAT and a custom first-person shooter reaction time game. Thousands of essays speculating on sleep deprivation proliferate on lesswrong.com, forum.effectivealtruism.org, astralcodexten.com, gwern.net and similar sites. The availability of peptides from Chinese manufacturers encouraged self-experimentation culture, gaining a critical mass by September 2025. Secondly, as EAs carefully choose high-impact careers, sleep deprivation could have disproportionate downstream effects. Thirdly, this population is cognitively sensitive to sleep disturbances. However, no peer-reviewed studies examine the EA diaspora as a collective impacted by sleep deprivation.

Conventionally, sleep is characterized by the two-process model: Process S builds up sleep pressure via adenosine linearly throughout the day, while Process C initiates sleep onset originally from the suprachiasmatic nucleus of the hypothalamus (Borbély, 2022). Both processes interact to generate alertness peaks in mid-morning and early evening, and troughs during pre-waking hours and afternoons.

During extended wakefulness, Process S continues to climb while Process C cycles through its natural rhythm. Thus, energy crashes hardest during the pre-early morning hours and in the following afternoon (Durmer & Dinges, 2005).

Vigilant attention and reaction time are the first functions to deteriorate in sleep deprivation, marked by increased P300 latencies and decreased amplitudes. In the following phase, higher cognitive functions are affected as a network. There is alternating activity between the default mode network during background tasks and the frontoparietal control network during task processing instead of prolonged activation of either mode (Khan & Al-Jahdali, 2023). The thalamus becomes noisy and frequently interrupts. More lapses, microsleeps, omissions and commissions become visible. Working memory capacity drops. Executive functioning and reward systems follow: Median PFC-amygdala coupling becomes more error-prone, and decisions take a longer time. Simultaneously, memory consolidation weakens. The Hippocampal cAMP/PKA and mTOR pathways are downregulated. In the third phase, longer-horizon mechanisms are more visibly impaired: poor renormalization of synapses as well as toxic waste buildup from lowered expressions in AQP4 and glymphatic clearance (Khan & Al-Jahdali, 2023). Crystallized intelligence remains the most resilient after sleep deprivation.

#networks. We discuss sleep deprivation as a neurobiological network. In the second stage, the DMN and FPC oscillate and are frequently interrupted. As sleep deprivation prolongs we notice the signals between network nodes ignore priors, become increasingly noisy, arrive unstructured and leads to the network destabilizing.

At the genetic level, humans carry variable-number tandem repeats (VNTRs) in PER3, with the most common genotypes being four repeats (PER3 4/4) and five repeats (PER3 5/5). PER3 5/5 individuals show greater slow-wave activity and slow-wave energy during sleep, indicating steeper accumulation and faster dissipation of homeostatic sleep pressures. Importantly, PER3 5/5 carriers face greater cognitive impairment during sleep deprivation, particularly in the early morning trough, despite having similar baseline performance to PER3 4/4 individuals, often linked to evening preferences (Viola et al., 2007). This pattern suggests that PER3^5/5 individuals have a "faster" homeostatic process: their sleep debt builds more steeply during wakefulness and clears more efficiently during sleep, but this also means they have less tolerance for extended wakefulness.

Likewise, the ADA22-A allele reduces extracellular adenosine accumulation. Lower adenosine release delays the onset of severe sleepiness, sustains subjective alertness longer, and reduces micro-arousals during sleep—indirectly increasing sleep efficiency (Rétey et al., 2005). Individuals carrying this allele report feeling less tired at equivalent homeostatic loads and show better maintained vigilance on objective tasks. This creates a virtuous cycle: efficient sleep leaves less residual homeostatic pressure at wake onset, permitting longer high-performance wakefulness, which in turn increases slow-wave pressure during the next sleep bout, further enhancing efficiency.

Finally, ecological niches describe the set of environmental conditions and resource requirements that allow a phenotype to persist. In hunter-gatherer societies, social vigilance and infant care formed different ecological niches, imposing disruptive selection on chronotypes; Modern human societies contain a rich diversity of cognitive and behavioral niches, created by the division of labor and technical complexity. Some occupations demand sustained vigilance (air traffic controllers, emergency physicians, security analysts); others prioritize creativity and problem-solving (research scientists, designers, entrepreneurs); still others require crystallized expertise and procedural skill (surgeons, pilots, software engineers). Night owls gravitate toward careers with flexible schedules, remote work, or evening-shifted hours; morning larks toward traditional office jobs with early starts. Many ecoservices are available on demand and the majority of modern humans are “lions” may indicate stabilizing selection on chronotypes.

#naturalselection. We speculate that hunter-gatherer societies may have disruptive selection followed by stabilizing selection in chronotypes. Similar selection forces among the EA community leads to less chronotype variance within the effective altruist population than the general population.

Occupation also shapes allostatic load along with chronic environmental factors: socioeconomic status, chronic noise exposure, access to green space, diet quality, and chronic psychosocial stress. Individuals living in quiet, low-stress, high-SES environments accumulate less baseline allostatic load, freeing metabolic reserves that can be deployed to cope with acute stressors like sleep deprivation.

With the three factors above, individuals with fast and efficient adenosine clock cycles, low allostatic load, and adaptive ecological niche constructions will be the most resistant to sleep deprivation.

Our hypothesis expects the EA community to have higher-than-population frequencies of PER3 4/4 and ADA22-A alleles due to self-selection into high-intensity, time-pressured work. While PER3 4/4 is genetically vulnerable, performance is mitigatable via niche construction: Teams that pre-commit to mandatory nap breaks for all members will be more effective on code quality than teams where naps are not a norm. We expect that in a 36-hour interventional study in the form of an AI safety hackathon, the vigilance decline will be steeper in genetically resilient individuals assigned to an eco-naive niche than in genetically resilient individuals in eco-smart niches by a significant amount.

#complexcausality. We discuss the genetic causes of having PER3 4/4 genes, ADA-22A allele present, becoming an indirect source of sleep quality improvement and hence sending the individual into a virtuous cycle. Additionally, we also think this study will contribute to an open data set that has second order effects inspiring replication studies, as well as indirectly contribute to AI safety and mental wellbeing.

The interventional study will include participants who (1) self-identify as EA-adjacent and (2) have coding abilities sufficient to attend a hackathon. We exclude any participants with (1) severe sleep deprivation 3 days prior to the hackathon, (2) recent trans-meridian travels, or (3) have night-shift work.

#interventionalstudy. The independent, dependent and controlled variables are discussed with two treatment conditions. The study has control, attempts to minimize extra cognitive load by mimicking the natural state of hackathons by having tracks and also performing tests during meal time.

Independent variables include PER3 genotypes, ADA22-A alleles. These will be gathered from saliva, with consent, to be used with PCR to correlate 4/4, 4/5 or 5/5 on PER3 VNTR and ADA22-A allele presence.

Dependent variables include cognitive attrition measures: Karolinska Sleepiness Scale for subjective state, performance vigilance tasks will be administered at meal times (9:00 am, 12pm, 7pm) to avoid additional cognitive load. We also measure work quality, as tracked by number of git commits per hour, quality of commit messages and code reviews analyzed post-event from the team’s public repo.

We establish two separate tracks. The treatment track imposes team-wide mandatory napping of 20 minutes beyond 00:00, and coffee beans with caffeine microdoses to flatten adenosine rebound. Dynamic lighting with distinct shades during 09:00–12:00, 18:00–21:00 and 00:00–06:00 is offered to mimic natural light. A second track allows participants to do whatever they want, referred to as eco-naive. Participants are prohibited from using any unprescribed amphetamines, modafinil or sleep peptides during the event. Mentors will circulate on site to verify rule-following.

We aim to recruit 270 participants, as hackathons have sign-up and submission attrition rates to satisfy a sample size of 30 per track.

#testability. The hypothesis is testable via psychometrics (vigilance task and karolinska sleepiness scale), code quality (commits and code reviews) and cluster correlation to performance is confirmable by PCR analyses after the hackathon.

We will sort the psychometric data, objective output and saliva samples into heatmaps. We expect to observe several different profiles:

Within eco-smart environments, genetically resilient individuals with ADA22-A and PER3 4/4 should dominate. We would observe a shallow decline in KSS and PVT as well as code quality. Following, genetically vulnerable individuals in eco-smart conditions should match or exceed resilient individuals. This resonantes with popular sentiment that conscientiousness drives the institutional engine more than raw genetic intelligence.

#phenotypicplasticity. We discuss the multiple variants of the PER3 gene (4/4, 4/5 and 5/5) and how genetically vulnerable individuals could be regulated by constructed eco-smart niches to perform better in hackathons and vice versa.

The hackathon provides "ecosystem services" analogous to natural habitats: provisioning (food, water, workspace, wifi), regulating (lighting, napping pods, coffee), and cultural (mentorship, peers, education). Just as organisms in resource-rich environments express broader phenotypic ranges, participants with access to eco-smart infrastructure can exploit cognitive strategies unavailable in resource-poor settings.

#nicheconstruction. We compare the hackathon, its two tracks as eco-smart and naive niches, elaborating on provisioning, regulating and cultural elements, and also shapes participants’ psychometric, real world outputs. We also frame ancient societies as niches that demand diverse schedules while contemporary moldable environments are adaptive.

Project start date: Jan 1, 2026. End Date: May 30, 2026

This study indirectly contributes to the cause of AI safety by understanding the mental clarity of researchers especially when peptides are becoming an increasingly chosen solution. Mental wellness is classified as secondary priorities by effective altruism.

In the best case, this generates replication attempts at hackathons and follow-up studies. In the average case, incoming EAs could copy the PCR techniques to self-screen for their chronotypes and take relevant steps followed in the paper. In the worst case: null or insignificant result, however, this study becomes recognized within the EA community. In all cases, the deliverable could be a replicable paper and an open dataset that lets EA quantify any sleep ROI beyond individual anecdotes within its community. This is comparable to the annual surveys of Astral Codex Ten and LessWrong with data that could be analyzed broadly by the community.

Alexander, S. (2025, January 29). ACX survey results 2025. Astral Codex Ten. https://www.astralcodexten.com/p/acx-survey-results-2025

Borbély, A. (2022). The two-process model of sleep regulation: Beginnings and outlook. Journal of Sleep Research, 31(4), e13598. https://doi.org/10.1111/jsr.13598

Durmer, J. S., & Dinges, D. F. (2005). Neurocognitive consequences of sleep deprivation. Seminars in Neurology, 25(1), 117–129. https://doi.org/10.1055/s-2005-867080

Guzey, A. The effects on cognition of sleeping 4 hours per night for 12-14 days: A pre-registered self-experiment—Guzey.com.

Khan, M. A., & Al-Jahdali, H. (2023). The consequences of sleep deprivation on cognitive performance. Neurosciences (Riyadh, Saudi Arabia), 28(2), 91–99. https://doi.org/10.17712/nsj.2023.2.20220108

Long-Term future fund. (n.d.). Effective Altruism Funds. Retrieved December 20, 2025, from https://funds.effectivealtruism.org/funds/far-future

Rétey, J. V., Adam, M., Honegger, E., Khatami, R., Luhmann, U. F. O., Jung, H. H., Berger, W., & Landolt, H.-P. (2005). A functional genetic variation of adenosine deaminase affects the duration and intensity of deep sleep in humans. Proceedings of the National Academy of Sciences of the United States of America, 102(43), 15676–15681. https://doi.org/10.1073/pnas.0505414102

Viola, A. U., Archer, S. N., James, L. M., Groeger, J. A., Lo, J. C. Y., Skene, D. J., von Schantz, M., & Dijk, D.-J. (2007). PER3 polymorphism predicts sleep structure and waking performance. Current Biology: CB, 17(7), 613–618. https://doi.org/10.1016/j.cub.2007.01.073

Original PDF

#sleepinpublic places I have slept at:

• on the grass
• a park near CCNY
• banana lounge
• aircraft carrier
• vancouver convention centre
• YVR
• arcadia house

• Chapter 8 from Hormone Hunters: The discovery of Adrenaline (highly rated)
• Winners don't use costly punishment
• Why is cognitive effort experienced as costly? — talks about opportunity costs, large information shifts and searching through rare brain states.

A friend and I were playing a Song dynasty simulator and we noticed coal was often but not always the limiting factor. Coal also shows up in:

• 1st industrial revolution. The James Watt steam engines burn coal with a magnificent thermal efficiency of 3%…

• 2nd industrial revolution. Steam engines are stacked with dynamos that convert mechanical energy into electrical energy in Pearl station in Manhattan. A 162x jump from steam engines. So I just placed them on a Kardashev scale. Vibes:

  

• Nuclear. More steam engines, but replace coal with Uranium 235, keep water and add spinning turbines afterwards. The computer system "СКАЛА" — система контроля аппарата Ленинградской Атомной, lit. "rock" — is literally between Scylla (Xe-135 poisoning) and Charybdis (positive void coefficients) 🤔 Xe-135 started to accumulate in the fuel rods and the power decreased to 30 MW. So the control rods were removed in order to have more power, which surged to 200 MW. However at this point, there were so many steam bubbles that the reactor exploded. During the disaster it was estimated that reactor 4 surged to 3e10 W. Modernization, more production for raised standards in quality of life — health, education, welfare.

Prolegomena to Omnipotence.

• The learner; the explainer. "A reader of a popular book is not going to take part in resolving them anyway, while an expert in a given theoretical discipline must first of all get to know its weapons and the configuration of its battlefield before he can himself participate in scientific councils of war, once he has learned the basics of muster and tactics."
• The benevolent observers, featuring the local babushka and the philosopher. "The elderly lady has much life experience, and she advises the girl, on the basis of [statistics] to dump the reckless boy. The philosopher knows the history of science and advises the physicists to give up on their theory because this theory is 'betraying' them."
• The designer, distinct from the babushka: "he becomes involved in action — as have the physicists… He is not a narrow pragmatist, like a builder who is constructing his house from bricks, uninterested in where these bricks came from and what they are, as long as the house gets built. The designer [knows and is interested about their] bricks — except for what they look like when no one is looking at them. He knows that properties belong to situations, not things."

30 years of dolphin research in shark bay! Dolphins form duos/trios, first order and second order alliances, partially because they have no consistent social group it is difficult to estimate a stranger dolphin's reputation.

Ontology of Childhood. Protagonist grows up, and slowly realizes they're in a prison. "But so what? The fact is that the world was not invented by people — no matter how clever they try to be, they are not capable of making the life of the last convict even slightly different from the life of the head of the supply unit. And what difference does the reason make, if the happiness produced by souls is identical? There is a quota of happiness allocated to a person in life, and whatever happens, this happiness cannot be taken away."

Uhlenbeck interview; When math branched from physics in history and came together. Theory-builders, thinking in linear guaranteed steps collaborating with intuitive people who reverse fill in the blanks backtracking in textbooks to understand lemmas. Powerspace and the bulk theorem!

I gladly accepted, planning to make several stops in Europe before going to Cambridge, as I'd also been invited to Paris, Rome, and Finland, where I'd be giving a talk at the International Congress of Mathematicians in Helsinki. Travel was difficult, however, because the British Consulate had recently taken my Hong Kong resident card, maintaining that I could not keep it now that I had a U.S. green card. In the process, I had become stateless. I was no longer a citizen of any country, although I was a legal resident of the United States. For this period of time, until I became a U.S. citizen in 1990, I was literally a man without a state, stuck between two countries and between two cultures. Traveling abroad, as a result, became a real nuisance. I had to use my "white card" to apply in advance to leave the United States, and if I did not follow the proper steps, I would not be allowed back in.

I could not secure a visa to get into Italy, so I had to cross Rome off my list this time, even though I paid an extra "fee" to the Italian consul, who said he'd take care of it (and didn't). The same thing happened the next time I was invited to Italy; I again paid the consul an extra fee, but still got no visa. On another occasion, Michael Atiyah invited me to give a talk in Wales before the London Mathematical Society. When I showed my white card at immigration control in London, they gave me a hard time.

"What is the purpose of your visit in the United Kingdom?" I was asked. I told them I was there for tourism. "Where do you plan to go?" I was then asked. Wales, I replied. "Why are you going to Wales," the border agent persisted, "since it's clearly not a good place for tourism?" The questions finally ceased when I explained that I was going to Wales with my good friend Nigel Hitchin, an esteemed professor at Oxford. That did the trick this time, though traveling with a white card caused me no shortage of headaches.

Spartacist uprising and the bottom-up creation of the Twelve Tables in Roman empire; No Spartacist uprisings in Spring and Autumn periods, evolved from morality to law creating top-down 《法经》。 崤山大战 where some people escape on a boat. Qin official brought forth a yellow horse, but ran just as well as the black horses. Archery competitions to show social status. Reminds me of duck hunting in Shogun Era. Vietnamese History "a princess fell in love with a foreign king, spilled national secrets, and country fell to China" mythos in 2nd century BCE. French colonization. Japanese colonization. The red scare. Vietnamese tunnels — do those residents not get insomnia — and snake tubes used in Vietnam war. Triumph of the villages networks. Credits to Minerva classmates.

East Japan Earthquake. Failure modes being normalcy bias, sometimes people think they're okay and don't evacuate. Other times sensors malfunctioned and predicted a 3 meter wave rather than a 10 meter one. Encountering a magnitude 6.8 Earthquake while studying Earthquakes is dizzyfying.

Introduction chapter from China the industrial revolution book is nice, goes from rational self-interest theories to industrial revolutions to Soviet technocracy; People in control were technical. "there is no such thing as ethos over physics in an industrial setting."

Grothendieck exile. This translation is better "looking at things with your own eyes". Shifting between masculine and feminine selves (2.18 The child and the Mother from a more rambley translation that keeps the gender discussion). Ovid invents Iphis, born as a girl, becomes a boy, marries the woman he loves. Remember? he's the not-so-epic exile in history. Christine de Pisan's dysphoria speedrun and reinterprets the story as pro-woman in the Book of the City of Ladies. Other writers don't endorse androgyny as fusion.

Deformable mirrors.

We visited science agora (X thread).

an electricity generator display

"Takeda cleanroom, cleanest air within the Yamanote line" doesn't this description read like a meme? Ja papers about Insect radio. pH strips but with photos and machine learning. Open source, local fluorescent microscope. Flexible ECoG array to avoid squishing your brain, etc. Japanese science builds on top of a basis with fine-grained novelty which seems to be established by American rough prototypes that have high impact directionally.

So on Fri, Xu Zheying was talking about Qwen models having the best stock investing abilities 🥳 and lots and lots of new exciting audio and images from Wan 2.5 flashed on the screen. There's no logical connection in his talk. I thought, Chinese tech seems to love thinking shortcuts as long as there is market validation, is uneven depth, and then jumps to scale.

Since both countries are researching neurotech, it will be interesting to see how the structures and methodologies differ; The Japanese government funds "Moonshot Internet of Brains" (timeline 2050) and there's International Center for Primate Brain Research in China that is focused on mapping.

German neurotech programs would be fun to compare too: Last month I noticed German science is often presented with less narratives but as is with a progression from first principles — Thinking about the "On Quantum Mechanics" paper by Max Born in 1924.

Curious to how people react when an experiment yields null results and how to deal with their latent learning progress? In startups, someone may be "not qualified" to join a team because they need "a lot of training". What is too much training in science. This reminds me, supervision often is related to happiness.

Quantum computing is often motivated as a natural progression from physics. See, problem-centric motivation for QC sounds unnatural: Need better drugs. Need more secure cryptography — argument not exclusive.

Compared to fusion, which is often more problem-centric.

Common fusion reasoning: We're building AI. Need more data centres. (Alternatively: trying to ascend the Kardashev scale). More energy demands. So, invest in fusion.

I wonder if due to this nature of motivation, there are more Chinese people working on fusion compared to QC, and the other way around in Germany.

Crypto is having its time in Japan! There are not a lot of public centralized tech events. There was the liquid AI hackathon, fully connected by W&B, robotics workshop, etc. I could find one or two good irl ones per week in Tokyo, though, need to learn more ja to understand the lectures — See Humanism in AI lecture, Sakana AI defense study events.

This paper talks about community and document-level signals. It says

Community-level. Strong papers are highly connected to the ecosystem. So, a heuristic would be to build co-occurrence graph from a paper's references, then, calculate connectivity and clustering coeffs. While playing around with this, one blocker was how existing citation layers like Semantic scholar or Open Alex restrict how fresh an input paper could be, and then, only shows related papers by their own similarity metrics rather than citations.

Document level. Sometimes, a paper, or even an article, may be full of jargon! Do the authors compress knowledgably or obfuscate? Given the German style of science, exposition clarity is not central to research quality! We may be able to check if authors know their stuff by if the text matches the ground truth in the ecosystem. Well hold up, if a field has no goals there is no ground truth … cope measurement should stay separate from narratives.

Also, detect red flags like tortured phrases, question marks. Don't forget about green flags.

Claude asks if I'm translating from Chinese as my prompts are too paratactic, just simple phrases piling up without conjunctions. I haven't thought about it, so I went on a rabbit hole. First I wrote every kindness a category error 1/2 in parataxis and in hypotaxis. e.g. "Garden roof, a fresnel plate."

Then I had a 30 min break where I binge read 阿Q正传. More zh-cn bookstores than en here anyway. The story set in the 1911 Xinhai revolution — ended the Qing dynasty — describes bay area psychological horror and most people having anxiety problems.

阿Q is a labour drifting between jobs, housing; He loves gambling and opium. Read: software engineers drifting between jobs, tryna find a group house in SF who also happen to pump memecoins and psychedelics.

阿Q's neither fully a villager or a townie, yet he hates both '条凳', '葱丝大头鱼' used by townies or '长凳' and '葱叶大头鱼' as endorsed by villagers. Read: "Not really a rationalist. Not entirely a biologist."

Similarities: Four or five distinct microtribes, the constantly high cortisol levels are similar. Differences: Present day, present time… And feedback loops are faster.

阿Q is determined to draw the best circle in his life, fails, ragequits and now believes only losers draw good circles. His 精神胜利法 maps to a reactively agency moment.

赵庄赵太爷 ~= an evil group house that welcomes people to crash with spare bean bags.

假洋鬼子: Soham, or, whoever is struggling with performance while overemployed, updating their resume headlines to match the specific job.

Just look at these epic paratactic sentences:

[subject + verb phrase A], [s+vp b] … [topic A], [topic B], no pronouns (how convenient) and no explicit links.

Su Shi asserts even at 45, his exile hasn't debilitated him. 2nd favourite is 丑奴儿·书博山道中壁 — Young people write melancholic sentences from void of experience, older people just haha, yeah, what a cool autumn.

Why do I talk about cultural observations so much? (1) modeling people of different cultures is an interesting thought exercise (2) mythology natively models the world micro/meso/macro without borrowing from STEM fields, which have massive blindspots and strikes non-STEM folks as inflexible. To be clear, I'm not excluding STEM models, but pick the right tool for the task. (3) belief systems that can stand time and cultural tests.

So, school: readjusting is okay now! It definitely seems like a future self made the decision and I'm coming to terms with it.

a ceramics workshop at Kogei SHS

Source: feralscholars.substack.com/p/september-readlog — noesis

Major update: I have moved to Tokyo and resumed university! There's all of a sudden time to just read for no reason and (thankfully) no more death threats. A huge thank you to all enthusiastic nerds who have funded my passion projects this month!

So here is a readlog! It contains: Flowers for Algernon. Medea commentary. Moving to Tokyo. Chinese history commentary. Quantum chemistry.

My rant on Goodreads:

Charlie's intelligence speedrun seemed like a direct manifestation of psychometrics and 1950s stage psychology potpourri!

First up, we have WISC! Consisting of: verbal intelligence — split into grammar and vocabulary, spatial intelligence — Charlie could see things with his mind's eye, working memory, as well as processing speed (not emphasized). Additionally we also had developments in interpersonal intelligence, intrapersonal intelligence, self awareness and regulation and so on.

Stage psychology theories were popular around the time of writing — 1959. It was funny to read it from 2025, when they are out of fashion!

There's Kohlberg's moral stages: When the pregnant woman was propositioning to Charlie in the park, he instinctively called her a degenerate (pre-conventional morality). The next time was when he saw a dull boy being bullied in a restaurant. Charlie subconsciously used the Golden rule (post-conventional morality). Later he also argued for his own rights in addition to means to an end as an experimental animal. Kohlberg developed his theories of morality in the late 1950s.

Look, we also have Piaget's stage theory:

Sensorimotor (seen as Charlie regresses near the end)

Pre-operational. Especially egocentrism as Charlie flips between first and third person for the first time.

Concrete operational. Reasoning about himself and the mouse, Algernon.

Formal operational... This is brutal

Alice rages at Charlie: "All I can do is listen and nod my head and pretend I understand all about cultural variants, and neo-Boulean [sic] mathematics, and post-symbolic logic, and I feel more and more stupid..."

Charlie, criticizing other academics. "Physics: nothing beyond the quantum theory of fields. Geology: nothing about geomorphology or stratigraphy or even petrology. Nothing about the micro- or macroeconomic theory. Little in mathematics beyond the elementary level of calculus of variations, and nothing at all about Banach algebra or Riemannian manifolds. It was the first inkling of the revelations that were in store for me this weekend. I couldn't stay at the party. I slipped away to walk and think this out. Frauds — both of them."

Minor note, I was expecting to see more of Dabrowski's overexcitabilities which were also developed in the late 1950s but they weren't explicitly applied. That was when the label "exceptionally gifted" was coined. I feel the author didn't seem to like giftedness labels. "[Dr Strauss] says [psychology books] will confuse me and make me think about psychological theories instead of about my own ideas and feelings."

Other topics. Charlie dissociated, particularly around women. "Even a feeble minded man wants to be like other men. A child may not know how to feed itself or what to eat, but it knows hunger." Why does the author like dissociation so much? Why do gender dysphoric people particularly love dissociative psychological horror?! Minds of Billy Milligan was also about DID. The way Charlie talked about relationships reminded me of Qiaochu Yuan archetypes, and I won't be one of them.

Fay, speaking to Charlie: "And boy did you act queer-I don't mean fairyish or anything like that but strange." Why did the author specifically used the word queer here?

Freudian jokes: "Does an idiot have an id"

I absolutely adored the narrative quality, however, considering it's 2025, I'm disappointed by the author's imagination. Just when the Charlie's intelligence approached near author levels, Charlie… regressed! To become intelligent and deteriorate again is like vector operation. Even I felt vicarious frustration when Charlie explains to his mom the intelligence changes. It definitely seemed like pre-planned rather than hypothesis-driven sci-fi, right?! "Algernon Gordon hypothesis" says you can artificially inflate intelligence but it must fall down later. If intellectual arc loses it's spark, then I'd expect emotional arc to be much stronger. There was a nice tripartite soul-like balance throughout. However, the falling action was a bit abrupt. But overall, highly rated.

• Consciousness club Tokyo — Max Hodak talked about consciousness for 1 hour. I'm not really sure I follow this reasoning especially intermediary steps.

  

  There are some interesting papers:

  • Monkeys could use 200 neurons to represent an average face
  • The brain has a lot of loops
  • Max defines a moment as a snapshot where you can't order things temporally

• Harajuku area — I appeared as Yamikawaii!

• BioClub Tokyo — Bioenthusiasts can run experiments there for free!

• Localhost Japan

Watching the earliest 夏，商，周 dynasties with dearest Adelaide!

Phrases I took note of:

三年不飞，飞将冲天，三年不鸣，鸣将惊人

i.e. If you don't fly for three years you're gonna soar! If you don't chirp in three years, everyone will be surprised!

It seems like in ancient China three years of silence is associated with brewing greatness! This contrasts a lot with the build in public culture on X, but arguably this is just stealth mode in ancient China.

春秋后期，社会上有这样一批人，他们有着高贵的世系、受过教育，可以被称作贵族，但却很贫穷，实际的社会处境和普通平民差不多，这样的破落贵族后裔组成了贫困的士阶层

i.e. In late Spring and Autumn period, there was a group of highly educated people "士阶层" with noble lineage, but these people are broke (do they resemble broke grad students?) Yet in two centuries, many Shi became high status.

伊尹 (Yi Yin), the legend that originated from a kitchen. 伊尹 was the primary intellectual behind overthrowing the Xia dynasty ruled by 夏桀 (Xia Jie). Xia dynasty is the first recorded dynasty which also engaged in slavery. 伊尹 was discovered twice in a kitchen which reminds me a bit of Eigenlucy.

Now I can complain in Chinese. I can just steal some Chinese literary containers to express my deep dissatisfaction, whereas, there's a lot more sarcasm in English before expressing anything!

Nowadays, many talented engineers and friends are being hurled out of the United States. But being exiled is the commonest thing in history, there's no need to panic. I looked at epic 10 exiles, calculated distance and clustered them.

One interesting cluster was Su Shi (苏轼) and Dante. Su Shi was exiled for writing satirical poetry about the ruler Wang Anshi's new policies. He complained how loans and state monopolies on salt superficially, helped farmers, but were practically abused by local officials to squeeze more revenue. He spent 12 years in exile and produced ~800 poems during the time. Dante was also a writer-activist type that was reported for corruption, exiled and wrote the Divine Comedy. Their works circulated among thin, regional networks and were memetic in themselves, and probably sounded cooler due to their status.

For Aristotle, he died 1 year after exile, there's not much original stuff. He just focused on docs, and having strong students extended his reputation for millenia.

The most disappointing exile was Ovid. During Augustus' moral reform encouraging righteous marriages, he wrote a … seduction handbook. And after he was exiled to the middle of nowhere Ovid continued to write laments. My guy! This doesn't really work!

Theme: talk is cheap

So now, am I a tiger, Scylla? — Hurl at me whatnames you please I I've reached your heart; and that is right. But Scylla still wins."

Creon, exiling Medea: "You're wasting words; you'll never make me change my mind!"

The janitor from the modern version of Medea: "She say, trouble with you Marta, you think words is worth something. But words ain't worth a thing. Words is just women's money, that's all. Only thing it buys you is scraps from the man's table."

Cynic's not-really-wisdom

During the argument Jason ironically expresses "I am sane now; but I was mad before". This sounds just like false cynic wisdom. When cynics pick us apart, we (especially young people!!) can mistake corrosion with care ... seek to be affirmed and, all of a sudden go up in flames! Am I falling am I sinking what the hell have I been thinking! Modern Medea rewrite pokes at this: "All that anger disguised as if wit was like sugar-coated hate."

Depressing quotes

Jason is a prisoner in a princess's bed; And Medea is in her room Melting her life away in tears; No word from any friend can give her comfort… She will not eat; she lies collapsed in agony, Dissolving the long hours in tears.

The dark cloud of her lamentations Is just beginning. Soon, I know, It will burst aflame as her anger rises. Deep in passion and unrelenting, What will she do now, stung with insult?

Medea gazing at her children:

Strong reason forces me to weep, old friend. The gods, And my own evil-hearted plots, have led to this. Dear sons, why are you staring at me so? You smile At me — your last smile: why? Myself? I won't do it. I won't think of it again. What is the matter with me? Are my enemies. To laugh at me? Am I to let them offscot free? I must steel myself to it. What a coward I am, Even tempting my own resolution with soft talk. Boys, go indoors.

I learned about some chemistry theories from the 20th century and ran an approximation technique on a few molecules. Namely, Hartree-Fock, which calculates molecular orbitals by treating all electron movements as one, averaged out electron, and also assume the nuclei being still. Then the computer tries to solve a 1-electron Schrodinger equation. They collapse the infinite dimensions onto a finite basis set, rewrite equation as a big sparse matrix and find some eigenvalues.

• Everybody has vertical and horizontal traits of their parents (paraphrased) by Lionel Wong
• 19 failed experiments by Breaking Taps
• arXiv RAG by me! quick 3-day hack I made for querying papers. Not deployed properly, so you can't. use. it.

"clustering coefficient" is a highly active user in the Signal database with 922 messages spanning from July 2024 to March 2025. Their communication patterns reveal a multifaceted individual with diverse interests centered around neuroscience, genetics, AI, and social networking.

• Total Messages: 922
• Time Period: July 2024 - March 2025
• Peak Activity Months: February 2025 (264), August 2024 (208), December 2024 (198)
• Most Active Day: Sunday (191 messages)
• Peak Hours: 11 PM - 4 AM and 6 PM - 10 PM
• Average Message Length: 74 characters
• Median Message Length: 44 characters
• Question Frequency: 24% (224 messages)

```mermaid graph TD CC[clustering coefficient] --> Neuro[Neuroscience & Neurotechnology] CC --> Genetics[Genetics & Genomics] CC --> AI[AI & Language Models] CC --> Health[Health Optimization] CC --> Social[Social Networking]

Neuro --> EEG[EEG Technology] Neuro --> BCI[Brain-Computer Interfaces] Neuro --> Psychedelics[Psychedelics Research]

Genetics --> NRXN1[NRXN1 Variants] Genetics --> GWAS[GWAS Brain Regions] Genetics --> FOXO3[FOXO3 Longevity Variants]

AI --> Alignment[AI Alignment] AI --> Claude[Claude's 'Moral Preferences'] AI --> ChatGPT[ChatGPT Experiments]

Health --> Microplastics[Microplastic Concerns] Health --> Longevity[Repetitive Element Profiles] Health --> Rapamycin[Rapamycin Research]

Social --> Boston[Boston/Cambridge Hub] Social --> SF[San Francisco Hub] Social --> NYC[New York City Hub]

Autism[Autism as Framework] --> Genetics Autism --> AI Autism --> Social

Institutions[Academic Institutions] --> Harvard[Harvard Resources] Institutions --> MIT[MIT Resources] Institutions --> Journals[Journal Access]

CC --> Institutions CC --> Autism ```

1. "8458210232"
2. "OH WOW THE HACKATHONS ARE SO SMALL AND FUN THIS WEEK"
3. "PARROT-ON-DMT"
4. "TELL ME"
5. "DRUMHELLER MALLARD"
6. "840170"
7. "MINIMAL MICROPLASTIC FLOSS"
8. "ALOK SINGH"
9. "WITH JOSE AND JACOB"
10. "HE TOOK ALEX Z WITH HIM"
11. "CATS ARE INQUILINES"
12. "ITERATED DISILLATION AND AMPLIFICATION OF MICROPLASTIC NOISE TOKENS"
13. "LOL IN THE HOMM3 PLAYTHROUGH JIMMY KOPPEL WINS THE GAME"
14. "GUYS WHAT ARE THE NAMES OF UR CHATGPT FOLDERS"
15. "YES AND FUCK THE VOTERS FOR NOT READING OR CARING"
16. "THEY KNEW WHAT WAS COMING FOR THEM"
17. "OMG FORUMS ARE STILL THE BEST"
18. "CARL BERGSTROM IS A HUGE LEWONTIN FAN"

Their messages contain references that map their intellectual influences and affiliations:

"CARL BERGSTROM IS A HUGE LEWONTIN FAN"

This reference to Carl Bergstrom (complexity theorist and author of "Calling Bullshit") and Richard Lewontin (evolutionary biologist known for critiquing genetic determinism) places them in a specific intellectual tradition that values systems thinking and critical approaches to genetic reductionism.

"not that ADHD-associated PGS explain anything b/c a lot of ADHD is trauma or qiaochu-yuan-syndrome"

The reference to "qiaochu-yuan-syndrome" (likely referring to mathematician and rationalist blogger Qiaochu Yuan) suggests familiarity with rationalist community figures and concepts.

"openai research does a pretty good job of guessing every major person's DnD alignment.."

This casual reference to both OpenAI research and Dungeons & Dragons alignment systems reveals comfort moving between technical AI research and gaming/internet culture frameworks.

"clustering coefficient" emerges as someone whose cognitive style mirrors their username—a mind that naturally identifies and strengthens connections between disparate nodes in a knowledge network. Their messages reveal:

1. Disciplinary Boundary Crossing: Moving fluidly between neuroscience, genetics, AI, philosophy, and social coordination

2. Conceptual Pattern Recognition: Identifying similar structures across different domains (e.g., seeing autism as a pattern of "extremes" that applies across personality dimensions)

3. Social Network Weaving: Actively connecting people based on perceived intellectual complementarity

4. Health-Optimization Focus: Consistent concern with longevity, microplastics, and optimizing cognitive function

5. Institutional Navigation: Sophisticated understanding of how to access and leverage academic resources

Their communication style—characterized by brief, direct messages, strategic capitalization, and frequent questioning—serves their function as an information node in a complex social network. The name "clustering coefficient" aptly describes not just their role in connecting others, but their own cognitive tendency to identify and strengthen clusters of related concepts across traditional boundaries.

Based on message content analysis, "clustering coefficient" likely has strong research interests in:

1. Neurogenetics: The intersection of genetic variants and brain function
2. Cognitive Diversity: Particularly autism spectrum conditions and their relationship to information processing
3. Longevity Science: Especially epigenetic factors and environmental influences
4. Network Theory: Both in social and information contexts
5. AI Alignment: Particularly the philosophical and ethical dimensions
6. Microplastic Contamination: Environmental health impacts and mitigation strategies
7. Psychedelics Research: Neurobiological mechanisms and applications

"clustering coefficient" appears to function as:

1. Information Broker: Sharing specialized knowledge across community boundaries
2. Social Connector: Actively introducing people with complementary interests
3. Resource Navigator: Helping others access institutional resources
4. Event Curator: Organizing gatherings around intellectual interests
5. Conceptual Translator: Moving ideas between different intellectual frameworks

This multifaceted role makes them a valuable node in the community network, facilitating both information flow and social connection across traditional boundaries.

After 3 months in Tokyo, I relocated to near the Stanford campus. Featuring material access, seminars, furries, rationality-subgraph (eh), quotes

Importance of post: 1/5

2025 was full of startup hype about neuron computers. At Frontier Tower's neurotech floor, friends were discussing biological neuronal responses to nootropics and neuropeptides. However, the neuronal headstages were hard to get. Student clubs e.g. UC Berkeley didn't rent them out, nor did neurotech startups e.g. Science.xyz.

For reference, a single microelectrode array of 59 channels from multichannel systems costs $300.

Not to mention, integrated neuronal headstage setups from Axion biosystems cost $10k - $50k.

Thus, we experimented with building a cheap MEA, and so far graphene version costs <$1 and 5.5 min per unit.

1. Laser-induced graphene
2. Aluminum vapour deposition
3. Copper metalization We've looked into laser ablation: Indium tin oxide. This was not tested but may also be worth a shot.

Iteration 1: OnShape. Created at ~12 am from circles and linear patterns. I found out this shortcircuited, since I naively traced the well as a ring, so I created version 2!

Iteration 2. The improved MEA in KiCad.

This dotted grid directly come into contact with neurons, and extend out to the peripherals.

Iteration 3. Using existing MEA design files provided by Mind-in-Vitro, openable in Klayout.

At atopile's machine shop, we directed a laser beam at the polyimide tape, for graphene to become deposited on the surface. We are using microscope slides from frontier tower's biopunk lab run by Elliot Roth. Techniques were adapted from this porous graphene paper

The laser’s energy causes lattice vibrations within the polyimide: Under high surface temperatures of >2 500 °C, C=O, C–N, and C–O bonds are broken, freeing non-carbon gases from this tape. The carbon atoms rearrange into graphene with its hexagonal structure.

We tuned a CO2 laser operating at 10 600 nm to 5.5W with a scanning rate of 88 mm/s with 100 lines/cm and directed the beam at the kapton tape.

The centre-to-centre distance between electrodes: 0.15 cm, with a 0.25 cm electrode radius. The trace width: 0.2 mm.

A hurdle is creating continuously conducting traces.

Trace width microadjustments

Given that SVGs are a set of lines defined using the XML format, we just made them thicker. However, in practice, laser software, e.g. Xtool or Lightburn, treat all lines as infinitely thin.

We then tried blowing up the JPGs, which are bitmaps; However, a change from 0.2 mm to 0.3 mm in the JPG led to a fuzzy visual appearance and shortcircuited. A lazy hack we found was just to duplicating the traces three times in the laser software from Xtool. Additionally, we varied the scan rate (+/- 20 mm/s) and lines per cm (max at 200, min at 50) and set the laser scanning to be monodirectional.

Examining raw trail conductivity

The triple-layered electrode version was consistently conductive. However, the graphene easily wears off, or encounters insulation at corners.

The next step would be more functional testing (saline) and compare to the standard Utah array.

While implementing laser-induced graphene, we noticed dense metallic stripes appeared on glass--This certainly didn’t look like graphene. So we peeled off the kapton tape, put the glass slide into the laser chamber and turned the laser on again. This was in fact, the vapour from the underlying sheet of aluminum which had been sitting inside the glass in a diode-pumped solid state laser (DPSS) of 1064 nm (physical vapour deposition).

Review: In the traditional electroplating process, an object, the cathode, becomes coated with a thin layer of metal.

The cathode and the anode are initially submerged in ionic solution and a circuit is completed. When the battery applies its gradient, two processes happen:

1. Chemical. The +ve ions start migrating to the cathode to coat it
2. Electrical. The cathode "craves" electrons and the electrons move to the cathode. Simplified copper reaction: Cu²⁺ + 2 e⁻ → Cu(s).

In copper plating, a focused laser beam mimicks the voltage differential like a battery. As it is directed at the copper solution, it causes a thermal gradient: Heated and unheated surfaces. The heated area gains copper under the plate.

Results of the deposition

Preparing the electroplating solution

Add water first, and then add chemicals slowly from least reactive to most. i.e. Add distilled water, followed by alcohol, glycerin, NaOH and then CuSO4.

We placed the solvent on a hot plate, with a magnetic rod for stirring.

The biggest challenge here was when the scale started tweaking. We became suspicious when the 100 mL of water weighed 125 grams. Consequentially, in our 2nd copper replication attempt, the copper solution instantly precipitated into copper hydroxide--a cloudy, turquoise colour, not what we want! The ideal solution should look transparent, dark blue.

We tried a series of quick troubleshooting e.g. switching power supplies, using different scales, but there weren’t matching power supplies of 9V for any other scale in the lab.

Thus, I poured in some ethanol (1 mL of ethanol was 0.789 grams) to see how much the scale was off by--approximately 29.5875 grams. From then on, we performed everything according to the scale’s offset mass (1.25x). This method has been time-intensive and prone to error.

Motivation. Researchers may use red and green fluorescent proteins in their living neurons which includes filtering colours. So the yellow polyimide tape may present dark under a microscope. They have a perfect reason to want transparent MEAs?

Here are some materials to use (loosely inspired by this paper)

• Indium tin oxide coated borosilicate glass as substrate
• Brush-on negative insulator (1002F epoxy) - To cover the non-conductive regions
• PEDOT:PSS (PH1000 + 5 % DMSO or ethylene-glycol) - Making the surface biocompatible, although ITO is already safe.

Then, the plasma engine could increase the surface energy of glass.

It would be really good to replicate the entire neuronal headstage from Mind-in-Vitro to broaden the accessibility. Please get in touch corpus.logica@gmail.com if you want to support this project such as by offering ephys lab spaces.

Eigenlucy (main collaborator)

Supporters: Morgan Hough, Paul Han, Elliot Roth, Lucas Mair, James Mavo at frontier tower, San Francisco as well as Atopile for letting us use their office.

Funding: 1517 Medici project, merge grant, Lucas Chu and individual donors.

In comparison, North American grade 3 students are expected to be able to comfortably read and write, as well as begin understanding comprehension and analysis.

That being said, Nigerian students in grade 3 have gone through 3 years of education with absolutely no added value. Not only do these students deal with the opportunity cost associated with going to school in the first place, they’re unable to utilize their educations to escape poverty and properly stimulate the economy.

While lack of accessibility to education in Africa is a large and often brought up problem, lack of quality education is just as important. Despite the fact that many children of primary school age simply can’t access education, many of those who have access to education are still unable to become well-educated contributing members of society due to low-quality education in Africa. Educated or uneducated, the education system in African countries has proven to create a low-quality workforce with many of the population trapped in poverty without the ability to escape out of it.

It’s evident that educational quality is incredibly important in creating a valuable workforce, decreasing unemployment rates, improving quality of life for Africans, and escaping poverty. That being said, why is quality education so inaccessible in this continent?

Teachers are the centerpiece of a functioning education system. A teacher’s skill determines the quality of the students’ education, as well as whether going to school is worth it given the opportunity cost associated with doing so. In Africa, even the children who have access to school are gaining little value from their education, primarily due to a lack of teacher training.

Many teachers are often absent, as they don’t have incentive to show up due to the severely low wages. Teachers in Africa often work multiple jobs which take priority over teaching, resulting in a high rate of teacher a. Even those who do consistently show up are improperly trained and lack the qualifications to adequately teach students. In fact, studies in countries such as Lesotho, Mozambique and Uganda have found that fewer than half of teachers could score in the top band on a test designed for 12-year-olds. In addition, there is a high amount of understaffing in schools in Africa.

A major outcome of high quality education is a valuable workforce capable of contributing to the economy and the development of these countries. It also significantly reduced the strain on global healthcare systems, as well as the carbon footprints of families. Families are unable to escape the struggles of poverty without a high-quality education while access to it provides them a stepping stone to escape poverty, a profound issue seen in all major African countries.

The growth in the world’s labor market is in Africa. As other parts of the world begin to age, Africa will grow its population and today’s children will be the talent tomorrow’s global companies will be recruiting. In the next 30 years, it is projected that sub-Saharan Africa’s working-age population will increase more than twofold—accounting for 68 percent of the world’s total growth*.* Economists have shown that when controlling for other factors, increasing girls’ and boys’ years of schooling (and the skills they learn while there) has a positive effect on economic growth. Ensuring girls’ secondary education is particularly impactful. An increased workforce also leads to an increase in incomes, growth in economies of Africa’s trading partner

With a significant lack of funding in the healthcare system in African countries, support for trained teachers and adequate training for teachers simply isn’t accessible. The already inadequate education system makes the supply of trained teachers incredibly low. A lack of proper educators contributes to this low-supply of trained teachers, making it incredibly difficult to stimulate a well-developed education system in Africa.

As well as this, one of the major priorities in Africa is to provide access to an education as opposed to a quality education. This means that insufficient funding is allocated into ensuring that education is of a high quality, that students who are being educated are actually literate, and that curriculums are being updated and thoroughly followed to ensure a standard of education for all students in these countries.

This clear lack of funding and defecit in adequately trained teachers is evidently unable to support the large amount of students who require education. At the end of the day, students choosing to persue an education in Africa, if possible, comes with a significant amount of opportunity cost. With the lack of quality of education, many people of primary school age join the workforce to help provide for their family rather than educate themselves in a system that doesn’t necessarily make them educated and contributing members of society.

This organization divides African students from grades 3 -5 into groups based on learning needs rather than age or grade in elementary school to learn reading, writing and counting. It is currently being implemented in Côte d’Ivoire and Zambia. Since 2001, J-PAL-affiliated researchers have rigorously tested the theory of change underlying the TaRL approach. Through 6 randomized evaluations in India, as well as a growing body of research in Africa, they find that when TaRL is successfully implemented, learning outcomes improve.

Focuses on gender equality, as well as teaching quality and assessment, early and inclusive education and education in crisis. This organization was found in 2002 and operates mainly in Burundi, Guinea-Bissau, Lesotho, Mauritania, Sao Tome and Principe.

In 2020, the Coca-Cola Foundation spent $52,725 in funds for teacher training and support for virtual learning in 200 low-income schools in Nigeria. This initiative was found by Roberto Goizuela in 1984. They have a strong presence in all African countries.

• How to deliver proper cost-efficient training?
• How to offer incentives for teacher attendance?
• How to find more willing teachers?

Rebecca Winthrop - Director for the centre of Universal Education, Senior Fellow at Global Economy and Development. Authored articles for Brookings on improving quality public education in Africa.

Russell Rickford - Expert in African American Education, African-American political culture, history of education and American social movements

Robert S. Peterkin - Expert in education achievement, school reform and African-American cildren, equitable school choice leadership.

The authors extract neurons from the hippocampus of three-day-old rats and store them in a cold dissection medium to maintain tissue viability—from keeping an appropriate salt balance, pH levels, supplying glucose and inhibiting too much glutamate activity that can kill neurons. Then they apply a digestion medium with enzymes (Trypsin, DNase I) to break down connective tissue and isolate individual neurons.

The authors used a 10 x 6 microelectrode array supplied by MultiChannel Systems. An MEA is a 2D grid of electrodes, each capable of capturing the neuron's electrical activity and stimulating it with voltage pulses. A limitation from MCS is how each electrode can only be used as either stimulation or response, but not both. They’re characterized by the centre-to-centre distance between electrodes, ranging from 30 to 200 micrometres. Normally, the grid is surrounded by a ring on a larger board, which is the culture chamber.

Example array from multichannel systems

There is an underlying thermostat to maintain the temperature. One electrode is used as ground.

These researchers added a coating of polyornithine which creates a positively charged surface for subsequent attachments to stick. Note: cells and neurons are negatively charged on the inside with the -70 mV resting potential. But polyornithine alone is not enough, we also need biological cues for neuron connectivity. Thus we apply the BD matrigel on top 20 minutes before seeding the neurons.

100 μL of cell suspension is placed onto the MEA’s electrode area with approximately 80 000 cells/cm^2. Researchers let them sit for 20 minutes at room temperature. Then, we add 1 mL of culture medium to the MEA. Cultures are maintained in the incubator set to 37°C, 5% CO, and 95% humidity. Neuronal cultures older than 25 days usually exhibit robust and reproducible responses.

Supervised learning is used meaning labelled inputs and explicit feedback. This paper provides an L-shaped stimulus and a T-shaped stimulus through electrodes.

The stimulation is low-frequency and short bursts: it alternates between positive and negative with magnitudes of 0.2 V—1V each time (bipolar stimulation), lasting for 100 microseconds at each polarity using the STG1004 Stimulus Generator.

The feedback mechanism is high-frequency and long bursts. There are 40 trains of bipolar pulses +/- 0.9 V lasting for 200 microseconds every 2 seconds. Each train then contains 100 pulses at 250 Hz itself. The long burst means it is a tetanus.

As previously mentioned in Training biological neural cultures: Towards Hebbian Learning, we evaluate if learning occurred by checking for three values: Reaching convergence criteria (e.g. demonstrating the error function is converging), connectivity changes and controlled firing rather than random firing.

• Selectivity of neurons
• Reliability across trials
• Connectivity changes

Spatial maps of the pre and post-tetanus response to see if the activity becomes more like the image filter Controlled firing

Measuring average firing rates before and after applying the tetanus The total firing activity during a time window before and after the tetanus

• use 16 x 16 microelectrode array
• Given that the number of pixels is limited, we change from direct mapping to using Gaussian splatting.
• Predict the stock market

When silicon computers were created, people justified the utility being to calculate, and now we get beautiful phenomena like deep learning. This pattern is common throughout science. By the same analogy I’m guessing living neuron computers have unexpected purposes post-building them. The field is really really early (~10 years), it seems to be a pre-paradigm science.

If it’s used for computation:

Parallel processing: Neurons learn in parallel compared to serial tasks which traditional computers excel at.

Energy efficiency: The human brain processes information with immense efficiency, estimated at 20 watts, compared to supercomputers' megawatt-range consumption.

Data efficiency: Humans learn tasks with far fewer samples (e.g., humans need around 10 samples to learn a simple task, while machines can require millions).

Other applications:

Taste food quality, so I don’t have to worry about getting sick

Sniffing the quality of medicine! and neural-drugs. This seems fun as I’m curious about psychiatry. Medicine is complex and haywire.

Paper: Smirnova, Lena et al. “Organoid intelligence (OI): the new frontier in biocomputing and intelligence-in-a-dish.” Frontiers in Science (2023).

In the 1960s, Alan Turing’s work on machine intelligence and von Neumann’s discussion of computer architecture laid the fundamental idea that computation can be universal. Turing-complete machines can emulate each other’s algorithms. This indirectly raises the question of whether brains could be used as computers.

In the 2000s, researchers started developing bio-engineered tissue models replicating aspects of human organ function. In early studies, Shahaf and Marom reported that cultured neurons could adapt to stimuli: Rat primary cortical neurons could learn as they adapted to show a desired predefined response to low-frequency focal stimuli after a learning curve.

Neurons in closed-loop systems responded to stimulation adaptively and showed signs of learning when separating multiple signal sources. Then, the Potter group trained rat primary cultures to control a small moving device.

The concept of organoid intelligence gradually took shape in 2016 - 2018 as researchers started to recognize the potential of brain organoids to perform basic cognitive functions and name the field.

In 2022, Kagan et al. from Cortical Labs elegantly demonstrated aspects of learning by training human neuronal cultures (iPSCs) to play the simple computer game Pong. There is a feedback loop where the culture was penalized with white noise (uninterpretable input) whenever the controlled paddle missed the ball. They also used electrophysiological measurements to show that learning has occurred in the neurons due to changes in functional plasticity and information entropy. However, there are questions about replicability.

Distinction: BCIs usually interface with whole brains in living organisms, but brain organoids refer to neural tissues in vitro.

With just $2000, Viraj and his team have grown neurons on an MEA and communicated with them.

After running the scrappy approach with purchased rat neurons, we could also grow our own neurons: stem cells → neural tissue → plating.

1. stem cells To reduce ethical controversy, researchers often take cells from the human skin/blood and reprogram them into stem cells (iPSCs), which can go on to differentiate into a variety of cell types including neurons.

Using viral and plasmid methods

Obtain human somatic cells

Producing Yamanaka factors—which are proteins which reset adult cells into stem cells. We can genetically engineer E. coli bacteria to produce these proteins. First, we insert Yamanaka factor genes into a plasmid which then enters the E. coli. The bacteria can grow in a flask or a bioreactor. After a certain period of growth, we break open E. coli cells to extract factors.

Transduce these somatic cells with Yamanaka factors. To deliver genetic materials into cells, we could use lentiviral vectors or protein transduction mechanisms.

Now we have stem cells!

Although having multiple error-prone steps this is extremely cheap. For more details, refer to: Open source brain organoids

Questions we have:

How long can we keep these stem cells alive?

using mRNA transfection

This protocol has reasonable costs under $3000 and high efficiency. It produces around 160 colonies per chip (27 mm²) in 15 days.

We first get a microfluidics chip, which is a container for the stem cells Microfluidic Polymer Chips - microfluidic ChipShop We also start with somatic cells obtainable through a biopsy. In this case, the choice is fibroblasts—connective tissue cells. From days 1 - 8, we add mRNA carrying the reprogramming factors to the chambers on the chip, as well as materials that prevent cell stress and toxicity From days 9 - 15, we stop adding mRNA and instead add a medium that supports the transition from somatic cells to stem cells Finally, we extract hiPSC colonies from the chip using chemical processes or mechanically, using a scalpel under a microscope. Source: Microfluidic reprogramming to pluripotency of human somatic cells

2. neuron differentiation Example well plates, where the stem cells start differentiating. Create a 2d sheet of neurons with the help of Neurogenin-2, taking 2 - 3 weeks Placing the hiPSCs on well plates coated with poly-L-ornithine (PLO) and laminin, which provide a supportive environment for cell growth Use lentiviral vectors to engineer the cells to express neurogenin-2 (Ngn2), a gene that induces neuronal differentiation Activate the Ngn2 gene using doxycycline to start differentiation with the help of biological switch rtTA Use antibiotics to ensure only successfully transduced cells survive Plate the neurons on MEAs Materials list, cost estimation: ~$6000

Source: Rapid Neuronal Differentiation of Induced Pluripotent Stem Cells for Measuring Network Activity on Micro-electrode Arrays

Note here this protocol gives rise to 2d neuronal cultures. “Organoid intelligence” often refers to 3d with more similarity to the in-vivo brain: It has comparable density, higher myelination and support for synaptogenesis.

While it’s possible to clump a 3d culture together from a well plate, researchers use specialized bioreactors or shaker cultures to grow at scale.

During neuron cultivation, common obstacles include cultures suffocating and needing to restart the process, so presumably that's why the shaker cultuers are used to increase nutrient reach. Researchers add systems to provide consistent nutrient and oxygen delivery and are an artificial vascular system. Bioreactors enable the mass production of organoids up to 1 million neurons. These organoids typically grow over of 1 – 9 months and can develop architectures including glia as well as basic features like synaptic connections.

Johns Hopkins University is one of the institutes that produced standardized, scalable organoids. These objects are typically < 0.5 mm in diameter, one 3-millionth the size of the human brain (theoretically equating to 800 MB of memory storage).

We’re working on more formalization of neuron culture architecture with topology frameworks.

3. training

The paper Learning without (artificial) neurons in physical systems, describes learning as solving an inverse problem. Physical materials change their degrees of freedom to exhibit desired behaviours. In Training biological neural cultures: Towards Hebbian learning, authors evaluate learning by checking for three values. If a convergence criteria is reached (e.g. demonstrating the error function is converging), connectivity changes and controlled firing rather than random firing occurs. Some learning categories are unsupervised learning, supervised learning and contrastive learning. Stimulation: We can apply electrical, chemical and optical methods of stimulation to the neuron culture.

Neurons have advantages including parallel processing. One task it can perform is image processing. We can think of neurons representing pixels from the digital image. Then, through a supervised learning process where the filter is presented as electrical stimuli, the network could learn such as an L shape. Unlike conventional machine learning, there is no cost function, training is shown by changes in the physical material and feedback is via stimulation rather than error correction explicitly comparing to the desired output.

Microelectrode arrays (MEAs) can stimulate and measure electrical activity in neurons. The manufacturing is inspired by CMOS technology to increase the resolution. These MEAs take on many shapes:

Realistically, the electrode number on a MEA limits the image resolution. Relatedly neuromorphic computing emulates neurons via hardware, and the Spikemark set provides a list of good problems to solve like classifying MNIST digits, finding the shortest path with wavefront propagation which can be represented in the network, LASSO problems, Sudoku, map colouring constraint satisfaction problems and more complex problems like SLAM using Bayesian learning.

• Basic cell culture methods
• Clean room
• Image processing as one of the initial experiments when the materials arrive

What do these algorithms have in common?

• Start. Initialize the population including the positions and velocities of particles. Define objective function $f(x, t)$ which could represent metrics like pollen density, or pest density within a particular sensing radius. The goal is to maximize $f(x^\)$ where $x^\$ is the global optimal solution.
• Evaluate objective function. For every particle, calculate personal best $f(x_i, t)$ at time $t$. Compare the current $f(x_i, t)$ with global best $f(x^*, best)$.
• Update. Changes in the velocity and position of individual members when there is a new best:
  1. New personal best. If $f(x_i, t) ≥ f(x_i, best)$: then $f(x_i, best) = f(x_i, t)$ and $p = x_i, t$
  2. New global best. If $f(x_i, t) ≥ f(x_i, best)$: then $f(x_i, best) = f(x_i, t)$ and $g = x_i, t$
• Consider population topology. How many other nodes each node is connected to will affect the swarm algorithm efficiency. In some variants, individuals only directly share information with their dearest neighbours, forming a 'local best'.

We compare this with conventional optimization algorithms:

Key differences

Conventional optimization usually use deterministic rules to iteratively improve it's solutions. The search process is guided by mathematical properties of the objecive function. Risk of local optima and coverge faster, guaranteed under specific conditions like smoothness and convexity.

Swarm intelligence uses a population of agents, with stochastic components. It's better at escaping local optima, can perform better at non-continuous, non-differentiable and noisy problems. Likewise, also better at NP problems. NP problems like travelling salesman or knapsack problem involve finding the best solution in a very large number of possible combinations so parallelization helps. However, swarm intelligence is more exploratory and may take a longer time to converge.

Update rules - just take it and implement it in your own code:

Velocity updates: $v_i^{t+1} = v_i^t +r_1 \cdot c_1 \cdot(g-x_i^t)+r_2 \cdot c_2 \cdot(p-x_i^t)$

Position updates: $x_i^{t+1} = x_i^t + v_i^{t+1}$

$v_i^t$: the velocity of particle $i$ at time $t$
$x_i^t$: the position of particle $i$ at time $t$
$p$ = local best so far
$g$ = global best so far
$c_1,c_2$ = cognitive or social coefficients to pull the update to personal best or global best, ranging from 0 - 1
$r_1, r_2$ = randomization weights for velocity updates, ranging from 0 - 1

Assumptions

• The intensity of flashing lights is the sole factor that causes fireflies to be attracted to each other. The brightness is determined by the value of the objective function at it's current position.
• The brighter the light, the more attractive (hahaaa if it was only this simple). If there is no light, then the firefly will perform a random walk in the search space.
• The attractiveness between two fireflies is inversely proportional to the distance between them. The farther apart they are, the less they attract each other.

Brightness
Let $I_i$ be the brightness of firefly $i$, which is proportional to the objective function value $f(x_i)$.

The attractiveness $\beta$ between firefly $i$ and firefly $j$ is modelled by the exponential decay function $\beta(r_{ij}) = \beta_0 \cdot e^{-\gamma \cdot r_{ij}^2}$. This should make sense if a firefly descends into another in circles.

Where:
$\beta_0$: The maximum attractiveness experienced when the distance is 0 (this being a constant is really funny, interpretively)
$\gamma$: A light absorption coefficient that determines the decay rate of the attraction. $r_{ij}$: the Euclidean distance between firefly $i$ and firefly $j$ i.e. $r_{ij} = ||(x_i-x_j)||$.

Movement
Basically we can take the PSO update rule and replace the 'updating part' with attractiveness and randomization terms.

$x_i^{t+1} = x_i^t + \beta(r_{ij}) \cdot (x_j^t - x_i^t) + \alpha \cdot \epsilon_i^t$

Where:
$\beta(r_{ij}) \cdot (x_j^t - x_i^t)$ is the attractiveness term pulling firefly $i$ towards $j$
$\alpha \cdot \epsilon_i^t$ is the randomization term, which is a random vector drawn from Gaussian distribution.

Forms of pollination:

• Cross-Pollination: Pollen is transferred from one flower to another, often by pollinators like bees or birds from the anther to the surface of the stigma. This represents global exploration in the algorithm, where solutions (pollen) are shared between many distant flowers (solutions).
• Self-Pollination: Well, pollen is transferred within the same flower. This represents local exploitation in the algorithm, where solutions are refined locally. Only 10% of plants use this.

Pollen/flowers, what are they doing:
In the algorithm, one pollen grain represents a solution vector in the search space. The fitness of a solution is analogous to the reproductive success of the pollen's derivative--a flower.

And pollinators too?
They act as vehicles for the pollen. Actually, pollinator flying distances aren't Gaussian. Smaller steps are more common than longer ones in many fruit flies and animals. So we draw from a probability distribution that is heavy-tailed--Levy distribution.

You can visualize it in 2D to get a better intuition.

The area under the curve sums to one. So, we model distances using Levy flights to mimic the behavior of pollinators traveling between flowers.

I should clarify the initialization will create a population of flowers, rather than pollinators/pollen with random positions $x_i$ respectively.

The main loop of the algorithm is still, like PSO, with a pollinator quirk in the updating part.
For each flower $i$, 'flip a coin' with value r between 0 and 1.

If r < p, perform global/cross pollination.
$x_i^{t+1} = x_i^t+ L \cdot (g - x_i^t)$
where $g$ is the current global best solution and $L$ is the key difference--a step size from a Levy distribution.

Otherwise, perform self-pollination.
$x_i^{t+1} = x_i^t + u \cdot (x_j^t - x_i^t)$ where $x_j^t$ and $x_k^t$ are two randomly selected flowers, and u is a step size from a uniform distribution.

At this point you get the core idea of the swarm intelligence algorithms with their quirks in the updating rules so I will not belabour it further.

Here's the story behind the cuckoo search algorithm. Parasitic cuckoos often choose a nest where the host bird laid their fresh eggs. Cuckoo hatchlings can kick eggs out of the nest! Some cuckoo species have evolved in such a way that female parasitic cuckoos are often very specialized in the mimicry of colour and pattern of the eggs of a few chosen host species.

Egg mimicry is not a conscious process but rather genetically determined. Likewise, a cuckoo chick can mimic the call of host chicks to gain access to more feeding opportunities.

So in the algorithm, we roughly have the following:

• The basic unit is the egg. Each egg represents a solution, the new egg laid by the cuckoo represents a new solution.
• Each cuckoo lays one egg at a randomly selected nest.
• Best nests with highest quality eggs are carried over to the next generation (of cuckoo eggs). the cuckoo egg will be selected when it’s the most similar to the high-quality eggs are solutions near-optimal value. the new solution will replace the worse solution. For simplicity, you may assume each nest only has a single egg which can be replaced. Though this may be generalized to having multiple eggs
• Host could discover a cuckoo egg with a probability. If the cuckoo egg fitness is less than the host bird, the egg is tossed out. Following, cuckoo may create a new one with simple random probability.

There are a few hundreds of algorithms and an easy way to publish a paper is to laundry-list them and then benchmark on standard test functions. Which bio-inspired algorithms should we pay attention to?

Otherwise, the bio-inspired algorithms have been explored for various telecommunication applications, including routing in sensor networks, electromagnetic antenna design, mobility management, filter design, home automation networks, spectrum sensing in cognitive radio and Internet of Things as well as robotic bees!

"We mentioned swarm intelligence and bio-inspired algorithms. What is the relation between them? Swarm is a subset of bio-inspired algorithms, and descriptively, they're becoming more meta over time.

I made a 3D simulation in Unity during 2 weeks at Founders Inc AI x hardware residency with team Colony. My first time writing in C#. We also had a microdrone < 250g (material list)

video

• A salient application of swarm intelligence is UAVs for agtech or SLAM purposes (#startupideas). I'd be very curious to see your prototypes or consider collaborating!
• Bio-mimicry Institute Go ahead and get inspired by all sorts of natural strategies!
• Robotic bees. The vision is to apply swarm intelligence to embodied swarm of bees! I'd be curious to know what the progress update is with the wyss institute. Last time I checked (2019), they flew untethered for less than a min. If you're strong at hardware and looking for a mission to devote yourself to for at least several months, check it out.
• Another direction you can go in is cybernetics. In the late 1940s, Grey Walter sought was building turtoise robots to showcase emergence from simple non-linear dynamics. So it could move toward light (phototaxis), avoid obstacles and charge itself when power is low. It's groundbreaking at the time and fundamental in today's robotics. It seems like robotics is heavily bio-inspired in general but we don't see this frame.

Fig. 1. Causal diagram for hikikomori.

Psychologically, hikikomoris base their self-image on others’ expectations. Others expect one to excel in examinations and jobs. Some individuals fall behind; Rather than getting up from major failures, they isolate themselves to avoid the negative judgement of others and protect their self-esteem. As isolation continues, the expected and the actual paths diverge even further. Individuals cling to the expected self in an attempt to protect their self-esteem, avoiding any behaviour—conversations, job interviews–that further damages it (Kato et al., 2019). Maintaining reputation is a reinforcing feedback loop that contributes to one becoming hikikomori. Avoidant behaviour and inability to get up after major failures are necessary factors which reinforce each other. When this feedback loop lasts for more than 6 months, it is necessary and sufficient to cause hikikomori behaviour. These psychological factors are shown by purple nodes in the causal diagram.

In Japan, school refusal started to become prevalent in the 1970s and 80s (Cerantola, 2019). On “17 October 2019, the government announced that absenteeism among elementary and junior high school students had hit a record high, with 164,528 children absent for 30 days or more during 2018” (Cerantola, 2019). School refusal can be a salient factor for entering into a reinforcing loop as mentioned above. Anthropologist Anne Allison suggests that after World War II, the Japanese education system became focused on “highly competitive and rigorous high-school testing…The goal was to prepare students for equally arduous employment in Japan’s industrial capitalist economy” (Berlatsky, 2013). Some students experience bullying, which is even termed “natural selection” by some parents. In 1998, in response to increased school refusal, the Japanese government proposed the relaxed-education policy, which intends to give younger generations less pressure. It was implemented for twenty years, however, as reflected by Japan’s decreasing rank on PISA exams since 2000, schools became inattentive to fundamentals (Zhou, 2017). As capitalism continues, it means that students also have to attend cram schools. Competition, control, bullying and relaxed-education policy are four salient factors for school refusal, shown in blue in the causal diagram.

From the 1920s to the Japan post-war economic boom, it was easy for men to get a permanent position at a big company to support a family (Dwyer, 2009). The collapse of the bubble during the 1990s brought instability, uninviting work conditions and unemployment (Kato et al., 2019). After the 2008 financial crisis, many Japanese companies discontinued the permanent employment practice, causing job insecurity (Dwyer, 2009). Economic factors were displayed using yellow nodes.

Japan is a collectivist culture where common goals are emphasized above individual needs, leading to three main implications. Firstly, people are valued if they support the goals of others, forming an interdependent society. Dependence is an acceptable mode of behaviour in Japan even in adult life. For example, as part of the heightened industrialization in the 1980s, employees are encouraged to view their companies and colleagues as part of an extended family, forming close ties (Kato et al., 2019). Interdependence can become parent-child overdependence where the child resorts to parental support after failure. Next, parents have much determination in their child’s goals and high expectations fueled by capitalism. Note that interdependence and parental expectations don’t necessarily conflict with each other: It’s common for therapists to hear parents speak proudly about their hikikomori child’s excellence, believing they can still achieve something in the future while continuing to financially support unemployed children. There is a gender stereotype for men to be strong across cultures. Japanese parents have high expectations for their sons to enter elite universities and land a job in top corporations, and parents are alarmed by the slightest deviation (Ma, 2016). Parents’ expectations tie into the child upholding self-image as aforementioned. Thirdly, in a collectivist culture, one also wants to be quiet about their own needs, giving rise to indirect communication. Particularly, "Making oneself disappear” in social situations where one is shamed has long been considered a virtue, fueling hikikomori’s avoidant behaviour (Kato et al, 2019). Therefore collectivism has three implications that intensify hikikomori behaviour. Psychological, workplace, educational, cultural and global causes spanning the multiple levels of analysis are all salient factors to the emergent property of primary hikikomori.

Fig. 2. A 2-dimensional phase space for individual’s achievement style

Two factors: Time needed to get up after major failures, amount of avoidant behaviour to maintain reputation and the duration of this reinforcing feedback loop can be used to estimate if an individual is hikikomori.

Similarly, from the same two factors, an achievement styles phase space can be constructed. The phase space contains 5 basins around their respective attractors. Firstly our causal diagram explains the attractor on the top right. However, if the failure bounceback time is less than 6 months and an individual is highly avoidant, they’re classified as an independent person. Next to the hikikomori is the depression basin. The characteristic hopelessness contributes to an attractor with a long bounceback time and moderate avoidant behaviour. Similarly, achievers are attracted to showing only their successes and strive to get up from failures extremely fast. Finally, we have a white space since many vastly different individuals have low avoidant behaviour and their trajectory is unpredictable.

External forces can cause a regime shift. Kenta became a hikikomori after being ostracized at school for his high-pitched voice. A young woman known as a “rental sister” began sliding letters under his door, gradually inviting him to walk in nature, exercise and eat out, helping to reduce his avoidance. She supported his self-esteem, reducing the failure bounceback time. Kenta still lives on his parents’ money but has moved out (BBC News, 2019).

Rental sisters can help hikikomori break the bifactor feedback loop. They are a type of home-visit service offered by the non-profit organization ‘Newstart’. Later, hikikomoris will transition to a program to live together with other hikikomori, work as a team in restaurants and bakeries and receive employment seminars (Ismail, 2020).

Yuto became hikikomori after burnout at the end of high school. After being helped by Newstart for a year, he was studying for his driver’s license, attending school to become a graphic designer. He maintains a good relationship with his parents. Now, he lands in the white space on the phase diagram (Ismail, 2020). As of 2021, “New Start has helped around 3,000 Hikikomori go back to regular life” and about 80% of clients successfully reestablished themselves (BBC News, 2019). Network

Figure 3. A network diagram based on three hikikomoris at different stages of recovery.

Based on the case studies for hikikomoris at different stages of recovery, a sample network consisting of the individual and their respective activities were created. In the network, each node is a source of activity, such as Mom or entertainment. Since humans need all four types of support relationships, each activity is classified by its support role to the closest hikikomori: Emotional (red) refers to an activity which brings physical or emotional comfort. Esteem (yellow) means an activity which reminds one of their strengths. Informational (blue) refers to informative activities which suggest the next steps. Tangible (green) relates to offering resources for someone to solve a problem (Scott, 2023).

Each edge represents an exchange of information between two entities, with direction representing where information flows. A weighted edge represents an offline interaction while an unweighted edge represents purely online interactions.

The affected hikikomori (Mr. A) has a degree of 4: An in-degree of 4 and an out-degree of 1. He consumes entertainment sources, social media and hikikomori forums as well as gets tangible support from Mom. However, he only communicates in online forums. The surrounding network is not robust with only one source of each kind. For example, if the node 'online forum' is removed, the hikikomori will have trouble finding esteem support and may further avoid in-person communication. In the case of a recovering hikikomori (Kenta), degree centrality increases–The degree of the node is 8: 8 in-degrees and 4 out-degrees. The number of weighted edges increases. The surrounding network becomes more robust as more sources of emotional, tangible and esteem are present. However, parents are still offering monodirectional financial support.

For a recovered hikikomori (Yuto), the degree of the node is 9: 9 in-degrees and 8 out-degrees. The degree is even higher compared to the recovering individual. A key difference is parents became informative support nodes with bidirectional communication. Secondly, there are even more sources for each type of support.

In general, as a hikikomori recovers, the degree centrality, weighted edges and surrounding robustness for multiple sources increase. The network exhibits a small world effect: Hikikomori all receive information from online sources. Parents can use social media to contact support programs with a geodesic distance of 2 to reach any hikikomori. Aside from Newstart, corrective solutions include virtual AI companions like character.ai, and physical robots like “OriHime” while affirmative solutions include free schools and metaverse social events (Montgomery, 2021). Shut-in behaviour is also occurring outside Japan; In South Korea, 660 000 people between the ages of 20 and 39 live the equivalent life. Having given up on exams, work, raising children and N number of things, they are labelled as the “N-po generation” (Liu, 2023). In China, the Lying Flat movement which gained popularity within the last 5 years is a close analogy, featuring youths figuring out how to live the lowest stress life at the lowest cost. The collective isolation behaviour been reported in other countries such as India, Brazil, the United States, Iran, Australia and more.

Hikikomori is a problem which is difficult to understand linearly. Rather than a singular cause, a complex causality lens suggested that most factors across levels of analysis are salient and identified feedback loops that solutions can target. Similarly, the system dynamics lens highlighted plausible solutions that shift the regime into adjacent basins. Finally, network analysis showed relationship trends for recovering hikikomori. These insights from a complex systems perspective can act as important constraints to guide future solutions to a global problem.

AI statement: Grammarly was used for sentence-level editing.

• BBC News. (2019). Rent-a-sister: Coaxing Japan’s hikikomori men out of their bedrooms. YouTube. https://www.youtube.com/watch?v=q9IRmUEsz6g
• Berlatsky, N. (2013, November 22). Japan's Cutthroat School System: A Cautionary Tale for the U.S. The Atlantic. https://www.theatlantic.com/education/archive/2013/11/japans-cutthroat-school-system-a-cautionary-tale-for-the-us/281612/
• Cerantola, A. (2019, December 23). Why so many Japanese children refuse to go to school. BBC. https://www.bbc.com/news/world-asia-50693777
• Dwyer, J. (2009, January 25). Japan’s ‘employment for life’ myth. The Institution of Engineering and Techology. https://web.archive.org/web/20100506074824/http://kn.theiet.org/comment/management/Dwyer-comment-69.cfm
• Ismail R. New starts at New Start: Recovery and the work of hikikomori. Transcultural Psychiatry. 2020;57(5):698-709. doi:10.1177/1363461520958337
• Kato, T.A., Kanba, S. and Teo, A.R. (2019), Hikikomori : Multidimensional understanding, assessment, and future international perspectives. Psychiatry Clin. Neurosci., 73: 427-440.
• Liu, Z. (2023, June 30). Hikikomori, N-po and lying flat - Snapping in East Asia. Erasmus Universiteit Rotterdam. https://www.eur.nl/en/news/hikikomori-n-po-and-lying-flat-snapping-east-asia
• Ma, K. (2016). Why are there gender differences in hikikomori? Quora. Retrieved October 26, 2023, from https://www.quora.com/Why-are-there-gender-differences-in-hikikomori
• Montgomery, H. (2021, November 29). Japan Has an 'Alter Ego' Robot So You Can Go Out Without Going Out. VICE. https://www.vice.com/en/article/93bbaz/japan-robot-hikikomori
• 1.5 million people in Japan living as social recluses, many as a result of the pandemic. (2023, April 1). The Japan Times. https://www.japantimes.co.jp/news/2023/04/01/national/hikikomori-numbers-pandemic/
• Scott, E. (2023, September 18). Types of Support: How Does Social Support Work? Verywell Mind. Retrieved October 26, 2023, from https://www.verywellmind.com/types-of-social-support-3144960
• Suwa, M., & Suzuki, K. (2013). The phenomenon of “hikikomori” (social withdrawal) and the socio-cultural situation in Japan today. Official Journal of the Italian Society of Psychopathology, 19, 191-198. Zhou, J. (2017). How Yutori Kyoiku Affected Japanese Students’ Learning Abilities [Panel Presentation]. https://mds.marshall.edu/colaconf/2017/day1/8/

Loneliness hits the senior population the hardest, in terms of mental health--increasing the risk of depression and physical health--dementia as well as cardiovascular diseases. In North American, more than 43% of adults over the age of 65 frequently experience loneliness, with seniors accounting for one-tenth of the world’s population.

This also means, if senior loneliness is "perfectly solved", we'd reduce people suffering fromdementia by up to 7 people globally and heart disease by 3.36 million in the US.

The main buckets we looked into are mobility hardware, mobility software and combinations.

1. Overstocking is caused by misjudged consumer demand

   Due to inability to accurately predict consumer demand, retailers in the food produce industry constantly overstock to prevent “selling out”. This method results in constant food wastage.

2. Existing demand forecasting models are not scalable and inaccurate

   Innovative forecasting algorithms have only been tested in Russian regions, supports large businesses and doesn’t yet support automation. Meanwhile conventional forecasting software have only been tested in a few states.

3. Integrating a conventional and experimental model could reduce US food waste by over 30%

   Afresh technologies provides a baseline forecasting model based on Machine Learning with over 600 parameters affecting demand within US. The core algorithm can be enhanced with more accurate machine learning techniques from DSLab, based in Russia.

Overstocking is a $50 billion problem in today’s world. It is a common belief amongst the grocery produce industry that over-stocking will be more beneficial to the company rather than under stocking. This is because retailers can't afford to risk losing sales because they ran out of stock. Within this, fresh products accounting for up to 40 percent of grocers’ revenue and shelf life of 1 to 7 days.

Only large grocery stores have forecasting methods despite way more small grocery stores: There are only 21000 large businesses within the US while 32.5 million small businesses exist. However, fresh products are more difficult to predict than stock items.

Fear of going out of stock

Out-of-stocks cost retailers $1 trillion yearly. The consequences of out-of-stocks are severe, from losing profits to losing customers and inevitably market share. The direct and indirect damages of lost sales are so great that retailers prefer to markdown unsold inventory, or even get rid of it at cost.

Poor inventory management

New inventory can be acquired without accurate data showing your profit margins and cost of goods sold. This information is key to beginning proper inventory management and avoiding overstocking.

Sudden change in consumer patterns

With the demand of certain produce changing on a regular basis, it becomes difficult to judge the consumer demand. This change can be due to latest food trends, economic change, weather, etc.

During COVID, there was an increase in the number of people who couldn't afford groceries and a surplus of food where it wasn't needed. Some of the most popular destinations, such as chain restaurants, schools, and workplace cafeterias are struggling to adapt.

Walmart Canada

Walmart has been undertaking waste-reduction initiatives since 2005. Between 2015 and 2017 alone, Walmart Canada reduced food waste in its operations by 25%.

• Discounting repackaged bruised or peak-freshness products through a $1/$2 Bag Program and reducing prices on fresh meat, bakery, dairy and produce approaching best before dates.
• Reducing over-production in bakery operations. Providing additional processes, training and resources to help support associates.
• Implementing organic recycling programs in stores and distribution centres that take unsaleable and unsold food for conversion into animal feed, compost and energy.
• Partnering with local food banks to assist in maximizing surplus food donations.
• Working with suppliers to improve packaging and food handling processes to maintain quality and freshness.

Harvard Business Review

A December 2017 Harvard Business Review article, How Large Food Retailers Can Help Solve the Food Waste Crisis, presents a four-pronged strategy to help retailers reduce food waste throughout the supply chain, from farming and production to stores and end use.

• Upgrade inventory systems with the latest technology. Examples include software that helps retailers streamline deliveries from the warehouse to the store; or programs that reduce excess inventory and handling.
• Partner with farmers in the supply chain. Retailers are encouraged to share forecast data for specific food items to prevent overplanting.
• Modify or eliminate store practices that increase waste. For example, discounting “imperfect-looking” produce or donating surplus edible food to charities.
• Educating consumers to cut household waste. Content shared with customers can go a long way toward helping them understand the issues around food waste and the ways the industry is working to alleviate the problem.

Afresh Technologies

• Afresh offers an operating system that helps grocers reduce the amount of fresh food that is thrown away. Retailers can view their inventory within Afresh’s system and use its technology to more accurately forecast what foods consumers are looking to buy. The predictive system’s reliance on data means stores will produce less waste.
• Its algorithms also factor in the intangibles like weather and peak freshness of produce to help predict demand and manage inventory.
• Afresh says it helps customers reduce waste by 25% or more. These efforts can also improve profitability
• The company has announced partnerships with grocers in more than 3,000 stores across 40 states
• Company’s total funding has reached $148 million.

Operational Limitations

There is not yet an accurate forecasting model spanning all states. Afresh only has expanded to 17 states. Meanwhile, accurate forecasting software has only been tested locally such as in Russian retail stores and in canteens. The software are not yet automated into the demand forecasting. Similarly, the top influencing factors need to be identified to provide consumers with ease of insight. Data should be generated at the store level.

Technological Limitations

Existing forecasting algorithms require massive volume of training data to yield accurate results, rendering the model unusable for small grocery stores. Existing forecasting algorithms for fresh food usually deliver a median forecast i.e. How many items will be sold in the future? A single number doesn’t allow grocers to adjust to the shifting market values.

Hypothesis

If we combine the forecasting technology used at Afresh with novel algorithms developed by DSLab based in Russia, food waste will decrease by over 30% in grocery stores across the US.

Solution overview:

• Our solution uses efficient machine learning to identify the relationship between more than 600 variables and the demand of products. Sample internal factors include: weekly data of purchasing, variability of demand with promotions and shelf life of product. External factors include: weather, large football games or concerts.
• Machine-learning algorithms that stem from gradient boosting over decision trees are used to obtain highly accurate predictions,
• Rather than a single future amount, the Bayesian method is used to output ranges of future demand e.g. 90% confidence that 300 broccolis will be sold on Tuesday at Costco. 10% confidence that 100 broccolis will be sold.

Why is it better than status quo?

• This forecasting model takes into account promotions. With 60% of goods being on offer, this is a huge factor in changing demand.
• This model allows prediction of fresh food with shelf life of 1 - 7 days
• Machine learning models using Bayesian methods require less training data, meaning this model could also be adapted for small grocery stores, as well as retail which expands regularly.

Feasibility

• This forecasting method has previously been tested in one of the largest retail chain in Russia—Lenta. The result was 13% higher accuracy compared to existing forecasting.
• The novel method is developed by an expert team with 10+ years of experience in data analysis, ML and AI in world-leading IT companies
• There are many studies testing the Bayesian method on the food predictions of canteens. Canteen food are comparable to perishable food in grocery stores.

Unknowns:

• The policies surrounding demand forecasting may be subject to change within the US.

Understanding consumer demand is crucial to the reduction of food wastage. Predicting better consumer patterns will enable grocery retailers to make better inventory choices, and determining the amount of produce to order each week. If implemented just within one Walmart store, this could save up to 300 pounds of food every week.

Holistically, this means reducing more than 88.5 Gigatons of CO2 between 2020-2050 which ranks first out of any factor.

Since Afresh technologies is based in the US, the next step would be to expand partnerships to other North American countries.

Each country has patterns of food waste in each supply chain, which need to be determined before replicating Afresh technology’s model to the region.

• Root causes: https://www.npr.org/sections/thesalt/2020/04/03/826006362/food-shortages-nope-too-much-food-in-the-wrong-places
• Current solutions: https://www.relexsolutions.com/resources/machine-learning-in-retail-demand-forecasting/
• Russian retail Case study using novel ML algorithm: https://medium.com/@dslab/fresh-food-forecasting-challenges-of-demand-forecasting-for-fresh-and-ultra-fresh-food-ef733e8107f7, https://www.dslab.ai/lenta-automates-demand-forecasting-dslab-technology
• CO2 data: https://www.drawdown.org/solutions/table-of-solutions

Summary: I detected my blinking using the OpenBCI biosensing kit and wrote a python program to convert it into pressing the space bar.

How this works: Implemented electrooculography (EOG)! So, the eyeball is positively charged near the front (cornea) and negatively charged near the back. Blinks are vertical movements that cause significant spikes and valleys in the amplitudes of recorded brainwaves. A threshold is set in python script.

Video demo

Summary: How do your users feel about the ad? Getting non-sugarcoated feedback seems impossible--but wait: Smiling and frowning each correspond to activating of a muscle group. We can detect changes in voltage of these groups and connect the results to lighting up LEDs.

Unless the user is under muscle relaxants, this method always picks up the accurate responses, one-up surveys and computer vision analyses. This method can be combined to also output gaze direction and emotional response intensity, which has strong applications in marketing.

Summary:

First, a 20 sec musical sample is generated. The amplitudes of brainwaves are mapped into a good-sounding musical scale (C Locrian).

Then we iterate: The sample is repeatedly passed into MuseNet for 7+ times, which is an AI model for extending music.

Summary:

The current diagnosis for ADHD is subjective and slow. Some BCIs are approved to improve the accuracy of a specific subtype of ADHD.

The current medical treatment involves guess and checking the right drug, while CBT is an indirect solution and depend on willpower.

Some BCIs can directly alter biology by stimulating the brain. Some can provide fast feedback on focus while offering engaging games

Background

• The fundamental unit of the brain
• What are Brain-Computer Interfaces?

Neuralink

• Neuralink’s near term goal
• How the link reads & writes to the brain
• Pig demonstration
• Monkey demonstration
• Perspectives on Neuralink

Applications of BCIs

• Redesigning learning & notetaking
• Smart workplaces
• Everyone can be an artist
• Quality sleep

The future

• How to advance the field of BCIs

• The ethics of BCIs
• The metaphysics of intelligence and intelligence amplification

Before we learn about Neuralink, let’s look very briefly at how the brain works.

This is a single neuron, the fundamental unit of the brain.

Essentially, a neuron receives inputs, computes and decides to pass on or abandon a message. The input comes in the form of neurotransmitters, e.g. Dopamine, serotonin, at the dendrites.

Then enter the computation phase. The number being modified is called the resting membrane potential.

As a neuron is in a soup of ions (charged atoms) and the interior of the neuron is more negatively charged than the outside. This causes a voltage difference or a resting membrane potential of -70 mV. A number of ions, such as K+, Cl-, Na+ flow in and out of the neuron membrane. Example ions include: As there is a negative charge inside the neuron, Na+ ions flow into the neuron. The membrane potential begins to increase from -70 mV. If it goes above -55 mV, then an electrical current is generated and sent down the axon. This is called the action potential, spike, impulse, or firing of the neuron. If the neuron doesn’t match the threshold, then nothing happens.

As the neuron is now positively charged, K+ ions travel out of the membrane for the neuron to return to its resting potential. Meanwhile, the electrical message has travelled to the end of the axon (see diagram). A way to continue the message is to release neurotransmitters across the synapse into the dendrites of the next neuron. The next neuron then “computes, passes or abandons” accordingly.

Our “input, compute, pass or abandon” is a vastly simplified model, not to mention from a computational standpoint. In the real brain, neurons are more like tangled yarn. Large numbers of them are connected in series and parallel, the signals converge and diverge, as an example, a typical cortical neuron has 1000–5000 synapses.

Example: Rat cortical neurons.

Plus, neurons can fire 100 times per second.

To read the complicated organ, our device must be able to detect the location of the spikes, from a larger number of neurons and reflect neural changes quick enough.

These requirements are known as, high spatial resolution, scale and temporal resolution.

The device which reads the brain is a kind of a Brain-Computer Interface (BCI), sometimes called a Brain-Machine Interface (BMI).

An example of a BCI. It decodes mental handwriting.

BCIs could also be used to stimulate the brain. What is Neuralink Neuralink is making a BCI which is implanted in the scalp. Its near-term purpose is to solve brain and spine problems. Namely, the company mentioned depression, anxiety, insomnia, memory loss, hearing loss, blindness, kinds of paralysis, paraplegia, etc. Presently, patients with neurodegenerative diseases require risky surgeries:

7 possible risks of neurosurgery

1. The patient’s hair must be shaved prior
2. Wide opening on the scalp, possibility of scar formations
3. General anesthesia side effects
4. Surgeon errors
5. High cost
6. The need for more surgery
7. Scarring the brain

Some BCIs are already used to partially replace neurosurgery, e.g. deep brain stimulation, where the device can send electrical impulses to stop an incoming seizure. However rigid metal probes can cause foreign body responses from the brain, causing scars that eventually deteriorate signals.

The Utah Array

Often, large boxes, which contain algorithms and batteries, come out of the head, meaning more risk of infection. The device also requires experts to follow around the patient to use. Meanwhile, Neuralink is about the size of a small coin (23mm diameter, 8mm tall). It aims to influence a large volume of neurons while being as easy and fast to install as LASIK. it is barely visible after the implant.

The Neuralink chip, the Link, with extending electrodes

How Neuralink works

The chip has a total of 1024 electrodes, each able to record/stimulate between 1000 to 10000 neurons. This gives control over a total of 10 million neurons. Every 16 electrodes belong to one thread, with 64 threads in total and spaced out by 200 microns. As each electrode is below visible length and should be placed 60 microns near a neuron, a robot is employed for the surgery. An opening the size of Neuralink will be created in the skull, covered by the module and the process will only require partial anesthesia.

The robot can sew electrodes while dodging veins, arteries at a maximum rate of 192 electrodes per min.

The needle pincher cartridge of the robot implanting the electrode

Here is the process of reading the brain:

Neuron spikes → picked up by probes → signal processed on-chip → signal transmission → computer/phone responds When there is a voltage change in neurons, an electric field is produced. An electric current passes through the electrode. This is a flexible electrode made by a biocompatible material, polyimide, wrapped around gold.

Each of the 1024 electrodes amplifies electrical activity, filters noise and digitizes the spikes. From this:

To this:

As all 1024 electrodes are simultaneously monitoring electrical activity, there is a huge volume of data to be processed. Sending them to a computer would delay the BCI. If a user wants to catch a ball, the BCI has to decode the intent in split seconds which then enables the user to do something else. The set of instant interactions is referred to as a closed-loop system.

Thus, all 64 threads (each containing 16 electrodes) connect to the custom ASIC that contain algorithms, spike detection, battery, communication processes, etc. There are four 256-channel chips within the ASIC.

On the ASIC, machine learning algorithms detect spikes and assign them to neurons. Spikes are detected in 25 ms windows, or bins with the spike count represented by a 4-bit number. In total, there is 4041024 = 20 kB of information per second. Along with the number of spikes, there are extra bits of information about the width, height of the spike etc. Information is transmitted via Bluetooth to a smartphone app with controlling software.

Writing

Each electrode can release current to flow along the neuron, causing a chain of spikes. Current can be as low as 10 microamps as far as 4 mm away.

Charging

An implanted link can last around 12 hours and can be charged inductively.

This means an electric current is sent through a copper coil in the charger, generating a magnetic field that generates another electric current in Neuralink.

So here’s how you would control a cursor:

First, calibrate the chip. For example, imagine moving your hand up. Neurons that are associated with both “up” and voluntary movement — in the motor cortex — fire. The electrical activity is digitized; spikes and their locations are detected. The chip associates specific activities of the motor cortex to “up”. After calibration, the chip will knock the computer via Bluetooth to move your cursor.

There are no human trials up to date, but the link’s functionality has been demonstrated two times.

Pig-Computer interface A Link is implanted in the brain region capable of detecting touch on the snout — a pig’s somatosensory cortex. When the Link senses a touch, a series of beeps is emitted from the speaker.

The horizontal lines of the graph represent the activity in each channel, white showing a group of neurons being active; the bottom is a summation of spikes across all channels at any given time.

More activity → Higher hill → Higher beeping pitch

Meanwhile, an average pig, a pig with removed Neuralink implant and pigs with multiple implants are also presented, demonstrating the link’s reversibility and safety.

Limb motion prediction in a strolling pig is showcased.

Some are skeptical towards the accuracy, which we’ll discuss in a bit.

However aside from that, not much can be extrapolated from this graph. The pig’s unique set of experiences relating to each straw can’t be decoded at the moment, as the storage site in the brain differs from pig-to-pig.

Monkey Mindpong

In April 2021, team Neuralink demonstrated a macaque monkey, Pager, playing pong directly from thought.

There are two N1 links placed across the left and the right motor cortex.

Again, calibration first. Pager used the joystick to guide a cursor to a targeted square while Link recorded neural activities. For example, activity in channel 8 might be correlated to intent for rightwards movement.

Pager calibrating the Neuralink implant

The number of spikes is sent to a computer with decoding software, which modelled the relationship between motor cortex activity to joystick movement. Later, Pager can play pong relying on his brain activity and the decoder alone.

In the raster plot, the blue represents 100 channels with firing activity correlated to intended upwards movement. The red represents 100 channels with downwards movement. The firing intensity correlates to y velocity.

Stances on Neuralink

The majority of neuroscientists think that Neuralink is solid engineering but mediocre neuroscience.

Many lab BCIs have low bandwidth, implying that a very specific portion of the neurons needs to be trained to fire, largely unlike the fluency of everyday actions. Neuralink made lab products easier to use. Namely, installment ease, operating ease, bandwidth, latency, # of channels, easily generalizing the controls and the user’s physical comfort. As the chip is trying to save power, the algorithm’s accuracy is sacrificed a bit. Most agree that the robot is innovative. The bold vision and publicity can spark the next generation’s interest in BCIs, AI and neuroscience. Neuralink is based on existing work. For example, wireless implants in monkeys have been demonstrated in 2014. The limb prediction in the pig demonstration was similar to an algorithm in a semi-invasive BCI in 2013.

There are a few micro and macro challenges:

1. As we move on to decoding higher-level thoughts, a greater number of areas become involved. How signals are relayed between brain areas is not well understood.
2. There is no one-size-fits-all solution for brain or spine disorders. A blind person’s brain differs based on time elapsed. As time passes, the visual cortex rearranges into the role of responding to tactile and auditory tasks. The goal is to stimulate the right regions.
3. On the other end, gaining control over small groups of neurons is difficult in stimulation.

For these three reasons, “better engineering”, referring to (a) increasing the number of electrodes, (b) deeper electrodes and (c) better robots might only serve so much. It is still uncertain how long the electrodes can last.

The Neuralink team can continue to explore the balance between battery and performance. A Ph.D. student commented that due to the small size of Neuralink, the motion prediction accuracy was worse than the Utah array, and it wouldn’t be useful for mind-controlled cursors. Similarly, a strolling pig’s motion is relatively easier to predict, compared to one that runs.

A few scientists are concerned about skipping peer review, while a Twitter user countered, if Tesla has been peer-reviewed, we will still be waiting to this day.

This raises the question, how effective is peer review? And other problems at the heart of the academic publishing scene. This will be covered later in the article.

Overall, Neuralink at the moment focuses on translation of existing academic results into industry.

Four main applications would be examined here.

Redesigning learning and notetaking

One word sums up our current education system: Approximations.

The education system is divided into years to approximate the student’s ability to learn. Then approximating streams are created — applied, academic, AP, enrichment, etc. Tests then again approximate knowledge gained.

BCIs, supplying much data, can shift us from approximation-based learning to massively individualized learning.

As of now, teachers can give BCI headsets to students. Upon authorization, teachers can gain focus level, memory and perception data; they can thus alter the lesson schedule, length and order.

Soon, AI algorithms would start to aid the teacher in measuring the length and structure of the lesson. Eventually, it would be fluent enough to design content on its own.

Student confusion can be measured in place of tests. Another function of tests is acting as certificates. A replacement for this could be publishing a hands-on project, an explainer video, mentoring a younger student, etc.

A personalized system like this would be extraordinarily enjoyable for students. This is also time and resource-efficient.

Learning involves taking notes, and BCIs again provide us with a shortcut.

A functional BCI notetaker is being able to read words from the brain at ~130 wpm, read commands to search for images and commands on how to connect the concepts together. This is uploaded onto a computer, can be queried and can be viewed by eye.

A genius BCI notetaker would involve reading words at a faster speed, autocompleting the mental scene, sketches and how the concepts are connected. A viewable version may not be necessary, as concepts could be directly beamed into the brain whenever the user wants. This is an example of Human-AI symbiosis.

Superior learning speed

If you learn a new language after 18, fluency would be troubling to maintain when immersion stops. This is thought to be related to declining neuroplasticity with age. Neuroplasticity refers to the brain reorganizing itself — the synapses being strengthened or weakened to fit new habits. A stronger synapse means more ions are allowed to enter the neuron’s membrane per time.

The two types of tDCS are Anodal and Cathodal, differing by the direction of charge flow to either excite or inhibit neural activity.

This shift can be sped up by using transcranial direct brain stimulation (tDCS), where a small current is applied externally to the scalp.

How tDCS could look like

Some study examples are accelerating the learning of social skills, math skills, motor skills and aiding a stronger memory.

However as this is still a nascent technology, there are risks of scalp burns. The changes are reversible, depending on how you think after the tDCS session. And tradeoffs exist, for example between skill acquisition and fluency of the skill.

** Smart workplaces**

This has been implemented by a Toronto startup, “Muse” as a corporate wellness program. Giving employees BCI headsets could monitor their attention levels, and thus adjust the workplace’s heating, humidity, CO2 and lighting. An employee’s current attention could be gauged and compared to the one needed for a meeting. Focus monitoring could help people decide when to take a break or to stab laziness in its face.

Everyone can be an artist

A person hears a tune in their head, but there is a barrier: insufficient music theory knowledge. This prevents them from recording the melody and composing.

YouTube lowered the entry barriers to content creation, Netflix for film watching, and Replit for programming. BCIs coupled with AI services could lower the entry barriers to almost all creative mediums.

A BCI can one day be used to capture a phrase of the mental song and inputted into AI music generators e.g. FlowMachines, IBM Watson beat, etc. to complete the tune. The user wouldn’t need to study music theory or composing software.

Similarly, in this video, a working webpage is created from a prompt made in English.

The MOMENT is a brain-controlled film made in 2018.

The watcher wears a NeuroSky MindWave headset that tracks attention levels. It is sent wirelessly to a computer to recombine the scenes as well as the background music. The film contains 17 scenes, the user being able to choose from 6 shots per time. Thus, you could watch 16,926,659,444,736 unique versions of it.

Unique journeys through 17 scenes of the film. Here’s another example in fashion.

Instead of picking clothing and shoes from a variety of approximated sizes, they can automatically fit you. BCI then comes into play by enabling personal aura projection. As you stroll down the street, your clothes change in colours.

This is the vision: We use BCIs to record neural activity, then integrate with AI to allow composing, writing, filmmaking, coding, drawing, fashion, metaverse homes, etc. The far future might even directly project emotions or senses to the audience in media.

As there are thousands of possibilities for BCI applications, an app store will almost certainly be built in the near future.

However, as there are few entry barriers and many producers, it becomes difficult to differentiate. Policies for allowed software should also be decided.

Everyone can become an artist, whether you are creating the rules, like the film director, or creating from the rules, like the audience.

Health Only 35% of adults in the US report sleeping on average more than 7 hours per night. This includes multiple wakes, poor quality and others. It is taxing on productivity and is correlated with obesity, type 2 diabetes, etc. Sleep is composed of four stages: stages 1–4, which are non-rapid eye movement (NREM) sleep as well as REM sleep. There is a distinct brainwave associated with each stage of sleep, as shown below.

A non-invasive headset with tDCS could be used to influence our brainwaves.

During NREM sleep, anodal stimulation is used to excite neuronal activity. In REM sleep, cathodal stimulation would be applied.

This is one of the aims of Neuralink. Sleeping efficiently could free up time in the coming decades. A 2016 study showed that sleep time could be reduced by 25 minutes.

Other popular use cases of BCIs involve patients controlling prosthetics and wheelchairs with their minds. This article won’t go into details as there are many writings on this.

Overall, BCIs shorten the bridge between ourselves and goals — learning, work, entertainment and health — allowing everyone to run across the bridge with higher efficiency.

Obviously, risks do exist, e.g.

• Is the decision made by the algorithm, or by the human?
• Who can access brain data?

People worry that BCIs make us less human. A similar question would be: Is the 10 or the 20 y.o. you more you?

Human values are constantly evolving. Imagine if we didn’t have computers, as some person argued that it makes us less human. As a result, Google, Youtube, Facetime may not have been developed to boost our work efficiency. There are tradeoffs and risks with every innovation, but the trend is that ideas breed more ideas to shadow misuses.

Advancing research in BCIs requires research in AI and neuroscience, as well as intersections with fields. However, many researchers said the field looks more like:
Paper factory at work!

To get funding, positions and fame, researchers are pressured to continuously publish papers, with the number of citations as an indicator of success.

This leads to exaggerating the paper’s importance, cramming in intellectual nonsense and sacrificing the overall quality.

A large volume of papers and no incentives cause peer reviewers to slack off — publishing papers that only align with contemporary views, allowing poor quality papers to appear, and occasionally harassing authors by telling them to make unnecessary edits.

Worse, journals and papers are ever-increasing in price. This limits the number of quality research in-game.

Here are a few remedies:

1. Set up funds that require research done with the funding to have open access.

This is being implemented by the Bill & Melinda Gates Foundation. The model could also be used to set up BCI-specific funds, requiring software to be open source to spur collaboration. This accumulates free-to-read papers and code.

2. Make research readable

a) Fermat’s library is a software that allows crowd annotation of papers. The founders also suggested publishing research in threads like on Twitter. b) Researchers could be given an exposition quota, where they must give a certain number of expository talks and nontechnical introductions to their research. Here’s an example of Tai-Danae Bradley’s Ph.D. thesis:

c) Expositions on existing research. A number of videos, games, blogs, etc. were created for the Summer of Math Exposition contest in 2021. At the present, there is a lack of tutorials in the field of BCIs. A “Summer of Neuro Exposition” could be similarly organized, as well as hackathons e.g. BR4IN.IO. That said, it is speculated that BCIs will grow spectacularly in 10 years.

In conclusion, Neuralink’s chip could read and write to the brain and aims to tackle disorders recently. It could be a key player in hyping up the BCI industry and accelerating the coming of sci-fi technology.

• Prediction of Three-Dimensional Arm Trajectories Based on ECoG Signals Recorded from Human Sensorimotor Cortex
• How Neuralink Analyzes Your Brain
• [Brain-Computer Interfaces for Artistic Expression])(https://link.springer.com/book/10.1007/978-3-030-14323-7)

With the development of quantum physics, quantum simulation technology has become more mature and has many applications in real life. This paper introduces two methods of quantum simulation: ultracold atom simulation and integrated photonic simulation. First, ultracold simulation is easy to observe and highly manageable, and its experimental results often agree with the theory. Besides these, it also has limitations. Thus, how to surpass the general theoretical simulation is essential in the future. Then for integrated quantum photonics, it is easy to manufacture, available at room temperature and allows long simulations with a high degree of experimental control. With a few components optimized, large-scale quantum photonic circuits with many photons can likely become a reality in the coming decade.

Since the early 20th century, quantum mechanics has become one of the most basic theories in the cognition of nature, a discipline to explore the physical laws of the micro world. The microscopic behavior of all physical systems or phenomena should meet the basic principles of quantum mechanics, so the simulation principle of any physical system should be based on quantum mechanics rather than classical physics. As the main content of quantum computing and quantum information science research, quantum simulation is to simulate and study the system that is unfamiliar or difficult to control by using a controllable quantum system.

Quantum simulation is a concept created by Feynman in 1982. To solve the problem that classic computers cannot simulate the quantum system accurately, Feynman suggested using a manageable quantum device to simulate the quantum behavior in other systems, which was the original motivation for the development of quantum computing, as quantum simulation has high requirements for computers with a large number of calculations [1]. For example, in quantum chemistry, classic computers cannot even simulate the behavior of molecules in median size. In condensed matter physics, quantum simulation is an ideal platform for research in a quantum phase transition and high-temperature superconductor (HTS) [2]. Methods for quantum simulation have sprung up in decades, and therefore a paper to summarize and compare the methods is necessary for the efficiency of future study. This paper summarizes two methods for quantum simulation, ultracold atom simulation and integrated photonic simulation. Firstly, for ultracold atom simulation, its reasons for being an ideal platform for quantum simulation, such as the considerable controllability of ultracold atoms, will be introduced, along with a discussion of the technical advantages of ultracold atom simulators challenging ideas. Then for integrated photonic simulation, it aims to combine photon sources, routing, optical processing and photon detectors on one chip for practicality, and its necessity in the quantum simulation will be demonstrated in this paper.

Figure 1. The mind map of this section.

The speed of atoms at room temperature is about 300 m/s [3]. However, there is a downward trend in the momentum of atoms with a decrease in temperature, and therefore an increase in de Broglie wavelengths of particles, according to the momentum equation and de Broglie relation in quantum physics $p=\frac{3}{2} kt$ and $λ=h/p$ when de Broglie wavelengths of atoms are the same as the distance between atoms, Bose-Einstein Condensates (BEC) forms, as shown in Figure 2 [4].

Figure 2. The formation mechanism of BEC [5]. Particles move like billiard balls at high temperatures. De Broglie wavelength and volatility of particles increase while temperature decreases, and therefore it moves like wave packets. At the critical temperature, or degeneracy temperature, de Broglie wavelength of atoms is the same as the distance between atoms, atoms interfere with each other, and BEC forms, with quantum effects like tunneling effects clearly shown in the behavior of atoms [6].

Ultracold atoms are defined as atoms at extremely low temperatures, usually in the magnitude of $10^{-3}$ - $10^{-9}K$. Ultracold atom gases are neutral and thin, whose density is usually in the magnitude of $10^{13}-10^{15}cm^{-3}$ [4]. At a low temperature, the elastic and inelastic scattering of particles is controlled by quantum mechanical effects such as resonances or tunneling through barriers since the de Broglie wavelength becomes comparable to the range of the interparticle interactions. Prominent quantum effects, combined with high purity and high handling of electromagnetic fields, make ultracold atoms an ideal platform for quantum simulation.

The ultracold atom system is highly manageable. In solid materials, the internal structure is fixed by the rigid structure of materials. In contrast, in ultracold atom experiments, factors like the dimensions of the potential well, indices of the light field, number of atom gases, temperature, and the strength of the magnetic field can all be regulated according to the requirements of the experiment [7]. This is how quantum simulation is done in subsections.

Preparation of ultracold atom gases.

Preparation of ultracold atom gases needs the combination of many sets of technology. In the procedure, atoms are made cooled and trapped by the laser and evaporation cooling technique and the magnetic trap. The key of the first step is the Doppler cooling and Zeeman effect. The Zeeman reducer coordinates with a magneto-optical trap in this step. As atoms gain and lose kinetic energy respectively when transmitting and absorbing photons, it receives recoil force from photons. According to the Doppler effect, the frequency of incident light sensed by atoms with different speeds is different, and therefore they receive force with different magnitudes. Thus, regulating the incident frequency in different directions and altering atoms' resonant frequency according to the Zeeman effect, a force opposite to the direction of atomic motion exerts on atoms to decrease the speed of atoms.

Consequently, atoms can be trapped by the magneto-optical trap, and a further step of Doppler cooling makes the temperature of atoms down to approximately 100 μK. However, there is a lower temperature limit for the laser cooling technique, as the spontaneous radiation can warm up the atoms. In the second step, the evaporation cooling technique enables the atoms with relatively high energy to escape the system, lower the average energy level of atoms left, and therefore degeneracy temperature is reached, and BEC forms [8].

Figure 3. A typical experiment to cool and trap atoms.

The timeline in figure 4 shows the time point when some important ultracold quantum degenerate gases are realized, meaning that after decades of development, the technique to prepare ultracold atom gases is quite mature.

Figure 4. The timeline of the realization of important ultracold atom degenerate gases [9].

2.2 Bose-Einstein condensates and Feshbach resonance

The BEC is a gaseous, superfluid matter of a boson atom whose temperature is close to absolute zero. Almost all atoms are concentrated in the quantum state with the lowest energy, forming a macroscopic quantum state [10]. BEC has volatility and coherence like lasers, so interference between two laser beams can also occur between two condensates, observed by W. Ketterle [11]. Figure 5 shows the interference between two condensates, with a slit clearly shown. The properties of the wave and the unique state of matter explain why BEC is a good media for quantum simulation [5].

Figure 5. The interference observed by Professor Kettlerle between to BECs.

BEC has a high optical density difference. Lasers can change the state of atoms in the condensates so that there is a sudden increase in the refractive index of condensates, and the speed of light drops immediately, even to several meters per second. That is why BEC can be made a model for a black hole, as light is "frozen" in the condensate [12].

Thus, the unique properties of BEC make it ideal for lots of research in quantum physics and applications. In a closed channel, the atomic motion is constrained, and the channel wave function gradually disappears at infinity, while in an open channel, the atomic motion is not constrained, and the channel wave function oscillates at infinity asymptotically. The typical s-wave scattering length of alkali metal atoms is between 50 and 100 Bohr radii, and the distance between atoms is about 1.4 Bohr radii. In comparison, the interaction between atoms is very weak. Through Feshbach resonance, people can use the magnetic field to adjust the scattering length of the ultra-cold atoms in the open channel, and therefore adjust the strength of the interaction between the atoms. Feshbach resonance is an important technique for the simulation [7].

Figure 6. The diagram of the Feshbach resonance. By adjusting the energy difference between the open channel and the closed channel, the atoms in the open channel can undergo a resonance process with a divergent scattering length [4].

The application of Feshbach resonance benefits from the advent and development of the optical trap. Ultracold atoms are trapped in the optical trap, and therefore they escape from the attraction of the magnetic field. Feshbach resonance provides a method to adjust the external magnetic field so that the strength of particle interaction can be regulated artificially [13]. The transformation from BEC to Bardeen-Cooper-Schrieffer (BCS) superfluid is realized by regulating the interaction intensity between fermions. This technique can simulate the outburst of the supernova and black hole in BEC, control the generation of solitons, etc.

2.3 Spin-orbit coupling and artificial gauge field

In condensed matter physics, spin-orbit coupling comes from the movement of electrons in the inherent electric field of the atom itself, which is the core content of the Spin Hall effect, topological insulation, topological superconductors, Majorana fermions, spintronics, etc. Besides, it can also be extended to quantum computing [3].

It is mentioned above that ultracold atoms are electrically neutral. However, SOC merely exists on charged particles in the natural world. That is why the research of SOC only applied to the fermion system in the past, while the research on the boson system, which is the majority in the real world, was nearly empty. However, this limitation is removed in the artificial gauge field, with the interactions between atoms and light being regulated. In other words, spin-orbit coupling (SOC) on neutral particles is realized. It was first realized in 2009 by Spielman [14].

Figure 7. (a) General experiments with two light beams, (b) The energy level structure of three-energy-level atoms in the experiments.

Figure 7 shows the basic idea of Spielman's method. The particles considered in the methods are alkali metals Rb-87 and K-40, whose electron structure in the ground state is $2S_{1/2}$. Two back-propagation Raman laser beams trigger this kind of SOC along the x-axis, combined with a magnetic field on the z-axis. For one of the two laser beams, $\pi$, its polarization direction is along the z-axis, while for the other one, $\sigma$, it is along the y-axis. With the two laser beams, a biphotonic process is experienced by atoms. First, the photons are absorbed, and the particles are excited to the intermediate state $2P_{1/2}$ and $2P_{3/2}$. Then a photon in another type is transmitted so that atoms return to the ground state [15]. With indices flexibly changed in the ultracold atom system, the condition hard to reach in the real solid materials can be simulated, and new phenomena can be discovered. A typical example of this is topological insulators and topological superconductors.

2.4 Optical lattices

In the solid lattice, the distance between atoms is small, and therefore the interactions between atoms are great. On the contrary, the distances between lattice points and atoms are large in the optical lattice, so the interactions are small. As a result, particles' dynamical property and time evolution in the optical lattice are very slow compared to the solid lattice. However, the mode of motion and equations of motion of atoms in the optical lattice are similar to that of electrons in the solid lattice. Thus, using BEC in the optical lattice to simulate the solid lattice system is feasible. Phenomena that are hard to be observed in the solid, such as the Bloch oscillation, nonlinear Landau-Zener tunneling, etc., can be observed in the optical lattice. In addition, simulation of the complicated model in the condensed matter and the optical system can be realized as the optical lattice technique. The detailed calculations can be found in the review.

Figure 8. The general model of optical lattices [14].

The optical lattice is mainly realized by the AC stark effect between light and atoms. A periodic potential well is obtained using overlapping laser interferences in the optical lattice. The system can realize the initialization of many particles by using the existing experimental technology; the Hamiltonian quantity is also easy to adjust; by regulating the potential field of the optical lattice, the lattice structure and dimensions can also be changed. By adjusting the direction, quantity, angle, polarization, and frequency difference of the incident laser beams, a variety of 1-dimensional, 2-dimensional, and 3-dimensional optical lattices can be realized, and different potential fields can be added to the original optical lattice, thereby realize the quantum simulation of various Hamiltonian [16].

In a typical condensed matter system, electron motion can be seen as the lattice produced by the nucleus, which can be simulated by the optical lattice technique in an ultracold atom system. Take graphene as an example, one of the most promising materials in many fields, with electrical conductivity and strength. With the help of the optical lattice technique, Dirac points, the important structure in the graphene, can be simulated in the ultracold atom gas. Dirac points are the core of numerous new physics phenomena like the massless electrons in the graphene and the conductive boundary state of the topological insulation.

2.5 Results

The great achievements of the quantum simulation obtained in the decade have benefited from the favorable quantum nature of the ultracold atoms. Additionally, the optical lattice technique is also a key factor. The Feshbach resonance made the research on the intersection of BEC and BCS possible, and therefore the simulation of the complicated physical system like black holes, boosting the astronomy research. The artificial gauge field broke the limit that the ultracold atoms are neutral and realized the SOC on the boson system, filling the gap in this field. The research of aggregating dimensions can be seen as an expansion of the optical lattice, bringing new ideas for future research in quantum simulation.

Due to the purity of ultracold atoms, experimental results often agree well with the theory. Therefore, quantum simulations of ultra-cold atoms are often criticized as repeated verifications of theories. Nonetheless, while achieving great results in ultracold atom quantum simulation, we also see its limitations and bottlenecks. Thus, how to surpass the general theoretical simulation is a key point in the study of ultra-cold atom quantum simulation. The interaction between atoms makes our research work beyond the general single-particle image and enter the field of multi-body physics. This is a feature and advantage of ultra-cold atom quantum simulation and an important aspect of its scientific significance. Finding a new foothold is the key to development. In recent years, topological quantum simulation research has developed rapidly in ultracold atomic systems, which is an important direction. But at the same time, we must also see that topological quantum simulation is also developing rapidly in systems such as phonons and photons. In addition, in recent years, the direction of quantum computing has been greatly developed, especially the superconducting qubit system, which has greatly improved its quantum simulation capabilities. In addition, in recent years, the direction of quantum computing has been greatly developed, especially the superconducting qubit system, which has greatly improved its quantum simulation capabilities [9].

Figure 9. The flowchart of this section

Integrated quantum photonics aims to combine photon sources, routing, optical processing and photon detectors on one chip for practicality.

3.1 Parts of the circuit

Figure 10. The scheme of an integrated quantum photonic circuit.

3.1.1 Platform material.

Due to the variety of optical components involved, various materials can be used as the substrate. For example, LN, GaAs, and InP are materials whose electro-optical effects allow fast manipulation of single-photon states. At the same time, silicon waveguides(the photonic equivalent of a wire) can keep light well confined with silicon's high refractive index, layered on silica [17, 23]. A higher index contrast allows smaller optical components.

3.1.2 Photon source.

Integrated quantum photonics relies on single-photon sources on the circuit. A single photon can be generated when a laser pulse strikes a particular type of semiconductor aka a quantum dot. Presently, the best performance is achieved by self-assembled InGaAs/GaAs materials [17]. Likewise, a pair of entangled photons can be generated via a crystal, such that depending on the input photon's polarization, two output photons will either share identical or orthogonal polarizations [18]. (see figure 2) Many other processes including the spontaneous four-wave mixing process can be used to polarize photons, with two photons being inputted.

Figure 11. Parallel(Type 1) and orthogonal(Type 2) polarizations of entangled photons generated in the process of Spontaneous Parametric Down Conversion.

3.1.3 Qubit encoding and manipulation

Input variables in an algorithm are often encoded in the path, polarization, time of the single-photon, though many other methods exist [19].

Figure 12. The three main types of photons encoding.

Photons are routed through the circuit by silicon waveguides, and when two waveguides come close together, they can form directional couplers that effectively divide light. Some examples of directional couplers are power splitters, polarization splitters and wavelength (de)multiplexers. Aside from the above passive elements, there are active elements that reconfigure the circuit. For example, phase shifters controlling phases of interferometers; polarization transformers and space switches acting as tunable beam-splitters. These active elements cannot be expected to operate identically in each use. Thus, they are a source of stochastic noise [20]. If multiple substrates are used, it is necessary to have interfacing components such as fiber waveguide coupling and dielectric mirrors.

3.1.4 Measurement

Quantum information is eventually read out by on-chip single-photon detectors (SPDs). One such technology uses a superconducting nanowire for detection. A superconductor carries electricity without friction, yet there is still a limit on the current passing at once. In SNSPD, a segment of the wire is charged with maximum current, such that when a photon passes by, it causes a decrease in the current limit, leading to a short loss of superconductivity. This change creates an electrical signal to record the passing photon [21].

Figure 13. a. A photon enters the nanowire, creating a hot spot b. Superconductivity is disrupted c. A resistive region starts to span the entire region of the nanowire d. A measurable voltage is induced across the device

SNSPDs can achieve high detection efficiency. However, the chip needs to be placed in a cryostat containing, e.g., liquid helium [22].

Another detector, transition-edge sensors, has higher efficiencies, limited speeds and must be cooled to millikelvin temperatures for operation. Similarly, silicon avalanche photodetectors have a moderate efficiency but are significantly low speed [23]. Still, many quanta photonic functions are integrated with electronics to provide DC power, control and I/O with the classical world [2].

3.2 Quantum Walk Simulation

Photonics is an exceptional platform for performing large-scale quantum walks. A walk is a distribution occurring over a given graph following a defined equation of motion [24]. A quantum walk model describes the coherent propagation of quantum particles in networks [25]. (see figure 14.)

Figure 14. Comparing classical walks with quantum walks

The quantum walk is a powerful tool: It could explain energy transport in photosynthesis; emulate Darwinian evolution, quantum biophysics, systems with properties beyond standard models; solve isomorphism problems, realize universal quantum computation and build quantum algorithms [26, 28]. The well-known Grover's search algorithm can be viewed as a quantum walk algorithm [27]. Similarly, the computation for the boson sampling problem, which is essentially asking, "What is the distribution of photons?" can be massively simplified by running on a quantum photonic device and is key in demonstrating quantum advantage with current technology [29].

Due to the long coherence times of photons, long evolutions of quantum walks can be simulated while giving a high degree of experimental control. One technique using femtosecond lasers can draw precise three-dimensional waveguide networks within the glass. As the third dimension is used to propagate the wave, this allows the simulation of two-dimensional quantum walks [28].

In the past decade, demonstrations ranged from single particles walking on two-dimensional lattices to multiple particles interfering on one-dimensional structures to multiple photons evolving on a two-dimensional platform [26]. They have been simulated using bulk optics and waveguides.

3.3 Realized simulations examples

Molecules can be simulated using near-term devices. In 2018, the eigenvalues of the ground and excited states of molecules, e.g., hydrogen was approximated on a programmable silicon photonic chip, using a combination of quantum phase estimation and variational eigensolver [22]. The results were estimated with 32 bits of precision with fidelities > 99%. The vibrational dynamics of four-atom molecules, e.g., ammonia, sulfur trioxide, energy transport in a protein fragment and a water molecule reaching thermal equilibrium were simulated [37]. Simulations can be evaluated using Hamiltonian learning, where a small, trusted quantum simulator verifies a large untrusted one [31]. In 2020, a quantum advantage was demonstrated on the bulk photonics quantum computer Jiuzhang by implementing a type of Boson sampling with 76 photons [29].

3.4 Platform evaluation.

Integrated quantum photonics is a promising candidate in simulating complex systems. The speed of photons allows interaction between a large number of components that are already supported by the mature CMOS fabrication process, providing the basis for scaling up [20]. The current classical PICs contain thousands of components in a few microns, but millions of components may be contained in the future [20]. The platform can run at room temperature and does not need to be isolated, reducing costs. Since photons do not interact with the environment easily, they have close to maximum-possible coherence time meaning many operations can be performed [20].

However, this non-interactive property also implies difficulty in making two-qubit gates, which are key in realizing universal quantum computing in the gate-based approach. Extreme devices such as the ultra-high-Q resonators are used, generating massive overhead, and the gate-based approach is considered impractical for scaling up [32].

In the one-way quantum computing approach, the difficulty presents in generating multiphoton entangled states in a relatively easy-to-implement circuit [30]. This approach is widely considered more realistic.

Currently, eight photons can be generated and processed in the same silicon chip [33]. Constructing devices with many qubits requires efficient single-photon sources, generating multiphoton entangled states that travel in a network of low-loss optical switches or waveguides. These are foundational to a fault-tolerant quantum computer. Once a sufficient number of photons are available, slight qubit loss will be sufferable, as the information can be recovered by applying Topological Error Correction [30]. Presently, photon counters are the only components that operate at cryogenic temperatures. By developing room-temperature photon counters, completely integrated photonic chips are possible. The room-temperature advantage can then be fully exploited.

In the near term, machine learning could speed up the design and experimental process.

For instance, a model was recently shown to recognize and select pure photon emitters with 95% accuracy and is 100 times faster than the conventional method [34]. Similarly, learning models are shown to have designed experiments more efficiently than the best previous approaches while rediscovering experimental techniques that are only recently becoming standard in quantum optical experiments [35]. Chip imperfections arising from manufacturing could be mediated with a combination of machine learning and on-demand programming of the chip [17]. Finally, machine learning could be used for validating quantum simulation data [36].

With these challenges solved, large-scale quantum photonic circuits with many photons can probably be used for simulation in the next decade.

Quantum computing and quantum simulation develop very fast; numerous achievements and new applications have spawned. This paper makes an overview of the past achievements of quantum simulation.

The advantage of ultracold atom simulators is the flexibility and purity, simulating the especially complicated physical cases and producing results with a high theoretical agreement. However, the future challenge in this field is to improve the uniqueness of the simulator since other techniques in quantum computing and simulation like superconducting quantum computers with high qubit numbers. This paper suggests that the key points of this are to lower the cost of material preparation and simulation and enlarge the range of cases and models it can simulate.

Likewise, the features of integrated quantum photonic simulators are the feasibility of manufacturing, the speed and the long coherence times of the qubit. The device can operate at room temperature except for one component. Much effort is needed to integrate this component onto a room-temperature chip and optimize several other processes, which can be aided by machine learning.

Those technologies may be used in many aspects in the future, and there are still some obstacles that need to be overcome. Through this, we hope more people can understand and interest in quantum simulation, and there will be more remarkable advances and amazing applications in the future.

References

[1] Feynman R P. Simulating physics with computers. International Journal of Theoretical Physics, 1982, 21(6): 467-488.

[2] Georgescu I M, Ashhab S, Nori F. Quantum simulation. Review Modern Physics, 2014, 86(1): 153-185.

[3] Shen X 2018 Quantum Simulation and Detection of Topological Invariant in Ultracold Atom System D. University of Nanjing

[4] Pan J 2017 Quantum Simulation with Ultracold Atoms D. University of Science and Technology of China

[5] Ketterle W. Nobel lecture: When atoms behave as waves: Bose-Einstein Condensation and the atom laser. Review Modern Physics, 2002, 74(4): 1131-1151

[6] Staanum P, Kraft S D, Lange J, Wester R and Weidemüller M 2006 Experimental Investigation of Ultracold Atom-Molecule Collisions J. Physical Review Letters 96(2). doi:10.1103/physrevlett.96.023

[7] Peng P and Gao C 2020 The history of ultracold atoms J. Science. 72(01): 41-46+4

[8] Deng T 2019 Quantum Control in Ultracold Atomic Gases D. University of Science and Technology of China

[9] Yan B 2021 Quantum simulation based on ultracold atoms in a lattice J. Physics 50(01):31-36

[10] Han Z 2019 The Study of Numerical Method Based on Bose-Einstein Condensate D. Shenyang Normal University

[11] Xu Z, Li Y, Zhang X and Wu Z 2017 The physical exploration of Bose-Einstein Condensate J. Physical Experiment of College 30(03): 63-68.

[12] Zhou Z 2019 Coherent Control of Ultracold Atomic Regular and Chaotic Transport D. Hunan Normal University

[13] Feshbach H 1958 Unified theory of nuclear reactions J. Annals of Physics 5(4): 357-390

[14] Dong D 2015 Schemes in Cold Atoms about Quantum Simulation and Quantum Computation D. University of Science and Technology of China

[15] Goldman N, Juzeliunas G, Ohberg P and Spielman I B 2014 J. Rep. Org. Phys. 77126401

[16] Liu W, Zhou X, Jiang K, Liu L, Xia L, Li X 2019 Ultracold atom simulators based on artificial gauge field and the optical lattice J. China Basic Science 21(05):41-57.

[17] Spontaneous parametric down-conversion. (2021, March 16). Retrieved from https://en.wikipedia.org/wiki/Spontaneous_parametric_down-conversion

[18] What is a qubit? (2021, April 20). Retrieved from https://uwaterloo.ca/institute-for-quantum-computing/quantum-101/quantum-information-science-and-technology/what-qubit

[19] Rudolph, T. (2017). Why I am optimistic about the silicon-photonic route to quantum computing. APL Photonics, 2(3), 030901. doi:10.1063/1.4976737

[20] Kingery, K. (2017, December 15). Single-photon detector can count to four. Retrieved from https://phys.org/news/2017-12-single-photon-detector.html

[21] Pernice, W., Schuck, C., Minaeva, O. et al. High-speed and high-efficiency travelling wave single-photon detectors embedded in nanophotonic circuits. Nat Commun 3, 1325 (2012). https://doi.org/10.1038/ncomms2307

[22] Santagati, R., Wang, J., Gentile, A. A., Paesani, S., Wiebe, N., Mcclean, J. R., . . . Thompson, M. G. (2018). Witnessing eigenstates for quantum simulation of Hamiltonian spectra. Science Advances, 4(1). doi:10.1126/sciadv.aap9646

[23] Moody, G., Sorger, V., Juodawlkis, P., Loh, W., Sorace-Agaskar, C., Jones, A. E., . . . Vidarte, L. T. (2021). Roadmap on integrated quantum photonics. Journal of Physics: Photonics. doi:10.1088/2515-7647/ac1ef4

[24] Aspuru-Guzik, A., Walther, P. Photonic quantum simulators. Nature Phys 8, 285–291 (2012). https://doi.org/10.1038/nphys2253

[25] Department Physik - Quantum Walks (Universität Paderborn). (n.d.). Retrieved from https://physik.uni-paderborn.de/silberhorn/forschung/quantum-networking/quantum-walks

[26] Jiao, Z., Gao, J., Zhou, W., Wang, X., Ren, R., Xu, X., . . . Jin, X. (2021). Two-dimensional quantum walks of correlated photons. Optica, 8(9), 1129. doi:10.1364/optica.425879 [27] Quantum walk. (2021, June 10). Retrieved from https://en.wikipedia.org/wiki/Quantum_walk

[28] Gräfe, M., Heilmann, R., Lebugle, M., Guzman-Silva, D., Perez-Leija, A., & Szameit, A. (2016). Integrated photonic quantum walks. Journal of Optics, 18(10), 103002. doi:10.1088/2040-8978/18/10/103002

[29] Garisto, D. (2020, December 03). Light-Based Quantum Computer Exceeds Fastest Classical Supercomputers. Retrieved from https://www.scientificamerican.com/article/light-based-quantum-computer-exceeds-fastest-classical-supercomputers/

[30] Slussarenko, S., & Pryde, G. J. (2019). Photonic quantum information processing: A concise review. Applied Physics Reviews, 6(4), 041303. doi:10.1063/1.5115814

[31]Wiebe, N., Granade, C., Ferrie, C. & Cory, D. G. Hamiltonian learning and certification using quantum resources. Phys. Rev. Lett. 112, 190501 (2014).

[32] Chen, Z., & Segev, M. (2021). Highlighting photonics: Looking into the next decade. ELight, 1(1). doi:10.1186/s43593-021-00002-y

[33] Wang, J., Sciarrino, F., Laing, A. et al. Integrated photonic quantum technologies. Nat. Photonics 14, 273–284 (2020). https://doi.org/10.1038/s41566-019-0532-1

[34] Kudyshev, Z. A., Shalaev, V. M., & Boltasseva, A. (2020). Machine Learning for Integrated Quantum Photonics. ACS Photonics, 8(1), 34-46. doi:10.1021/acsphotonics.0c00960

[35] Melnikov, A. A., Nautrup, H. P., Krenn, M., Dunjko, V., Tiersch, M., Zeilinger, A., & Briegel, H. J. (2018). Active learning machine learns to create new quantum experiments. Proceedings of the National Academy of Sciences, 115(6), 1221-1226. doi:10.1073/pnas.1714936115

[36] Biamonte, J., Wittek, P., Pancotti, N., Rebentrost, P., Wiebe, N., & Lloyd, S. (2017). Quantum machine learning. Nature, 549(7671), 195-202. doi:10.1038/nature23474 [37] Sparrow, C., Martín-López, E., Maraviglia, N. et al. Simulating the vibrational quantum dynamics of molecules using photonics. Nature 557, 660–667 (2018). https://doi.org/10.1038/s41586-018-0152-9

1. Quantization: Spin is quantized, meaning it can only take specific discrete values. For electrons, the spin value is ½.

2. Directionality: The magnitude of spin is fixed for a given particle. The direction of spin can point in any orientation, but it can only be measured along one axis at a time. Measuring spin along one axis destroys information about its orientation along other axes.

3. Wavefunction Rotation: The spin value determines the angle of rotation required for the particle's wavefunction to return to its original state.

For spin-½ particles (e.g., electrons, protons, neutrons, neutrinos, quarks), the wavefunction returns to itself only after two full rotations (720°).

Particles with spin behave like tiny magnets. In a magnetic field, particles with spin experience a force, causing them to follow specific trajectories. The Stern-Gerlach experiment demonstrated this by splitting the beam into two distinct paths, corresponding to the two possible spin states (spin-up and spin-down).

Angular Momentum and Spin
Spin is a form of intrinsic angular momentum, independent of orbital motion. Faster "spinning" corresponds to greater angular momentum, represented by a larger vector.

Spin and the Pauli Exclusion Principle
Particles with half-integer spin (e.g., electrons) obey the Pauli Exclusion Principle, which states that no two identical fermions can occupy the same quantum state simultaneously. This principle is fundamental to the structure of atoms and the behavior of matter.

Photons travelling across the universe interacting with electrons produces higher-frequency radiation, enabling the observation of otherwise hidden objects.

• By increasing the energy of scattered photons, previously undetectable objects or phenomena can become observable.
• Different frequencies can reveal distinct properties of objects. For example, radio imaging provides insights into magnetic fields and free electrons.
• The effect can be used to estimate the thermal energy density of cosmic environments.
• A key application is the ability to generate ultra short electromagnetic pulses.
• In laboratory settings, electrons can be accelerated to high energies and scattered against photons, creating observable signals such as X-rays and gamma rays. This occurs when electrons are in motion and interact with photon fields.

Cosmic Microwave Background (CMB).
The Inverse Compton effect often occurs between cosmic ray electrons and CMB photons. As electrons interact with CMB photons, they lose energy and slow down. This explains why most cosmic ray electrons detected on Earth originate from within our galaxy. Meanwhile, the photons gain energy, leading to slight changes in CMB temperature. This phenomenon is linked to redshifting, allowing scientists to measure the average temperature of gas at various redshifts throughout the universe's history. For instance, observations show that the gas was cooler around half the age of the universe.

Crab Nebula. The Inverse Compton effect is observed around the Crab Nebula, where it contributes to X-ray emissions.

Radio Lobes. Low-energy interactions produce detectable X-rays in radio lobes.

Pulsar winds. The effect plays a role in the dynamics of pulsar winds