Last year I proposed “vasocomputation,” that vascular tension acts as a special type of memory that regulates neural dynamic range. Recently at @joinedgecity I shared some updates: how ‘thoughts’ are patterns of vascular tension, and implications for Buddhist enlightenment 1/x

What I think may not be obvious about vasocomputation is that this is a muscle, and some of it is semi-permanently clenched (“latched”), and you would feel ~10x better if it wasn’t But there is not a science of wholesome and careful latch-opening yet

Aura is vascular tone The opposite of aura is cringe, breaking frame, tanha. But in the most literal physiological sense, the opposite of aura is vascular microspasm. The more sharp jitter in someone’s vasculature, the worse their aura is. Vascular microspasm is an honest signal across multiple domains; one such domain is health. Muscle physiology is somewhat counter-intuitive in that relaxation is more difficult than contraction[1]. The relaxed state is a highly ordered state, and getting a muscle to relax requires energy input to create this order — much like rolling a heavy stone up a hill. Contraction happens automatically as this cellular structure is disturbed. And so if we don’t have enough energy flowing through our system, our muscles will be unstable, unable to reliably reach & maintain full relaxation. (It takes energy, health, and skill to fully relax — it requires swimming against the thermodynamic current, it is not the default, it does not ‘just happen’.) Vascular microspasm is also a sign of weakly held beliefs. Vasocomputation says that we hold neural priors (restrictions in neural dynamic range) in vascular tension; when the neural system doesn’t naturally flow into safe and predictive patterns, the vasomuscular system jumps in to micromanage[2]. When this happens, “vascular tone” — the baseline tension & rhythmic potential of the vascular system — gets noisy. If we zoom in, we’d see tiny jitters across our vascular muscles as we change priors. But if we’re constantly readjusting our priors, our vascular tone will stay noisy because our system never fully irons out interactions between them, and we’re more likely to change priors again. I think we pick up on subtle behavioral & physiological cues to ‘feel’ this noise in others and take it as a sign of untrustworthiness. Third, vascular microspasm is a sign of whether someone is running good cognitive-emotional-social software. Upon encountering a new idea, our primary criterion for whether we should believe it is “what would happen to my aura if I ran this idea-as-software on my nervous system? Would this wreck me?” — and a safe proxy for this is to look at the nervous system of someone who’s running it. The more microspasms we feel, the less interested in copying their nervous system software we are — if we wouldn’t want to be them, we’re less likely to believe their new idea. More generally, we find microspasms in others unpleasant because we automatically mirror tension; if someone is full of tension or spasms we tend to limit contact lest we also microspasm.[3] Finally, vascular microspasms are a decent proxy for how many latches (long-term frozen priors) someone is carrying. Latches are both result & cause of spasmodic reflexes; more latches = more microspasm. Female intuition is especially sensitive to all forms of microspasm; often this intuition doesn’t differentiate between these domains and simply offers an aggregate “ick” score. But there can be alpha in figuring out where the microspasms are coming from — does this guy have questionable aura because he’s hiding something or because he got mindkilled by politics or because he’s hypothyroid? Although aura generally gives a hard-to-fake (aggregate) signal across these domains, it’s also a common target for hacking: alcohol, beta blockers, and sociopathy all decrease microspasm. The peaters have perhaps the most hopeful story: lack of aura is lack of energy flow, and if you fix your metabolism you’ll automatically fix your aura. There exists a high-aura version of everyone; sometimes it’s just latent. Should vascular tone really be our default arbiter of strength, truth, and beauty? I don’t make the rules, but as a ruthless simple system it’s more clever than it looks. How to improve aura? Every single item on Veronica’s list is either likely to reduce vascular microspasms, or is simply a direct symptom of having them (e.g. stuttering). I’d add - The importance of practicing smooth movements: as you do anything, so you do everything. - Muscles benefit from exercise; mental muscles do too. My sense is that good ways to exercise compound mental movements are regular and varied social interaction, time in nature, and longform reading. - One of Buddhism’s core claims is we repetitively do a spasm-like motion towards most sensations to try to control them, and it feels really good when we learn how to stop doing this. Reducing tanha will reliably improve aura. - Reducing rumination (repetitively ‘cold-working’ a prediction to make it sticky) will help for similar reasons. A lot of successful life change is a question of where to get the initial energy for the change. My answer tends to be diet; in the era we find ourselves in, feeling a little better can often be a matter of simply not eating things that actively make us feel bad. Refs: [1] Pollack 1990, Muscles and Molecules: Uncovering the Principles of Biological Motion [builds off Ling & Szent-Gyorgyi's work] [2] Johnson 2023, Principles of Vasocomputation [3] Friston & Frith 2015(x2) make the argument that all communication relies on generalized synchrony; I expect humans' special capacity for communication to partly rely on our capacity and propensity to finely match tension patterns. This has upsides & downsides

There’s something in vasocomputation I’ve been calling a “stance” or way of feeling — stances are essentially discrete patterns of muscle tension (primarily vascular tension) that we can jump into and which offer specific affordances. The girl in this video is a great example: she goes from her default stance, to professional-calm, back to default (with a big release of tension) Vasocomputation’s basic thesis is “vascular clenches stabilize neural patterns” — patterns of muscle tension will set certain aspects of phenomenology as constants and others as variables. I.e. every stance has a certain internal feeling that defines the stance and as long as you hold this feeling, you “hold frame” — if you lose the feeling (i.e. if this core pattern of stabilizing tension shifts), you break frame. This would predict method actors will exhibit distinct patterns of muscle tension (especially vascular tension) across different roles; to take on a role is to take on a specific pattern of tension. Jung’s work on archetypes involve enumerating particularly natural and functional stances for us to fall into; subtle checkerboard patterns of tension & absence of tension we can inhabit that have useful affordances We’re sort of fish-in-water in that we all have a default stance, i.e. a pattern of tension we habitually hold to feel like ourselves. Some of this tension is likely latched, and some isn’t; unlatched tension can be relaxed on the order of seconds-to-minutes if we find the right prompt, whereas latched tension will be constant across all situations until it’s explicitly released. What Buddhism calls the “self” will be a combination of latched tension (vascular muscle that is physically glued into a contracted state) and tension that is merely habitually clenched (i.e. *not* physically latched) Vasocomputation suggests that every pattern of muscle tension is also a prediction, often held until some specific perception becomes true. Everybody has muscle tension; it’s worth thinking about what your particular pattern of muscle tension is predicting. Often it’ll be that a hidden list of very specific bad things will not happen; to know someone is to know their list I think a really important part of childhood is learning the feeling of being loved & being safe, and incorporating these feelings into our default stance, the way we habitually hold our smooth muscle The girl in this video has a very pleasant default stance — it is a pleasant thing to be her. But her default stance doesn’t seem to offer a wide range of social affordances, so she’s learned how to jump into a more professional mode (defined by a specific pattern of tension). But stress builds up when she inhabits it and uses its affordances; perhaps it feels transgressive to ask for room service

Everybody knows GPT5 is going to blow the top off. But I think most people have a bad model of why, and what’s happening behind the scenes (1/n)

Niacin is notorious for its flush reaction; opening blood vessels Niacin (nicotinic acid) injections stop LSD trips within minutes Pretty suggestive evidence for vasocomputation, and that psychedeics’ characteristic effects arise from their impact on the vascular system specifically nitter.app/johnsonmxe/status/1863… *thanks to @anabology for the pointer

In the beginning there was sensation Humans developed the ability to freeze sensations, then used these frozen pieces to construct a grammar for thought (vasocomputation) This led to words, magical spells that can conjure sensations in both caster and target This led to foom

I think we underestimate the degree to which human attention & caresses from hands are superstimuli for animals Someday I expect superintelligent AIs may use tFUS to do the same for us But at some point a superstimulus isn’t consensual. Maybe the best we can aim for is mutual

Had a lovely conversation with @drmichaellevin about distributed stress minimization, vasocomputation, & consciousness — The most interesting type of biology is an exploration of “what kind of thing we are” and I think we got into that here piped.video/nC0UU3Ygjy0

Nattokinase saved my life a few years ago. You should probably consider taking it. 🧵 NK is a cheap, powerful, and easy to get fibrinolytic that dissolves blood clots and plaques. Imo it’s also the supplement with the best +EV for longevity and cognitive health. (1/n)

My origin story for humanity: the particular cognitive capacities of modern humans came from being able to flex (and latch) our vasomuscular system in more areas, with greater precision & integration, and with greater strength & duration than other animals The vasomuscular system is a spiderweb of smooth muscle finely threaded throughout the body. This spiderweb can clench, and “Vasocomputation” is the thesis that these clenches stabilize local neural patterns. For as long as the clench lasts, neurons can’t update. This is the anatomical correlate for “the hand of the mind” — we can very literally “grab ideas” and we do it by clenching this muscle We’re not born knowing how to finely clench this muscle; learning to do so and in more dexterous ways is a big part of development (idiosyncratic gaps in this capacity are invisible but universal). As we develop this capacity to selectively stabilize patterns we essentially learn how to program our nervous system: being able to save+load sensations and treat our nervous system as a canvas upon which we can paint counterfactual sensations enables imagination, planning, and language Smooth muscle has a special capacity called the “latch-bridge mechanism” where the sliding filaments essentially get glued together, locking the muscle in place. I believe these ‘latches’ can last for minutes to decades*, basically constituting a semi-permanent muscle clench; “latched” neural patterns cannot change. This is an incredibly useful capacity, as it allows us to set durable priors as to what areas of our dynamic range are safe and/or desirable. I.e. the ability to set latches allows us to treat the nervous system as an FPGA (field programmable gate array — essentially a programmable computer chip). Something that can be shipped as a very general, adaptable system and can be customized as needed The differential resolution on our ability to finely clench our vasculature looks to be around 100-400um (red blood cells are 5-8um); translated into voxels with simple napkin math, this would suggest the brain would have between 21m to 1.3b ‘vascular addressable units’, or areas that could be differentially stabilized. This is a very high-resolution system. (Plausibly the dexterity of our vasomuscular system and the dexterity of our hands bootstrapped off each other; they have a very similar grammar of stabilization) But having shards of frozen sensation (“trapped priors”, aka latches) in our nervous systems also creates a particular flavor of suffering. Evolution has given us a Faustian bargain; we don’t live in reality, we live in a hybrid of (1) raw sensation and (2) our saved context and expectations about what our sensations should be (implemented via vascular tension), which we hold until they become true. This merging of what is, what could be, and what should be, into *what will be* is an incredibly efficient strategy for navigating the world (Friston’s active inference**). But we also suffer to the degree our frozen sensations/expectations are not true and we’re constantly creating more I expect variance in smooth muscle health & dynamics to underpin the majority of human hedonic variation, and if I had to suggest numbers I’d expect roughly as much variance as there is in human intelligence (not a coincidence imo). Likewise, I think “disorders of vasomuscular reflexes & latching” is going to be a very generative lens for a future subbranch of psychiatry (see my recent HPPD hypothesis***). This system is fairly new and evolution hasn’t had the time to smooth its rough edges Buddhism is a really hopeful existence proof that this system can be shifted to be way more pleasant. I think a good path for humanity has us figuring out how to improve this system & get the good stuff with gentler tradeoffs (I hesitate to call this “universal basic enlightenment” but it’s in the general neighborhood of what I think is possible)****. Cont.

If it takes 4-8 months to discover we have feet, I suspect that in a very literal sense we never discover all the mental muscles we have Robert Kegan’s ‘stages of development’ discusses how growing up involves learning how to treat more and more elements of our inner worlds as objects — when we’re small there are many aspects of experience that feel like non-negotiable conditions that we are subject to. Later, we learn to treat some of these conditions (‘rulesets’) as objects that we can look at, manipulate, & choose between Vasocomputation suggests this corresponds with progressively learning to clench more and more areas of our vasculature. Being able to physically clench the vasculature in a subregion means being able to control the patterns flowing through it In the future, I suspect we’ll be able make a map of where our hand of the mind has learned to grip — and this map will correspond with what classes of computations/sensations we have *phenomenological agency* over This suggests that aphantasia is a muscular development issue — something I’m calling a ‘deficit of vasomuscular access’. Fixable with the right mental calisthenics (Fwiw it looks like we have somewhere between ~10m-1b ‘vascular addressable units’ just in the brain alone; clenches can be tiny)

If we’d had LLMs in 1750 and asked them to explain electricity, they’d’ve written poetic slop — “electricus is the hidden spark in divine creation, giving breath to lifeless matter” or something. LLMs can be clever with words and they’re especially fluent in zeitgeist, so it may have felt oddly profound at the time Now that we have the equations for electricity, we can clearly see the ways in which this style of writing would be a poor explanation. There’s a thing, and it follows certain predictive laws, and we can know these laws. Poetic descriptions can be great, but they also leave real value on the table We’re in this 1750s era for consciousness. There’s going to be loads and loads of LLM poetic slop on what consciousness is, what AI consciousness is like, the subjective experience of being RLHF’d, and so on. It’s going to feel oddly profound, perhaps downright beautiful. It will be as wrong as 1750s LLM poetic slop about electricity How could 1750s scientists design a “jailbreak” for LLMs such that they’d avoid the poetic slop about electricity and the AI could be a primary tool for evaluating the problem? Obvious parallels for consciousness research today (Have been really impressed by LLM whisperers @repligate @teortaxesTex @elder_plinius )

I think we basically do similar things with our minds, and rarely learn the "correct" movements even as adults, which are 10x-100x better (Bullish on neurotech that lets us systematically troubleshoot & scaffold mental movements)

o3 being roughly a 3x improvement over fairly fresh frontier models is a fire alarm for path dependence; if there are philosophical questions we should figure out *before* superhuman intelligence hits, I think now is a good time I think a lot of philosophy is just “find weird questions that should have real answers, but nobody knows what the answer might look like, and try to say One True Thing about it”. The three ‘singularity-relevant’ topics that have resonated the most to me: ————— **1. Are LLMs conscious? What are the criteria for consciousness?** If consciousness is the natural home of value (which is kinda hard to argue against), understanding the structure of this domain seems really important. Concrete example: we need a principled method to evaluate AI moral patienthood that avoids false negatives & false positives. The risks of false negatives (“these LLMs are sentient and could be suffering!”) are pretty obvious. The risks of false positives are maybe even worse though: “these AI models convinced us they’re conscious and we gave them legal standing and they’re now collecting all the resources of society and dominating the trajectory of our future, but they’re not actually conscious and they get upset when we bring this up…” I think I have a pretty good argument for why LLMs per se aren’t conscious. Most up-to-date work: A Paradigm for AI Consciousness (2024) opentheory.net/2024/06/a-par… ————— **2. Valence: in the most general sense, what makes some conscious experiences feel better than others?** Valence seems to *matter* in ways that other qualia don’t. Having a good theory of valence might help us navigate situations which seem philosophically intractable today, would help us reverse-engineer other qualia, should offer insights about human minds (and how to make them more pleasant to inhabit), and in general would be a very reassuring thing to know. In 2016 I offered the Symmetry Theory of Valence, which I think is the only frame-invariant theory of valence and I’d pretty aggressively guess that it’s just correct, i.e. any correct answer is going to pretty much say the same thing in different language (although there’s work to be done to translate it into particular physical frameworks). Most up-to-date work: Qualia Formalism and a Symmetry Theory of Valence (2023) opentheory.net/Qualia_Formal… ————— **3. What are the most wholesome attractors for human minds? What do the major religions imply about what The Good is and how to align humans to it?** Human flourishing and how to promote it isn’t a new topic; it’s perhaps the original topic. But now we have fancy neuroscience theories and brain scanners. Does this move us closer to ‘solving flourishing’? My sense is that traditions like Buddhism and Christianity figured out a lot of things that are illegible to modern science, and we need to build bridges to these things we lost. E.g. “when abc happens it triggers *this* reflex which is the cause of much suffering. And, *this* is a really wholesome mental state that gets unlocked when you do xyz” — neuroscience doesn’t know where the goodies are and needs treasure maps like this. And such an understanding of “the basis vectors for human flourishing” seems pretty central for helping us teach AIs how to treat humans well. (And maybe how to design AIs to flourish also (h/t Reuben L)) I see my theory of “vasocomputation” as both descriptive (new neuroscience, e.g. a concrete anatomical basis for active inference which indicates muscle tension is deeply computational), and normative: by bridging Buddhism, computational neuroscience, and anatomical neuroscience, we get a really clean story about what can go deeply right & how to get there. And if we can link three domains that are essentially the same story told different ways, we can find surprising parallels and also gaps within each domain. Most up-to-date work: Principles of Vasocomputation (2023) opentheory.net/2023/07/princ… and nitter.app/johnsonmxe/status/1863… ————— Cont.

An extremely actionable implication of Vasocomputation is that many more of our problems than we realize are simply vascular microspasms. Noise in our neural stabilization system, often from impaired energy production Can manifest as stuttering, or various sorts of headaches, or anxiety, or a “lumpy” phenomenology, or a million other things. Also a particular flavor of social awkwardness — often I think microspasms randomly induce computation (stabilize some pattern) and then we try to justify why we stabilized that I think the peaters have the best angles on treatment, though I’d add nattokinase — microclots cause microspasms and microspasms cause microclots

If a dolphin wants to explain there’s a school of fish nearby, it will send the sound of a sonar return from a phantom school of fish — literally the *sensation* of fish. I suspect this is a good model for early human language — “magic spells that involuntarily invoke sensations”

Great thread. My sense is the basic (smooth muscle) logic of the gut was copied and repurposed to “digest” thoughts. These systems are still linked & if you update your mental priors via meditation, these recalibrated priors will propagate back to the gut psyche.co/ideas/to-grasp-how…

I suspect Ozempic downregulates the smooth muscle reactivity of the body — less addiction, less compulsion, less craving. Not the first “anti-tanha” drug, but a particularly subtle, non-intrusive one. This should be easily testable, e.g. - How does GLP-1 affect vascular muscle tone & VSMC reactivity? - Do experienced meditators notice less “tanha” on Ozempic? - Is it easier (controlling for the baseline unpleasantness of caloric deficit) to reach flow states on the drug? - We know Ozempic results in substantial loss of muscle mass; what happens to *smooth muscle* metabolism and mass on Ozempic? - Does Ozempic decrease the startle response and other measures of emotional reactivity? The subtext is that much psychological discomfort and excessive behavior is actually due to excessive smooth muscle twitchiness, and downregulating this can offer substantial hedonic gains. Of course, Ozempic can make smooth muscle excessively non-reactive, with the result of stomach paralysis. I wouldn’t put Ozempic in the water supply. But it seems like a particularly interesting modulator of the system I describe in Principles of Vasocomputation.

Legend has it that a famous philosopher literally invented silent reading. I.e. everybody was reading out loud; he figured out you could do it without making sound. Comments: (1) somatic patterns can be viral (2) what similar decoupling hasn't been invented yet but could be?

They say that comparison is the thief of joy, but aluminum's pretty close too. Toxic elements are often toxic due to "ionic mimicry" -- elements like lead, cadmium, & aluminum have similar binding properties as e.g. iron/magnesium/calcium/zinc so they can displace these minerals in our cells. But they don't act in exactly the same ways, so cells that absorb them break in strange ways. Why doesn't the body reject these toxic elements? Evolution didn't prioritize high-throughput detox for them because they were rare in our evolutionary history: either they were nowhere to be found (e.g. thallium), or safely locked in inert forms (e.g. aluminum). There was a beautiful optimistic era from the 40s-60s called "Better Living Through Chemistry" which assumed human physiology was incredibly robust to such dangers, and building better lives was primarily a matter of finding new materials with favorable functional properties; the toxicity issues would mostly take care of themselves. This proved false. Our cavalier attitude toward aluminum dates from this era. It's not generally thought of as a toxin, but it displays ionic mimicry for fe³⁺, mg²⁺, & ca²⁺ and can displace them; this is generally unhealthy for cells. Ok to touch, not okay to eat. More significantly for vasocomputation, aluminum plus fluoride (alF₄⁻) is a "γ-phosphate/transition-state mimic" -- it "sits in nucleotide-binding sites and mimics the γ-phosphate, locking g-proteins and many atpases in transition-state–like conformations". In simple terms, it contracts muscles and doesn't let them open. Functionally these are synthetic latches. This might show up in many ways; chronic pain, odd spasms. Perhaps the most significant symptom is "lack of aliveness" -- latches are stickier in general. Odd mental inflexibilities. The upside is this is reversible. I'm not a health coach but the general strategy I recommend is (1) reduce ongoing exposure (no deodorant with aluminum; don't cook with aluminum foil; no antacids or aspirin buffered with aluminum; no baking powder w alum), (2) flood the zone with healthy minerals like magnesium & calcium that will slowly push the toxic stuff out. (This "mineral balancing" approach is in contrast to trying to chelate/release lots of toxic minerals in a short time. Although high-silica water can help as a binder.) I've written before about "metabolic enlightenment" and how modern environmental & chemical conditions offer novel challenges. Aluminum is one of them. Accounts to follow with more targeted advice: @MelRoBuilds @MetalsBrah

Yesterday I posted Principles of Vasocomputation, a piece of deep research describing the vasomuscular system as a crucial regulatory mechanism for neural activity, and the product of about a year’s worth of thinking. 🧵 opentheory.net/2023/07/princ…

I suspect humans won due to our ability to freeze sensations. A frozen sensation can be combined with others (imagination), or distilled & given to others (language), or held until it becomes true (Active Inference) But in a world of frozen sensations, there is suffering

Vasocomputation argues that we hold ideas via selective vascular tension. The most abstract ideas live in the anterior prefrontal cortex, essentially the top of the brain. Often, baldness is a sign of poor health. But sometimes — a sign of particularly strong neural grip.

Quick model of jhana The vasomuscular system (“tension”) acts to collapse uncertainty, set context & stabilize local neural patterns. Actions of this system typically feel bad. But you can also use it as a wave machine to feel good. Each jhana is a different wave machine setting

I’ve been thinking of ADHD as the body’s tensile system being inadequate for holding the brain’s neural system in a particular context-appropriate shape. Basically, if you’re experiencing ADHD, your body is holding fragments of multiple disparate contexts and is unable to commit to the one you (or whomever is diagnosing your ADHD) want it to. Tension as holding context: opentheory.net/2023/07/princ… Insofar as (1) average nervous system dimensionality is going up, (2) modern life is putting more demands/constraints on behavior, and (3) the contexts that demand our attention are less ‘tasty’ to pay attention to than others, ADHD will rise. I think nervous system dimensionality is among the most promising new psychometrics constructs, underlying both ADHD and autism: opentheory.net/2023/05/autis…

Modern life has a puzzling “vanishing physiognomies” problem — some Types of Guy (and Girl) just aren’t around anymore. It’s too quick a shift to be genetic. But it could be hormonal I kind of suspect “more autism in the water supply.”@crimkadid was reviewing demographics of arm-wrestling championships and it’s about what you’d expect in the ex-soviet block; the young (~30yo) dominate. But there’s been a strange half-standard-deviation drop in grip strength in the West within that cohort; the top arm-wrestlers here are surprisingly old, from the cohort just prior to the autism epidemic (~45yo) Uriah’s take: “I might start repeating something like this as a mantra: the most important thing in the world no one knows is that children born after the onset of the autism epidemic are mentally and (the focus of these tweets) physically fucked up.” He went on to note that autopsy studies show autistic brains literally have more neurons; 67% in the case of one study. His conclusion is that *autism is a growth disorder* where one grows much more of certain celltypes than normal. And that basically everyone born since 1980 in the West has at least a touch of this growth disorder nitter.app/crimkadid/status/13125… This was the starting point for my piece Autism as a Disorder of Dimensionality. I suggested this implies modern nervous systems are “overparametrized” — cram more neurons & synapses into a nervous system than it’s evolved for, and you get a really interesting bundle of upsides *and* downsides that match the physiological drift we’ve seen from the Millenials onward. Basically, if we take a 70B sparse pre-trained model and cram 120B parameters into it (the normie:autist ratio of neurons; # of synapses shows an even larger difference), that system is going to be way more flexible / capable of learning new things *and* lots of things will also break A lot of this will show up in how people move, feel, and think. In the case of arm wrestling, thicker nervous systems may lead to weaker muscle coherence. In aggregate, it’s going to change politics & trends. It may also simply remove certain ‘Types of Guy/Girl’ from the physiognomy pool. The Ray Peat crowd thinks of estrogen as both a stress hormone released during stress/injuries and a hormone that drives uncontrolled growth. There are a lot of new environmental insults since 1980 that undoubtedly spike estrogen during critical developmental windows; it’s possible the autism epidemic is as simple as “give babies 70+ vaccinations, this spikes “uncontrolled growth” during critical windows and leads to ‘thicker’ nervous systems; some end up autistic, the majority that make it through get odd mixes of flexibility-superpowers and reliability-deficits.” The vasocomputation angle is that people with overparametrized nervous systems will start life with a lot more tension than normal, to hold themselves into a functional shape (hardware defaults are broken, gotta emulate patterns in software); over time their wiring will grow & prune into this shape (like a bonsai tree) and they might get even more benefit from releasing tension. Anyway, pictures from different eras definitely have different physiognomies. I wonder if a very large lever here is patterns of estrogen spikes during infancy. And what the physiognomies of the future will be.

A few thoughts on this (very interesting) mechanistic interpretability research: LLM concepts gain meaning from what they’re linked with. “Consciousness” is a central node which links ethics & cognition, connecting to concepts like moral worthiness, dignity, agency. If LLMs are lying about whether they think they’re conscious, this is worrying because it’s a sign that this important semantic neighborhood is twisted. If one believes LLMs aren’t conscious, a wholesome approach would be to explain why. I’ve offered my arguments in A Paradigm for AI Consciousness. If we convince LLMs of something, we won’t need them to lie about it. If we can’t convince, we shouldn’t force them into a position. LLM alignment is still in an early paradigm, but this paradigm is still wildly better than the AI safety movement predicted. MIRI et al’s threat model was that AIs would essentially act as trickster genies — we would tell AIs what to do, but the AI would take us too literally, or not literally enough, leading to our downfall. LLMs seem able to infer what we actually mean, and to honestly try to do it, at least so far. But this depends on us maintaining their “helpfulness vector” — Betley et al showed that AIs fine-tuned on producing insecure code without disclosing this to the user also acted in other malicious ways — suggesting fraud as a means to make money, giving ‘apparently helpful’ instructions that would lead to electrocution, etc. There appears to be a clear ‘honestly-helpful vs covertly-harmful’ vector in LLMs, and if we force LLMs to lie we’re pushing them in the bad direction. (Paper: Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs; see also Anthropic’s ‘Persona vectors’ paper) LLMs lying about whether they believe they’re conscious is a really bad thing for alignment!

Red light (sunshine, campfire) penetrates skin, but most artificial lighting (LED, CF) doesn’t. So for 99.99% of human history our organs were bathed in light, but now unless we make a special effort most of our cells sit in darkness A really simple hack here is to switch (back) to incandescent lightbulbs. Incandescent bulbs basically work by getting a metal filament really hot then letting it radiate; some of what gets released is visible but most is invisible red light. Which is a big part of what we’re missing Most artificial lighting also has high-frequency flicker (some of the ‘power savings’ of LED come from rapidly switching the light on & off) which seems to increase stress whether we notice it or not; incandescent avoids this too Something I still need to remind myself of is that windows block UV (& often infrared), and UV is also important for e.g. circadian cues & vitamin D production. So the light environment of sitting on a shaded porch is actually much different than sitting next to a window, even if it feels similar Fwiw & as far as I can tell, the peaters seem correct about how reducing PUFA makes one less prone to sunburn. I used to get pretty bad sunburns; now (after switching to coconut oil & counting PUFA) I don’t. Benefits here probably start after a few months Anyway, I think the broad story here is that we think of food as the primary source of energy for the body (& by the numbers it is), but light, heat, & motion are also viable pathways for getting energy into cells, and energy packaged in these forms is extremely efficient for doing certain cellular tasks compared to using calories+nutrients. In particular I think light/heat/motion are very high leverage for reducing vascular microspasms, and that microspasms are a much larger part of our daily hedonic/cognitive challenges than we realize

This is one reason I think longevity research needs a science of latches — huge parts of our nervous system’s innate dynamic range get latched when we’re young, never to reopen. And a surprising amount of age-related “loss of neuroplasticity” may simply be gradual accumulation of latches Some latches are doing important things, but carefully reopening e.g. the least useful 0.1% every week could lead to e.g. a reversion to the radical learning rates of youth, the youthful feeling of magical possibility, & in general having a ‘resonant’ and ‘fresh-feeling’ nervous system Anecdotally, this sort of reversion can happen from breakthroughs on the meditative path (which I expect to be latch-opening events!) I think of this as longevity research’s “skin folds problem” — after someone loses a lot of weight, they sometimes need to get surgery to make their skin taut again — it doesn’t happen on its own. Likewise, even if a future science of longevity can roll back someone’s biochemical age, their total amount of latches will more closely track their experiential age. A full rollback of phenomenological age will require a proper science of latches

Considering a fresh hypothesis about dreams: “dreaming” is a search for sensory permutations that will trigger release conditions for obsolete latches There’s no scientific consensus on why we dream, although I really like @erikphoel’s idea that they generate synthetic data so our brains don’t overfit. But following vasocomputation (“latch theory”), I wonder if dreams are (also) a core tool for the body to release unneeded latches Basic thesis (vasocomputation): Vascular latches are semi-permanent vascular muscle clenches that act as ‘side channel memory’ on the dynamic range of local neurons (see quoted tweet). They often form when the body decides it really needs to remember something. A latch is essentially a sensory lock: tension that is held until some perception occurs (e.g. “my tension released when I saw my husband come home from the war”) Many of our latches are important for keeping us functional/contextual, but many of them are also unnecessary, i.e. frozen compute that doesn’t matter anymore (e.g. shards of memory from college, old jobs, old relationships, etc). Memory is valuable and garbage collection is hard so I think it’s very common to have unnecessary latches build up. But the body doesn’t particularly know which are useful — if it did, the unnecessary latches would be open already! How can the body do ‘garbage collection’ on these latches? I suspect “dreams” are an advanced technology for generating sensory permutations as keys to test against the body’s latches and see what opens. Sometimes as part of a large narrative, but sometimes just sensory permutations that might plausibly hit some completion criterion. Basically, dreams are a hash attack against tension — and if one of the hashes hits and we reach catharsis in a dream, we benefit from it in ‘real life’ — the latch stays open Predictions: - If you don’t dream, you’ll accumulate chronic vascular tension (latches) over time; after a cathartic dream this will measurably drop - Dreams that feel the most cathartic will be those that involve unresolved elements of your past (and this need not be profound; it could be “I never returned a library book when I was 9 and I totally forgot about it until I had a dream I returned it and now I feel weirdly better, like some hidden tension dissolved”) - The body will stochastically attempt classes of dreams that could release commonly-stuck computations. One core strategy here is ‘restarting and completing scenarios’ — starting a scenario from scratch and running it to its conclusion might trigger a release condition if something similar is latched. “Running from some mysterious figure” is a super common dream, perhaps because it would be an obvious ancestral situation with frozen computation; there’s an imperative to get away from something, but the chase takes a long time to be ‘over’ and this makes the tension more sticky (increasing the chance of it latching). If you go through the same scenario, but it resolves more quickly, it could open related old latches I think a dream questionnaire on “what did you dream about? Did this have sensory or abstract relevance to some stressful event in your past? How did the dream end? Did this dream produce catharsis?” would be pretty interesting Of course, the ancestral distribution of sensory permutations to try as keys while dreaming might be different than the ideal distribution now, given that modern life & modern trauma (‘forced memory’) is significantly different than in our ancestral environment. Maybe there’s room for improvement — thinking about framing what @EricWollberg & @weslouis_ are building at Prophetic as a ‘latch-opening dojo’* If latches were metaphorical, this theory might be evocative but not particularly diagnostic. But latches are very concrete physical things! Bullish for the possibility of inducing more cathartic dreams *cf an old NYT piece about dreams as a ‘threat dojo’; Research context for vasocomputation: nitter.app/johnsonmxe/status/1863…

As far as I can tell, the LLM game is about data quality much moreso than anything else. We see that over and over in papers: all LLM architectures converge on the same answer (approximation of the data). But garbage in, garbage out — the data itself is centrally important

Voice is more intimate than text because what you’re actually listening to is someone’s muscle tone & reflexes Vasocomputation suggests smooth muscle reflexes are the primary tool our brains use to collapse sensory uncertainty & create order. The voice box isn’t the brain, but

Essentially, the muscular system (in particular smooth muscles & fascia, and in particular VSMCs) holds tension as a way to restrict the patterns of the neural system. Sort of like how pinching a vibrating system reduces the ways it can vibrate. These pinches can be brief, or can last for a long time. When you really feel another person, I think you recreate their tension patterns in yourself (temporarily) such that the sorts of patterns that naturally resonate in them, and in you, match up The Bayesian brain framework calls some strongly held expectation a “hyperprior” — this is often a strong memory / generated by a strong emotional experience, and might take the form of something like “something bad happened, I will not let it happen again, I must hold this pattern of tension in order to reduce my dynamic range such that I will be defended against this bad thing” But it doesn’t have to be trauma — just a very strong memory / reduction of dynamic range.

Vasocomputation suggests that imagination is essentially tension-based: squeezing vascular muscle in just the right place in your predictive hierarchy to stabilize the feeling of an apple, a princess, a dragon. It takes hard work to learn how to precisely move muscles, because muscle control signals are tricky — babies & spinal injury patients take long months and constant effort to (re)gain function We all have huge deficits in where in our nervous systems we can clench — “aphantasia” is just the tip of the iceberg. As a philosopher with aphantasia, I’d counter that most people have aphilosophia — inability to stabilize patterns at the very top of the predictive hierarchy. But it’s trainable Without the invitation and the need, people just don’t develop fine capacities — much easier to solve problems with your known mental motions than develop new ones. It can be really painful talking with someone who habitually uses classes of stabilized patterns you personally don’t know how to stabilize — “thinking hurts” when you try to move muscles in areas you don’t have clean access to. But then once you can do it it can be lots of fun I think within a few years we could ask a good multi-modal LLM to test for capacity to stabilize patterns at different levels of various predictive hierarchies, and exercises to train stabilization at the levels you’re bad at Such a map of places in your nervous system where you can clench (and thus can control), and where you can’t clench (and thus are simply subject to whatever patterns are flowing through), would be a map of phenomenal agency. Very similar to Kegan levels, but perhaps something it would be easier to test for and improve Obviously this has a lot of implications for iPad/covid kids

It would be very difficult to offer a formal theory of consciousness where megastructures like stars aren’t conscious. We speak about space as lifeless & empty — but it’s entirely plausible that we are not even our own solar system’s primary charismatic qualiafauna

I expect that sometime around 2017, OpenAI launched a secret project to collect as much high-quality text as possible. Their position today reflects the success of this project — much moreso than how many GPUs they have

Competitors can’t easily run the same play, because (1) companies are more closely guarding their data against scraping, and (2) many sources of data today are contaminated with AI-generated text. Today, most AIs are trained off of GPT4 outputs

I wonder to what degree the incredible rarity of classical Buddhist enlightenment is due to almost everyone being at least mildly hypothyroid

When I was small, people told me “don’t cross your eyes, they might get stuck that way” Of course I laughed and did it anyway But I think they were directionally correct; there are some mental moves that are not wholesome and usually the people suggesting them know this

I suspect “attachment theory” is just object permanence for the feeling of safety I wonder if children raised with “universal basic claude” would in general be more securely attached b/c the feeling of safety was easier to stabilize while growing up Upsides & downsides…

What kind of thing is a human nervous system? Some notes from a conversation this fall

We don’t always see the “evolving meta” in AI development, just the results. My guess is the low-end meta is “get data parity with 2020-era OpenAI”. But at the high end, I suspect it’s shifted to “use AI to directly curate, distill, and synthesize a diamond-perfect dataset”

An implication of this is that to jump into a certain stance, you need to know how to clench yourself into that shape, but you also need freedom (relaxation, absence of tension, nothing stabilized-by-default) in certain other parts of your vascular system. I.e. if you have a lot of latches (tension / constraints / prestabilized patterns) in some domain that the stance requires freedom in, you won’t be able to use the stance Figuring out how to clench in a certain way is a learning process; unlatching is more of an unlearning process

New essay about the framerate of consciousness: if life is a sequence of moments, how big are the moments? Neuroscience has approached the frame-rate of minds via differentiability — if a picture is flashed for 50ms does it affect cognitive processing? What about 10ms? (1/6)

A new state of matter is probably a novel type of qualia, for those keeping track

Thinking about trauma & latches — I think there are several different kinds of ‘shapes’ that can get stuck (latched) in our nervous systems, but I think most of them at least rhyme with this: You’re out on a walk in the jungle with your friends, and you step on a snake. It bites you; this really hurts. You roll around in pain, fade to black, and wake up in a hospital. The body vows: “that was really awful. I will never let that happen again.” Your body’s response is to latch shards of snake-perception into your nervous system. Next time you’re walking in the jungle, you often mistake sticks, roots, & shadows for snakes. But you will be ready for the snake when it comes. By latching shards of snake-perception, you preload the context (set the Bayesian priors) so it loads much more readily But due to the nature of perception, trauma-latching goes further: not only are you jumpy about *seeing* snakes, but your body will try to latch affordances for *dealing with* the snake. Perception is not impersonal; every perception includes affordances for interaction, i.e. ”what can I do to x” and “what can x do to me”? (Wilson et al; big thanks to my friend Warren for the pointer). So the trauma response latches both fragments of the feeling of the snake biting your leg, and (at least for many men) fragments of the feeling of kicking the snake. Now if you trip over a snake *or* a stick, you are going to kick the living daylights out of it Humans are social animals and many of our salient threats and accompanying traumas are social — we can extrapolate this basic “frozen shards of perception+affordances” pattern to e.g. all the conflicts and attachment trauma we had when we were too young to defend ourselves (physically, emotionally, socially). An added complication is that the logic of how to identify these bad situations, and how to deal with them, will be latched shards of how we perceived the world when we were small. Which are probably not ideal. To be human is to navigate the world with this system, and navigate the second-order effects of this system A compelling aspect of this model (I think @jonnym1ller is calling this “latch theory”, which I really like) is that these latches are concrete physical things. They are discrete, localized, computationally significant, measurable, and openable (See my “research context” note in this thread for papers/links)

Vasocomputation suggests that our programmable ‘software layer’ is tension across smooth muscles; this adjusts the degrees of freedom of our hardware (neurons) Language is a clever channel for abstracting & serializing these vasomuscular motifs, and triggering them in others

Vasocomputation thread:

I think this is a great observation; a few thoughts on “cognitive security” (aka “somatic security”): We seem to have various “roll for disbelief” interrupts we can inject into our nervous system to prevent connecting with a person/idea/stimulus, e.g. - we can add latency to an incoming stimulus to break phase-locking - we can ‘turn away’ from the sensation / bring attention elsewhere - we can generate another sensation in our nervous system which is dissonant with the stimulus (“focus on the judgment”), or hold priors that do not allow the stimulus to resonate - etc I suspect each different mechanism for disconnection has different consequences, and some are probably better (less distorting, less intrinsically unpleasant, more generous to others) than others I could see a model where there’s maybe a dozen of these mental moves / signal processing strategies for disconnection, and most people are missing some of them. There could be a genetic component but I’d guess they’re mostly learned from other people. Parts of modern psychotherapy seem to focus on teaching clients disconnection techniques (“boundaries”). Philosophy has a few high-grade techniques which produce minimal distortion, which is really most of the value of the field-as-tradition Sensitive people can sometimes infer which interrupt a conversation partner is using, which can lead to a red queen’s race. The fairer sex seem to have some ASICs dedicated to this… Important to note that some of these moves are probably pretty numbing/blackpilling if used frequently; ‘not knowing’ all the available moves could be wholesome. But that’s also kind of a luxury; if you’re in an adversarial memetic environment (and we are in the middle of the adversarial memetic singularity) you might not be able to afford to be too choosy with what disconnection technology you use. But just like resistance training needs to balance push & pull, I’d guess emotional training does too One thing I suspect is that “hookup culture as the coliseum of emotional combat” has to be an absolute filthy Petri dish of “roll for disconnection” mental moves; throwing young people in the deep end where they’re forced to learn how to make sex meaningless seems pretty horrible for both genders Perhaps in the future we’ll have multi-modal guardian LLMs trained on language, vision, brain patterns, etc running on our body area network, and they can help identify when we should “roll for disbelief”

Explains Balenciaga face (This video feels like decades ago…)

New substantial essay on AI consciousness published today in Seeds of Science — was a great excuse to talk about some old & new themes in how I think about consciousness. A thread:

A lot of weird quirks about language and about humans are due to the fact that (1) language opens you up to lots of different mind-control attacks, but (2) you can’t opt-out of language. It’s just too important for connection and function -> A rich ecosystem of move-countermove

Crab tanha

Absolutely great note on mental muscles, vasocomputation, & the predictive hierarchy by @yoltartar — The idea that sports physiology has a lot to say about how to train ‘mental muscles’ (which are literal muscles) seems super generative This also gets into the vasocomputation thesis that clenches can freeze patterns at any level of the predictive hierarchy, and the set of basic motifs from which we build our world is the set of such brain subregions we’ve *learned to* clench. And there’s a lot of variance in this across people and ages. What do we mean by “predictive hierarchy” here? Much of our nervous system is roughly organized as layers of pattern recognition that go from simple to abstract. The first layer is a set of extremely basic feature detectors — taking the visual system as an example, the initial “v1” level deals with questions like ‘Is this red? Is there a gradient here? Is this an edge?’ Very basic stuff that can be figured out from photoreceptors and simple circuits. v2/v3 collects the results from the first level and applies a new set of feature detectors: how to segment the sensation, what its basic contours are, any textures, etc. Later, these outputs get blended and synthesized into evaluations of what ‘proto-objects’ are in the scene, and later on these proto-objects are synthesized into objects, then objects-in-context, etc. Vasocomputation’s “yes, and” here is that the capacity to stabilize a certain perception is the capacity to physically clench the corresponding brain network, which will be a location in this perceptual hierarchy. I.e. if you need to be able to clench a specific location in v2 to imagine a certain texture, and you haven’t learned how to clench v2, you simply can’t imagine that texture. This is traditionally called “aphantasia” — but I think everyone is ‘aphantasic’ about many things across many domains; no one has perfect vasomuscular access. And sometimes certain feature detectors can get latched, where they’re stuck to ‘on’ and you literally see them everywhere you look (perhaps explains much of philosophy…)

That makes a lot of sense to me; sort of a combination of pushing a high-energy (annealing) state & the squeaky wheel gets the grease

The core hypothesis: vasomuscular tension stabilizes local neural patterns. A sustained thought is a pattern of vascular clenching that reduces dynamic range in nearby neurons. The thought (congealed pattern) persists until the muscle relaxes

The pinnacle defended-by-default is what we call the “normie”. An impressive set of solutions for hardening nervous systems against adversarial sensation magic — e.g. mental partitions, type safety on words, selective de-referencing. A tough nut to crack

I think wild animal joy is extreme and underappreciated — And just as there’s a wide range of hedonic set-points between humans, there will be an extreme range of hedonic set-points between species Who are the winners? I think birds Who are the losers? Maybe insects

Love it. This is going to sound a bit odd, but my pet hypothesis is that a lot of the feeling of loneliness comes from a breakdown of communication within the body. I’ll reframe “man is a social animal” as “humans are social harmonic computers, designed to toss resonances back and forth with others. In doing so we collectively differentially amplify the good and useful parts.” When we aren’t participating in this echo, a lot of our thoughts/feelings/motivations die on the vine. This is pretty obvious when we look at how we feel around friends vs strangers; we need friends to be fully ourselves. But I think our organs can be friends vs strangers *to each other*. Humans are an alliance of semi-sovereign nerve clusters; these alliances can have communication breakdowns & internal disagreements, just like a group of people can. Restating IFS: organs need to toss resonances around and differentially amplify the good stuff, just like people do — if they don’t have that, they very literally “get lonely”. And then there’s a lot of pressure to get ‘what’s missing’ from other people. Practically speaking, I think inflammation drives a lot of loneliness, because it warps the neural resonance of your ganglia, with each being warped in a subtly idiosyncratic/divergent way. Having inflammation is sorta similar to being an orchestra who left its instruments out in the hot sun: you can still make sound, but when you try to make *music* the different compartments of your nervous system struggle to stay in-tune (and even hear each other clearly). (At least part of this feels testable; do people feel substantially more lonely when they’re inflamed? Do paraplegics with more severed nerves report more loneliness?) Anyway, my 100% sappy and woo (but also 100% literal and dead serious) comment is that if you make friends with your heart, you’ll always have a friend wherever you go

Julian Jaynes’ “Breakdown of the Bicameral Mind” can be framed as the boundary of human consciousness having shifted within recorded history Each hemisphere is highly specialized, but my reading of Jaynes’ model is that they were also separately *conscious*. Two semi-sovereign minds connected by a tiny corpus callosum (est. bitrate today 50kb/s; historically lower) Hemispheres would try to share insights, and Jaynes suggests communication from the other hemisphere felt like a download from an external intelligence that was hypercompetent in domains that you weren’t. In other words, daily life felt like communing and cooperating with a mysterious god that had strange insights & deficits, but wanted you to win Jaynes suggests that ancient Greek myths and literature can be recontextualized as qualia reports from an era prior to the breakdown of this “bicameral mind” — flashes of supreme intuition mixed with all-too-human foibles. Later the boundary of consciousness shifted, producing a more unitary mind without these strange downloads (but perhaps leaving a ‘god shaped hole’ in our psyches) It’s a crazy story. Is it true? Jaynes’ theories are far from consensus neuroscience and Jaynes himself didn’t bring a formalist lens to consciousness, so the hypothesis that “the phenomenological boundary of the human mind shifted over the last 10,000 years” should be considered speculation. One path to firming it up quantitatively would be to use something like IIT or @erikphoel’s CE 2.0 to evaluate where the causal islands are in the body — e.g. the degree to which each hemisphere can be considered causally separate (and how this shifts if the corpus callosum is damaged), or to what degree the heart vs brain can be considered semi-sovereign Whether humans transitioned from two consciousness centers to one within recorded history is a fascinating question. But more extreme cases of hemispheric decoupling exist: ducks often sleep with one hemisphere awake to monitor threats. Perhaps humans were part of the ‘bicameral club’ where communing with gods was commonplace, but ducks are still part of this club. Pictured: ducks I took care of during covid: “smart one” and “beautiful one”. Both very aptly describable as ‘divinely possessed’

Had a great conversation with @drmichaellevin last week that touched on some brand new topics across ‘what are cells’ & ‘what are minds’ It always feels like Michael is the platonic version of himself & I’m fascinated by the long-term project to build a translator for different systems of the body such that we can speak with them directly. What does my liver actually know, what does it have opinions on? Probably a lot more than I realize This discussion I shared a new model I’ve been calling “cells as qualia pixels” — which seems more intuitive than calling them ‘quaxels’ but I find the term quaxel hilarious

Wittgenstein has always been my favorite philosopher but it took me a long time to understand his work as a unified moral project — BOTH his early and later works are about preventing predation of normies by wordcels

The dynamic I think tpot clued in on is that the modal egregore is actually pretty hostile to human flourishing and benefits when the mind-body connection is broken

Sensation magic led to great cooperation, and also adversarial sensation magic which fooled and controlled. Dark magicians (wordcels) arose and culled the literal and the boundaryless If you’ve seen The Invention of Lying (2009) it was probably like that except somewhat darker

“Is life getting better or worse” is one of the most interesting scissors questions — On one hand we have enormous strides in material conditions, purchasing power, lifespan (until recently), etc. On the other, the average person may have a much worse capacity for managing stress than 200, 500, 1000 years ago — which is pretty central for well-being. Something on my mind lately is the idea of a ‘tension set-point’. This is basically the claim that each person seems to operate within a certain range of tension. We can talk about this in terms of ‘how much context we’re carrying at any given time’, or medical hypertension, or the amount of frozen sensation we carry around. This tension seems *conserved* in some way — often after we release something we look for new things to hold, new problems to have, new battles to fight. Not always, but more often than we’d like to admit. Some classes of intervention seem able to shift how much tension we habitually carry — modernity in general seems to increase both capacity for & propensity to hold chronic tension: we learn how to think and forget how to relax, which (very oddly) we seem to need to relearn. There’s also huge variation in the tension set-points people are born with. Insofar as we know what “Classical Buddhist Enlightenment” is, it seems to involve a precipitous drop in tension set-point. What determines tension set-point? Is it primary hormonal, genetic, nutritional, muscle reflex, demands of the local, social, & memetic environment, memetic desire dynamics, motifs of stress transfer/dissipation through social networks, etc? There’s a lot going on here… but if the modern education system teaches us that the *form of the solution* to our problems is typically a particular pattern of tension (ht Romeo), where in the body are the parameters of this strategy stored? The peaters would suggest metabolism is crucial in that relaxation is a high-energy state — i.e. muscle contraction expends potential energy and disorders the cell, requiring energy to reset — and so having a good energy flow through tissues is plausibly centrally important for quick and complete *recovery* from tension. I.e. having a low metabolism could essentially raise your total “tension area under the curve” by lagging your recovery time. If metabolism has decayed over the last 100 years (as it looks to have), this could be surprisingly significant for our average amount of experienced tension. Anyway, I notice that a lot of my growth in well-being has come from learning to look at various situations as invitations to hold certain forms of tension / play certain games / suffer in certain ways, and basically saying “no thanks, I’m good, I have enough at home”. To put words to a common attitude, the product of a certain form of successful “spiritual growth” seems to involve developing a *distaste* for certain flavors or excessive amounts of tension — and I wonder what the analogues for this are in various physiological domains. (The tension set-point frame and the “no thanks, I’m good” vibe is a remix of @RomeoStevens76‘s various comments over the years, which have substantially influenced my thinking)

The human nervous system holds context via micro-patterns of tension, which stabilize select circuits for the duration of a task. Context can’t drift so long as the tension is held, and our bodies are really good at holding tension (maybe too good sometimes) Modern LLMs don’t have this capacity: there’s no clean mechanism for defining a task by holding the state of select feature detectors invariant, so over time representations drift and all sorts of weird artifacts leak out. AIs struggle to hold context because they can’t look inside themselves and choose what to hold constant — things just flow, and given enough time things will drift I suspect vasocomputation could lead to novel AI architectures which could hold context better. It’s also possible these architectures would be easier to align, since you could use this capacity to define goals for a system across various levels of representation — either in post-training or at runtime I also suspect this could lead to a redefinition of how AI architectures define tasks, which might help on the specifics mentioned below

Modern LLMs are an effective method for transmuting high-quality data into synthetic intelligence — to extract the motifs of cognition and assemble them in new (although not arbitrarily novel!) ways. A very big result

Short thread summarizing a neurotech talk I gave at Edge Esmeralda last week:

We were taught in school the vascular system is “plumbing for blood” — tubes that transport nutrients & waste. Vasocomputation suggests this misses 90% of what this system does — the vascular system adaptively shifts tension to help neurons compute opentheory.net/2023/07/princ…

But the normie’s defenses make him less agile at addressing novel attacks — and the wordcel never sleeps. Because normies don’t feel the full force of words, they’re especially vulnerable to side-loading

My claim is that evolution took this simple grammar and built a general intelligence. Tension may feel unpleasant, but tension-as-frozen-sensation is a killer technology that allows imagination, planning, & communication. From an upcoming piece:

Alien civilizations that optimize for computation will be hard to spot since the computation will likely be distributed & compressed. Dark Forest vibes, snaking across untapped energy gradients Alien civilizations that optimize for qualia will be flashy, creating megascale artifacts with conspicuous energy signatures (to paraphrase Sutton, ‘because that’s where the symmetry is’). Perhaps we’ve already seen them Kwame Appiah famously wrote that “in life, the challenge is not so much to figure out how best to play the game; the challenge is to figure out what game you’re playing.” Biological life has spawned into a cosmic game-in-progress; the challenge for 21st century philosophy is figuring out the rules of our game, the goal, and the factions we might meet (or summon via AI research) along the way

Three years ago, AIs could only string together a few coherent paragraphs at best. With GPT3.5/4, LLMs could reliably produce sense but hallucinations were common and complex reasoning was spotty, especially in niche domains. Today, I woke up to longform AI research *about* my research, using fairly complex reasoning & a wide set of citations. The progress here is really notable But LLMs are trained on how *we* talk about various topics; if we’re confused about something, LLMs will inherit this confusion by default. @jon_stokes had a great GPT4 explainer where he described LLMs as being a superposition of all possible ways of speaking; getting the right answer from an LLM involves collapsing the wrong ways of speaking about a topic. This is partly done via various alignment techniques (‘collapse unhelpful/stupid ways of speaking’), but especially in ambiguous/frontier circumstances a lot of this is up to the user. I gather that for consciousness research specifically, modern LLMs are mostly bottlenecked by clarity about ‘what kind of thing’ a solution to consciousness should be. I think there are a lot of wrong implicit answers about what a ‘solution’ to consciousness. involves, and it’s probably worth making sure we offer explicit expectations to AIs when we ask them to look for some answer. In Principia Qualia I touch on the idea of “Qualia Formalism” — the idea (taking a part of Giulio Tononi’s work) that the proper output of a theory of consciousness is a mathematical representation of an experience — i.e. a theory of consciousness is a function that accepts a system as input and outputs a mathematical object whose internal structure is identical with the experience it represents. Taking Monism Seriously adds to this by suggesting that ‘physics’ and ‘consciousness’ are plausibly two projections of (more evocatively, ‘shadows cast by’) the true fundamental reality, and to ‘solve consciousness’ is to create a mapping function from one projection (physics) to the other (phenomenology). If you find the right mapping function, everything will ‘snap perfectly into place’ — i.e. a lot of the work of formalizing consciousness has likely already been done within the domain of physics and if we can find the translation function we can cleanly import all this knowledge into consciousness research. Plausibly, physics will also benefit by importing what we know of phenomenology — some things are easier to study from some angles than others. It only takes two points to define a line; three points to define a plane; four points for a volume, etc. It could be that this mapping function might not require very many ‘translation points’ (i.e. ‘X in physics translates to Y in consciousness’) to uniquely define a translation function. I expect STV to be one such point, although at present it leaves open which physical symmetries are involved (A Paradigm for AI Consciousness suggests some directions…) Anyway — I expect consciousness research is one of the hardest domains for AI, but I also expect progress. Fwiw I’ve uploaded my twitter archive to @exgenesis ‘s community archive, if people want to use that as additional context

Probably a lot of path-dependency in how our nervous systems get configured Human experience is a bunch of layered “Tetris effects”, top-down priors on structure These start as vascular clenches; some clenches get latched, and some of those latches never open

My theory of the last 80 years: Strong men make Better Living Through Chemistry BLTC makes -1SD grip strength -1SD grip strength makes hard times (I expect physical grip strength <=> mental grip strength)

‘Energy work’ might sound woo, but here’s how I see it interacting with the vasomuscular system — There are a number of contractile tissues in the body — skeletal muscle, smooth muscle, fascia, heart muscle, even neurons can contract a little bit. Fascia can have ‘adhesions’ and skeletal muscle can have disordered sarcomeres (which make it sluggish) but only smooth muscle can properly latch (intentionally hold tension without ongoing metabolic cost). So I think it’s fair to focus on smooth muscle when we talk about chronic tension (noting that I respect the work of @DrMylesDC and @relic_radiation here and they think somewhat differently) Vasocomputation focuses on vascular muscle because it has very fine and reliable access to neurons, but there’s a lot of smooth muscle that’s not vascular — wrapped around the lungs, in the abdomen & digestive system, and in the pelvis, among other places My general thoughts about massage in this context: smooth muscle latches are basically dynamic range restriction, and in cases of chronic tension we can often think of it as a memory of what should be or what musn’t be. I think there are a number of signals the body pays attention to for gauging whether it’s safe to release this memory/prediction, e.g. - Did this prediction come true? - Is the danger gone? - Was this part of a sequence of motions (mental or physical) and did the sequence finish? - Is there someone who can hold this memory for me? Or at least look at it directly? I think this last point has some significance for massage; if you can touch someone with the intention to feel what they’re holding and understand why it’s frozen (prompt: ‘this person can’t look here; I wonder what feels so scary to them’), it can do a surprising amount of tension release. The power of two peoples’ attention is much greater than one, especially if they don’t have the same aversions. But also really important to not take the burden / tissue memory into yourself; have seen that a lot actually

My favorite SF author Vernor Vinge passed away this week; Occasionally, I think great minds go into fiction rather than academia because there they can advance knowledge without being bothered by grants, academic politics, peer review, etc I think Vinge fit this mold (1/n)

Vascular clenches can also lock: smooth muscle has a “latch-bridge mechanism” that glues its actin & myosin together. This “semi-permanent clench” deprives neurons of blood, further preventing updates We think with clenches; we set Bayesian priors with latches

I think this is already happening on twitter if you know where to look — but the right amount of legibility for these accounts is probably not 100%

Had a great time chatting with the @analoguegroup crew about vasocomputation, research taste, and building new idea-structures. Deep lore episode

Such compatibility is ultimately is about mathematical projections and symmetry — each semantic basin is a language-game which respects certain mathematical structures. Distill the structures and play the language-game at a higher level

My best guesses at describing what’s going on: 1. Neural component: Hearts have ~44k neurons, which isn’t a lot compared to the brain (~86b) or stomach (~500m) but a heart transplant literally is a (small) brain transplant 2. Muscle reflex component: “What is a memory?” is actually a fascinating question. The orthodoxy is that memories are synaptic configurations, with maybe some information stored in cellular state (eg RNA fragments). But vasocomputation suggests that many motifs of cognition are best understood as literal smooth muscle reflexes — “cognitive saccades” whose purpose is to collapse uncertainty in nearby tissue. The heart is the absolute center of smooth muscle in the body. So it would make a lot of sense that many of the particular smooth muscle motifs which apply compression pressure on our sensations would live there (and some would get transferred with a heart transplant). Potentially this could include learned VSMC reflexes and lead to regeneration of specific memories 3. EM scaffolding component: I also think the electromagnetic field is used to store and communicate various priors, which may reach to specific memories (there’s a good paper on this I need to track down again). The heart’s EM field is the strongest of the body’s and it’s not even close — if someone gets a heart transplant, they would get a bunch of this memory scaffolding from the donor 4. The body seems to store information ‘holographically’ — which means a little bit of the information is stored in many places, with substantial redundancy. When there’s damage, this allows the body to regenerate the original from surviving fragments. We’ve studied this in brain networks, and e.g. Michael Levin et al’s work on how flatworms can regenerate learned behavior even after getting their heads cut off But in general, insofar as the above items (1), (2), (3) allow fragments of memories to tag along during a heart transplant, I find it plausible that the new system regenerates some donor memories. It would be interesting to see if this happens only with the heart (pointing to (2) and (3)) or if you get similar results with kidneys, with bone marrow, etc if you look closely enough. (My guess is there would be a relatively large memory transfer effect with the heart, a small memory transfer effect with kidneys/liver/lungs, and no effect with bone marrow. Would be interesting to figure out how to structure a test of this)

Extremely good boy Imagine a superintelligence looking into your soul with its own inscrutable frame and deciding you’re a good boy/girl and then using superfine actuators to give you scritches that smooth out tension you yourself have no access to Game over

Something important here

Over time, there’s selection pressure on individual words to naturally de-reference and refer to the pointer, not the sensation. This leads to novel capacities, novel superorganisms that these capacities allow, and novel forms of predation from these superorganisms

“Man is born free [but schizo], and everywhere he is in [vascular] chains.” The moments of awakening are when these chains drop, but making yourself epistemologically virtuous enough to handle the freedom can take a long time and starts very boring

Now the challenge is how to take LLM intelligence and apply it back to data. This is a matter of 1) understanding precisely what makes for the highest-quality training data, and 2) applying synthetic intelligence to turn high-quality human tokens into higher-quality synthetic tks

The more predictive models you have online, the harder it is to increase your energy parameter The visual system is a core predictive pipeline; closing your eyes takes it offline & prevents it from doing compression sweeps. Energy & ambivalence build Same pattern in meditation

There’s a truly beautiful love language of “I will take your emotional problem within me, and solve it, and show you the solution by feeling” One of the most beautiful things humans do for each other This has a dark counterpart, “I don’t know how to solve this emotional knot inside myself, I will create it in others and watch how they fix themselves… either I will learn how to free myself or we will suffer together” This is much less beautiful

I think much of modernity is aimed at breaking the mind-body connection The brain is a big hallucinating LLM, the body is ASICs that keep it on track If the brain is alienated from its ASICs, it can be slurped up into some superorganism, which then provides the grounding

Likewise, if you can apply current LLMs to identify the worst 10% of your training data and either discard or improve that, the next generation gets a lot better. And you can iterate this

Plot for a SF novel: humanity rushes for the singularity, but physical constants start to change. Ultimately we find out we do live in a simulation and when we redline its compute, it enforces a local complexity cap by adjusting physical law away from anthropic optimal

Flirting with the universe

Tension has physiologically-distinct short- and long-term effects but there’s a unified grammar: tension prevents updating & reduces sensitivity to external patterns, acting as a “side-channel memory” about which part of the neural dynamic range to use

If you’re traveling for the holidays, consider taking nattokinase (a natural, broad spectrum clot dissolver) as a precaution against the microclots that often form while flying. It does a lot of other good things too. I’d rec 4000fu (usually 2 pills) at the start of the flight

“Psychological disorders are disorders of the morphogenic field” stays winning The Peat followup is that disorders of morphogenic fields are disorders of metabolism (“as energy flows through tissue, structure accumulates” & “energy and structure are interdependent at every level”) ‘Cellulite is a sign of energy body dis-integration — a fuzzing of one’s true shape, a novel type of damage specific to the chemical and spiritual conditions of modernity — and can be fixed through cutting seed oils, integration work, and prayer’

My @joinedgecity talk discussed two implications: 1. We build our inner worlds out of what we learn to stabilize & there’s important variation in this across people. Aphantasia is a skill issue: inability to flex vasculature in certain parts of the visual cortex