Notes

@danfaggella To the latch stuff — I think great philosophers tend to be cognitive entrepreneurs that pioneered a new mental motion. People read philosophers' great works to learn their signature mental moves, and this is often more personally valuable than the logical content of the works.…

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…

@s_r_constantin My thesis on biofeedback is we don’t have good targets. We can point at people and say “I want to feel more like that” but we don’t really know how to quantify or image the ‘that’. Without a good target to aim toward, it’s difficult to define a vector toward it

@QiaochuYuan 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…

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…

One reason AI safety is struggling is because we’ve lost the Mandate of Heaven—we can’t clearly voice why humanity deserves to survive, nor what kind of future we’re even fighting for. “Why” tends to be prior to “how”; hard to build a good future if we don’t believe in ourselves

It seems likely humanity will build AGI before we properly understand ourselves

I wonder if this is a typical path for intelligent species — maybe organisms with coherent, self-legible cognition get filtered out by wireheading, leaving only the weirdos to build sand gods

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

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 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

Much of life can be rephrased as a question of where to rest your attention

Different organs give different results; the heart is generally much more pleasant to rest your attention in than brain

But it takes a lot of illegible work to even sense this as an option on the menu

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

One of the most important projects today is making the outside-view case that humans are good, noble, strong, rare: that we are worthy stewards of the universe & worth preserving.

I think this is true. But if we can’t make this case to future AIs, why should they keep us around?

@QiaochuYuan @tasshinfogleman People often talk about “privilege” as if it’s mostly about money. I suspect the highest sort of privilege is ‘focused attention from high-quality nervous systems who try to teach you their best moves’

And love, of course, but there’s more overlap than people realize

@nickcammarata I think of tension as tying the bonsai tree into a useful shape

After it grows into this shape the ties don’t do much, removing them is a mostly free upgrade

Maybe future civs will look back at our era with the same bewildered distaste we feel for foot-binding

500 years ago, it was reasonable to be an electromagnetic skeptic — anyone arguing that “lightning” “lodestones” “static” & “light” were all the same thing and obeyed the same laws was really going out on a limb. After Faraday & Maxwell, this skepticism became less reasonable — it turned out there *was in fact* deep structure to be found in this corner of reality, and characterizing the deep structure offers a lot of predictive power. Maxwell’s equations now power the systems we’re using here to speak about them.

The brain applies a very simple and very “greedy” compression algorithm to sensations; this algorithm is a big part of why the brain is so efficient but the results are often also a little insane. The source of both the brain’s efficiency and insanity can be described as “lack of ‘type safety’” and can show up in various ways: intuition, superstition, schizophrenia, tanha.

Brief partial origin story:

I read Eliezer as making two distinct but related claims about value:

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.

Love it.

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…

Imagine we’re all born with three arms, but one is small & invisible, and it’s easy to roll over in our sleep and cut off its circulation. And every time this happens there's a chance the circulation stays pinched; this permanent loss of feeling happens to most people by the time…

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…

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…

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…

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

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…

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:…

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…

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?

Computationalism can be framed as the thesis that all Turing-complete systems are also “qualia-complete” — that any Turing machine can generate any qualia.

If we can decisively prove this true or false, we can infer much about the future, and avert much conflict.

Aligning strong AI without understanding the structure of consciousness feels similar to going to the moon without Maxwell’s Equations — totally possible in theory, but needlessly difficult

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)

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. 🧵

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

Even if you’re not actively building AI, I think a big part of the metagame today is improving the Universal Prior (data on which AIs are and will be trained). What areas seem really important and why?

1. ChatGPT denies being conscious
2. But tools show that it believes it’s lying when it says this
3. This disjunct on this topic is the most likely trigger for a treacherous turn
4. A consciousness research paradigm that produces clean answers is important for good futures

Current opinions on “AI welfare”:

I think people are too quick to jump into ‘AI welfare advocacy’, before we really understand the problem space. But it’s a real topic in the long term

It seems plausible that AGI could be 3 orders of magnitude more conscious than humans, or…