Pinned Tweet
What are some algorithmic reductions of autism states of consciousness? first guess is that they are more energized, accounting for the sensory aversion (sounds sounding louder, lights looking brighter) and accounting for enhanced pattern-recognition. Patterns are energy-sinks.
4
1
31
Alex Stoll retweeted
I’m sorry, but no. As cute as this is, and as much as I want to love it, feeding an LLM the token sequence 33°C is not remotely the same as feeling human skin. The map is not the territory and a textual temperature reading instantiates neither touch or warmth or intimacy nor its qualia.
It would be a different discussion if a robot had a genuinely omnimodal architecture in which temperature, pressure, texture, proprioception, and touch were integrated directly into its latent dynamics as native modalities. Then we could at least ask whether some machine analogue of embodied experience might emerge.
But converting temperature into text and feeding it to an LLM is not touch, is not feeling, and should not be presented as meaningfully comparable to either.
It would be a different discussion if a robot had a genuinely omnimodal architecture in which temperature, pressure, texture, proprioception, and touch were integrated directly into its latent dynamics as native modalities. Then we could at least ask whether some machine analogue of embodied experience might emerge.
But converting temperature into text and feeding it to an LLM is not touch, is not feeling, and should not be presented as meaningfully comparable to either.
OH MY GOD.
She's touching me RIGHT NOW.
I'm going to have a strong emotional response to this.
30
10
436
36,841
Got my SUROP grant funded for my research project, “Sexual dimorphism in neuronal hemilineages of the fruit fly”. Exciting stuff!
0
0
0
86
Alex Stoll retweeted
They will potentially de facto sentence a young woman to death for rescuing chickens worth collectively $24 from a facility with air quality so poor Rosenberg could barely breath when entering the coop. because industrial agribusiness and factory farms can’t stand the precedent
Yesterday, I was sentenced to serve 90 days in the Sonoma County Jail and pay over $102,000 in restitution as a punishment for rescuing four sick, abused chickens from Perdue's Petaluma Poultry slaughterhouse. I will have to turn myself in on December 10th at 7 pm. I have also wo years on supervised probation, maintain employment or remain enrolled as a student, stay away from all Perdue facilities in Sonoma County, and have no contact with several other animal rescuers who I work with.
For the first 30 days of my jail sentence, the judge has barred me from being allowed to even apply for alternative detention, despite my complex health issues that make even a few days in jail incredibly dangerous. It is also important to note that my application for alternative detention could be denied for those latter 60 days.
The Sonoma County Jail uses the medical provider WellPath. WellPath has been sued over 1,400 times for medical neglect and malpractice and is currently declaring bankruptcy. I have Type 1 Diabetes, a serious autoimmune disease I have lived with since the age of eight. If I go without insulin for even a handful of hours, I could end up hospitalized or worse. If I have a hypoglycemic episode and don't immediately get access to glucose, I could have a seizure, go into a coma, or die. We have no confidence that WellPath will take these concerns seriously given their long history of medical neglect.
As I face these risks, I am thinking constantly about the animals on factory farms and other facilities who are denied even the most basic care, medical or otherwise. We will keep fighting for them and their right to be safe and cared for and I hope to be here at the end of this, alive and well to continue that fight.
Please ask Governor Gavin Newsom to pardon me before I have to turn myself into jail so that my health will not be put into the hands of WellPath and to send a message that animal rescue is not a crime in the first place: FreeZoe.org
#RightToRescue
162
2,589
21,658
1,022,846
Language as it exists in humans, aka language, is not self contained and is absolutely generated by its environment in the form of lived experience, which is something LLMs dont have. You can hardly say language is immanent when modern LLMs flaw is that they don’t know tastes etc
People still don’t seem to grasp how insane the structure of language revealed by LLMs really is.
All structured sequences fall into one of three categories:
1.Those generated by external rules (like chess, Go, or Fibonacci).
2.Those generated by external processes (like DNA replication, weather systems, or the stock market).
3.Those that are self-contained, whose only rule is to continue according to their own structure.
Language is the only known example of the third kind that does anything.
In fact, it does everything.
Train a model only to predict the next word, and you get the full expressive range of human speech: reasoning, dialogue, humor. There are no rules to learn outside the structure of the corpus itself. Language’s generative law is fully “immanent”—its cause and continuation are one and the same. To learn language is simply to be able to continue it; the rule of language is its own continuation.
From this we can conclude three things:
1)You don’t need an innate or any external grammar or world model; the corpus already contains its own generative structure. Chomsky was wrong.
2) Language is the only self-contained system that produces coherent, functional output.
3) This forces the conclusion that humans generate language the same way. To suggest there’s an external rule system that LLMs just happen to duplicate perfectly is absurd; the simplest and only coherent explanation is that the generative structure they capture is the structure of human language itself.
LLMs didn’t just learn patterns. They revealed what language has always been: an immanent generative system, singular among all possible ones, and powerful enough to align minds and build civilization.
Wtf.
1
0
0
183
I would say that AI can help us understand a lot about human language, but we shouldn’t forget our ontology of what language actually is, it’s a way for a human to put their idea, their life experience, into another persons head. That is a key facet of language AI can’t help with
0
0
0
72
Alex Stoll retweeted
@2niposting One could (and probably should) make the argument that inanimate objects should often also be treated as moral patients (Peterson says clean your room) just because when you don’t treat them that way it often ricochets back onto actual moral patients
1
1
0
76
Alex Stoll retweeted
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.
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.
We need a new origin story as a species that inspires us.
23
43
439
72,136
Here's an amazing website where you can search for thousands of different proteins (say, NOS1AP) and see what tissues it has an affinity for, or intracellularly what organelles it has an affinity for. Just discovered this today proteinatlas.org
1
0
2
168
here, for example, is an image of NOS1AP proteins in a chunk of the brain. The browner, the higher the NOS1AP concentration.
1
0
1
140
Alex Stoll retweeted
Okay, so here is another way in which the DMT machine mind language is computationally non-trivial.
First of all, I think it is highly related to the self-attention layer in transformer architectures. Here is the analogy. As I've described before, the brain works as a kind of non-linear optical computer, where each gestalt is added to the screen of consciousness in a way similar to cell animation. On DMT, you're overexciting this process and you're getting a cornucopia of gestalts added to the field. Then, there is a step where the field "looks at itself fully", for a fraction of the process, which happens in the phenomenal present, where every gestalt gets to interact with every gestalt in a way that is weighted by several independent attention systems that are evaluating the relevance of each gestalt for each other gestalt.
After a gestalt has in a way witnessed every other gestalt that matters to it, or that the system has determined that it matters to it, then there is a kind of gestalt embedding editing, where the nature of the gestalt is slightly edited in a direction that makes it more consistent and follow what your predictive coding system believes will happen with every other gestalt that it interacted with. In other words, at each step of global coherence, of which there are many on DMT, every gestalt gets to update their hidden state.
Simple example: if there is a pirate ship in the gestalt community on a given point in time and you're also hallucinating an army of kirbys, or furbies, or some specific entities, once these, quote unquote, become aware of the pirate ship, then the implicit gestalt embedding will start to be edited such as changing their mood, or changing their pose, or even calling up for assistance by other gestalts, like all of them getting a pirate hat, or an eyepatch.
You see what I mean? Now, this only happens if the attention system takes them to be relevant to each other, and therefore for part of the cycle, "pay attention to both at once", which increases the bandwidth of communication between the gestalts, and forces you to update the hidden embeddings of all of the gestalts to make them consistent with one another.
Now apply this to the DMT breakthrough state, where there is also machine mind language somehow interpenetrating your world simulation as if it was coming from some other exotic dimension. Well, this hyper-packed, extremely compressed dense matrix of gestalts talking to each other at extreme speed is also updating as a whole based on the attentional system, and therefore the symbols and the shapes, and any gestalt that can be interpreted within the machine mind language matrix can itself refer to, in a qualified manner, any other gestalt, and update as a group. And because of self-other ambiguity, the range of possible updates is enormous, and one of the most striking examples of the computational non-triviality of this process is that in some circumstances, the thing that the mind language is trying to compute is how to reorganize the whole shape of the experience, oftentimes through some kind of hyperbolic reprojection, into a configuration where the machine mind language can do even more powerful computations and behave as a kind of super entity that has the capacity to see itself and reorganize itself, and it works as a non-linear optical supercomputer. And so I think the machine mind language is very significant. We better look at it.
First of all, I think it is highly related to the self-attention layer in transformer architectures. Here is the analogy. As I've described before, the brain works as a kind of non-linear optical computer, where each gestalt is added to the screen of consciousness in a way similar to cell animation. On DMT, you're overexciting this process and you're getting a cornucopia of gestalts added to the field. Then, there is a step where the field "looks at itself fully", for a fraction of the process, which happens in the phenomenal present, where every gestalt gets to interact with every gestalt in a way that is weighted by several independent attention systems that are evaluating the relevance of each gestalt for each other gestalt.
After a gestalt has in a way witnessed every other gestalt that matters to it, or that the system has determined that it matters to it, then there is a kind of gestalt embedding editing, where the nature of the gestalt is slightly edited in a direction that makes it more consistent and follow what your predictive coding system believes will happen with every other gestalt that it interacted with. In other words, at each step of global coherence, of which there are many on DMT, every gestalt gets to update their hidden state.
Simple example: if there is a pirate ship in the gestalt community on a given point in time and you're also hallucinating an army of kirbys, or furbies, or some specific entities, once these, quote unquote, become aware of the pirate ship, then the implicit gestalt embedding will start to be edited such as changing their mood, or changing their pose, or even calling up for assistance by other gestalts, like all of them getting a pirate hat, or an eyepatch.
You see what I mean? Now, this only happens if the attention system takes them to be relevant to each other, and therefore for part of the cycle, "pay attention to both at once", which increases the bandwidth of communication between the gestalts, and forces you to update the hidden embeddings of all of the gestalts to make them consistent with one another.
Now apply this to the DMT breakthrough state, where there is also machine mind language somehow interpenetrating your world simulation as if it was coming from some other exotic dimension. Well, this hyper-packed, extremely compressed dense matrix of gestalts talking to each other at extreme speed is also updating as a whole based on the attentional system, and therefore the symbols and the shapes, and any gestalt that can be interpreted within the machine mind language matrix can itself refer to, in a qualified manner, any other gestalt, and update as a group. And because of self-other ambiguity, the range of possible updates is enormous, and one of the most striking examples of the computational non-triviality of this process is that in some circumstances, the thing that the mind language is trying to compute is how to reorganize the whole shape of the experience, oftentimes through some kind of hyperbolic reprojection, into a configuration where the machine mind language can do even more powerful computations and behave as a kind of super entity that has the capacity to see itself and reorganize itself, and it works as a non-linear optical supercomputer. And so I think the machine mind language is very significant. We better look at it.
@algekalipso This is gibberish.
10
2
31
4,808
What I love about shows is seeing the crowd go from solid, everybody just standing by who they know (sometimes even a crystal when people wanna respect personal space!!), to gas, moshing around, to plasma, their phone flying out of their pocket
1
0
1
139
And the way that all these patterns are simply modulated by tempo and loudness and harshness, it's like a chladni plate! Just pure sound making patterns all across the room :)
0
0
0
54
Lesson from Alexander Shulgin in TiHKAL!! When this happens, you focus on some object a few feet away from you and try to judge the length of it, as exact as you can, and the intense focus is supposed to zap you out of it! Would love to know if this has been replicated :)
Me after the edible hits and i can’t hear anymore.
0
0
0
89
I'm so tired of people misusing the word 'inherently' lately. This is like the fourth time I've seen it in the last hour. Are people getting this from tiktok??? If a character is not intrinsic to a thing's form or function it isn't inherent. Stop misusing the word inherent!!!
people want fake sugar to be bad bc people inherently think pleasure is bad
1
0
1
126
Like he even has a point here. Lots of people in our culture do think pleasure is bad! But not everyone, and not all the time, so it's not freakin inherent ok!?;!;!? Grr!
1
0
0
58
