Imo, there isn't anything fundamentally special about the brain as a cognitive substrate. Even a pool of water could have cognition (reservoir computing), i feel that it's a general property that can emerge in any sufficiently high-dimensional physical system with enough internal degrees of freedom to explore many configurations, and enough coupling and stability to remain coherent, and enough plasticity for experience to leave persistent changes in its future dynamics.
And subjecting such a system to sustained external objective pressure...where some internal organisations perform better than others and it would progressively accumulate some useful structure about its environment.
The brain is just a really old artefact of billions of years of blind environmental weathering of matter being repeatedly shaped by the requirement that its host continue to exist as a coherent, persistent entity.
In short the brain is not fundamentally special in the principles that make cognition feasible.
What i'm trying to say in response to this article is that he's working at a far higher level of abstraction than is required. It's putting software before physics.
Also this somewhat chimes the same tune from the bitter lesson; it’s extremely non-trivial to reverse-engineer from the top down the residue of eons of evolutionary search. You’re basically trying to infer the process from this one absurdly overfit artefact it left behind.
What if its actions are completely determined by the rules of physics? I.e., it cannot do anything other than one thing, whether it applies thinking or not.
I think the question and article is fascinating. I do think that one should not only look at the genetic representation, and the algorithm a cell runs, but also the formalized environment that runs the simulation.
I’ve been thinking about this for a few years and I truly believe these cells need to be able to “move” in either a 3d lattice similar to a cellular automata or become nodes in some cyclic graph (or maybe both).
For the genetic representation, I had an insight that DNA/RNA looks strangely similar to SKI calculus (combinators) and have been using a Church encoding to translate the SKI program to an “action” per turn (move, connect, spike, divide, etc.). It has a nice property that the cell’s “program” and the spikes between them, and the input/output to the simulation itself can be the same thing (just a string of combinators).
Either way the author is spot on when it comes to this sort of thing having the properties 1) “always on” (no separate training mode), 2) local interactions, 3) everything stems from a genome, 4) each cell is its own little mini-program with the same genetics as everyone else.
All I can say is that evolutionary program is hard and takes either massive compute or large timescales to run the simulations. Fascinating stuff though
> The program must fit in a genome-sized instruction set of about 1 gigabyte
True but this can be a lot larger, maybe even a 1 exabyte instruction size since it depends on he programming language used to express the program. And even accounting for the invariance of runtimes at scale*, the constant factor might be gigantic since we have to build up a lot of cellular machinery first. Reminds me of the Carl Sagan quote: To make a sandwich,you must first construct the universe!
* For example, to translate a program from language A to language B, you can dedicate a constant size to write a language A to B translator. Thus at large program sizes , the kolmogorov complexity (i.e instruction set size) is fairly similar between programs since the size of the program dominates the size of the translator program which is constant. But the constant factor for cellular machinery might be gigantic**
If I crumble an assembled hot dog, I do not get nachos.
If the two were equivalent, performing the same action on each should produce similar results.
I suppose you could say that hot dogs, subs, hoagies, etc are kinds of tacos, but I can't see what element of the taco should give it this priviledge. It would seem, rather, that tacos, hot dogs, and sandwiches are all examples of the ursandwich where some (hot dogs et al) are members of the triwich family (being surrounded on three sides or containing a single continous substrate on three sides) and that some (burgers et al) are members of the duwich family.
This allows us insight into the great pizza debate. To the uninitiated, the thin crust and deep dish appear similar. We can see, however, that they are merely related cousins on seperate sides of the urwich family. One is a rare example of the unwich having bread on one side, while the other is a duwich. This easily explains the preferential division in humans. Some see no familial distinction while others care deeply as to which derivation of the urwich they are eating.
edit: on an entirely unrelated note, I will die on the 'cereal and oatmeal are soups' hill
One aspect missing from TFA is that in biological brains a lot of the training is performed during query time. If we assume 1 GB of DNA is enough to encode the brain's overall structure we still do need training data (e.g. visual/auditorial/tactile) to build out the strength of the synaptic connections.
Yep and most classical computer engineering approaches have rarely tackled it appropriately in my opinion.
Expecting a von-neuman-machine with at most viaion/audio encoded tin binary to replicate a quantum processing machine with multi spectrum inputs audio/visual/tactile/electromagnetic/olfactory encoded in ternary(ACGT) is only going to simulate a very basic image.
Not to mention the philosophy of the mind/brain dichotomy (including the gut brain[1]) that connects the physical/enviromental/emotional.
In the instance that the physical (silicon/photonic) recreation is not possible the next best direction would be learing how to program the DNA for wetware devices[2].
This isn't quite the question the article is asking, but if you expand it just a bit to a brain rather than a simulation of a biological brain specifically, the specification complexity of a modern AI is not that large. You can get a good idea of the foundation of the system with some 3blue1brown videos. While just watching those and deeply understanding isn't enough to score a job at a frontier AI lab, I'd expect that a lot of what the labs do is similarly not that complicated to specify... it's coming up with the ideas, tuning them, and then trying to test them at the now-staggering scale it takes to prove them out for commercial use that is most of the issue.
The complexity of the AI itself, the final weights, is vastly higher, because that incorporates all the data that was flowed through the "brain".
In a decent encoding you could probably still fit the initial state of the frontier models comfortably in just a few kilobytes of code-golfed code, including the update functions and every necessary for the actual training. Recall the specification complexity is the smallest program that can create the initial state, not the initial state itself; the specification complexity of "give me a trillion 64-bit numbers generated from this psuedo-random number generator" is on the order of that very English sentence in size, not 8 tebibytes of specification complexity.
Is the result a biological brain? Obviously not. Biological brains do not do what the LLMs are doing. But unlike someone asking this question 20 or 30 years ago, where "yeah, but what if some other architecture could work too?" was still largely hypothetical, and neural nets were still largely toys, now it isn't. We may argue if an LLM is as smart as a human but I would say that at least on its playing field it is quite clearly smarter than most biological brains in existence on most measures we care about. I have to qualify "on its playing field" because it is fundamentally a text completion engine, so for instance even very very tiny biological systems still beat it on things like "ability to drive an ant body around to do useful tasks". That's a separate field of AI right now. Stay tuned on that matter, but it's not how things work today for sure.
There is still something to what biological brains do that we are not matching; as I like to say, humans do what they do without the entire contents of the internet being poured through their head multiple times over. But the idea that maybe we don't have to exactly match biology to get something useful is no longer just a theory.
I have this science fiction story in my head that consciousness can evolve from organic to mechanical and back to organic as it is perhaps the most energy efficient method to encode complexity.
"Consciousness" is tick based and it does not require a massive human brain. Even the smallest of animals react in real time to stimulus. There's a filter that lowers the amount of tokens the next tick has to process. That's why you can watch a video and two different people will notice entirely different things about it. Forgetting is our brains culling what they process. Being awake eventually degrades that process. You can see this in every study on tokenized llms showing a decrease in accuracy as the context fills up. So too does your brain. That is why we sleep.
Short answer yes. We can build a brain.
As for quantum effects. It's just the machine. Not the mind itself. It's sort of like asking how a CPU works and then expecting that to inform your science on cognition.
> "Consciousness" is tick based and it does not require a massive human brain. Even the smallest of animals react in real time to stimulus.
Reacting to stimulus is a behavioral response. Consciousness is subjective experience. We know humans have subjective experiences (and suspect many other animals do), but not every stimulus reaction is conscious, and consciousness is not limited to perception (feelings, empathy, dreams, imagination, visualization, earworms, inner dialog, memories).
No, we don't. The humans claim that they do, but they may as well be philosophical zombies who claim to have such experiences without actually experiencing them (just as animals don't actually experience pain even if they behave as if they do).
Well, they certainly have behavioral response. But do they experience it? Descartes believed they didn't, most people today believe they do, but don't believe that web servers or LLMs are conscious; who knows what people a century in the future will believe?
The tick is at the top of an async jobs queue. The jobs provide the tick with constant updates. And this is not in real time. If you are injured say a sudden injury. You won't even notice until 2-3 ticks after.
Behavior is an emergent property of this. Not it's base.
What I'm actually talking about is decision making. This applies to higher and lower life forms. Even a bacteria will decide things from moment to moment. Well okay how does this happen? Well something in the background must provide new data to the "tick". Ticks don't rebuild on each turn. They get re-used. But that re-use is updated from sensory inputs, memory, and then is filtered to limit how many tokens each congitive tick is forced to process. Anyway consciousness is an emergent property of this system. We're just state machines ultimately.
Personally, I get a lot more milage out of "human brains are 100-1000 trillion parameter odes" than a discrete temporal model. There's probably some isomorphism but I find ODEs to be a much more natural model with less impedance to real world biological systems.
> LLMs suggest that human-like "cognition" can be somewhat mimicked by purely mechanical means.
This is not a strike against that hypothesis. Quantum systems can be simulated using non-quantum computers but that doesn't make non-quantum computers into quantum computers. Similarly, mimicking some aspects of human cognition doesn't mean you can replicate all of human cognition with the same method.
when mice were exposed to cosmic ray bursts (to test for effects on past and future astronauts traveling to moon and mars)
the mice learned much much more slowly and forgot things much much faster
how's that for "cognitive corruption" ?
(btw this is also why I laugh when Musk thinks he's going to Mars ever, it's not just a one-way trip, he wouldn't even survive the journey itself with current technology and unlike earth there's no magnetic field or enough atmosphere on Mars to deflect them)
> unlike earth there's no magnetic field or enough atmosphere on Mars to deflect them
These are solvable with known methods.
They make the whole thing rather less romantic.
If you're going to the trouble of e.g. burying your base under regolith or wrapping a planet in a big coil and lots of PV to make an artificial magnetic field, it very quickly becomes apparent that for essentially all possible disasters short of "oops we made a black hole and it caused a magnitude 15 earthquake and all the crust everywhere liquified and mixed with all the water so even ocean bases weren't safe", it's much, much easier to survive disasters on Earth without actually leaving Earth, than to go to Mars and have a normal day.
If someone made a robot factory that went mad and converted all the nitrogen in our atmosphere into ammonia (killing all the plants that no longer had access to the atmospheric nitrogen), while also producing enough smog to halve how much sunlight we got, and dumped a lot of endocrine disruptors into our water cycle, the top of Mount Everest and the Sahara desert would each still be much more hospitable than Mars is today.
It shouldn't be a surprise that it is very hard to keep humans alive outside the exact environment we evolved in. Current squishy humans will never colonize space, only intelligences using a vastly more durable substrate will be able to.
I do like the theory of using a massive amount of water to surround a ship as a shield, it solves a lot of radiation problems and something humans desperately need for long travel (with super-recycling) but not a solution once you get there
and we'll definitely have super-advanced space-drones, eventually
> ... effects on past and future astronauts traveling to moon and mars ... wouldn't even survive the journey itself with current technology ...
It's hard to tell given your phrasing but it almost sounds like you're insinuating that traveling to the moon is technically impossible, which would be quite a bold claim given the success of the Apollo missions. Apologies if I misinterpret.
LLM emulate cognition, there is a huge difference, it's a mimick process not genuine
if you read about psychopaths/sociopaths, they do not comprehend emotions but they can learn to emulate them by watching others and mimicking them
that's exactly what LLM is doing, it is taught to assemble words in a way that mimics all the humans in its huge datapool, so it comes off as "cognition" because that's what the humans it's emulating were doing
I don't see how this adds much to either a pro or con argument.
An inner monologue isn't something all humans have[0]. It is unclear to me how well (or poorly) Chain of Thought mimics intrapersonal communication. J-space probes (and other things) suggest there are, in fact, other aspects to LLM "thought" besides a CoT scratchpad[1].
However, regardless of how those questions pan out, qualia (of emotions in this case, but the problem is in general) is its own mystery, one for which we are still, so far as I can tell, unable to make progress with.
We kinda know what emotions are for, what function they serve in our genetic fitness, but why do they feel like anything? We do not know, they just do.
Without a falsification this theory is simply an extension of the age old dogmatic view that humans are somehow divine or superior to all other organisms, in a way that views were held before Darwin.
Everything is subject to quantum mechanics. If you meant something like a non-trivial quantum interaction that is nor modeled in the chemical and electrical theories of the brain presently, still, as far as I am aware quantum effects don't need anything beyond Turing to simulate. If I am wrong, someone let me know.
I think the elephant-in-the-room arguments against Penrose's quantum microtubules [0] are the complex behaviors exhibited by organisms without neurons hosting quantum effects. As a result, I think the more promising area is in bioelectrical explanations [1].
And subjecting such a system to sustained external objective pressure...where some internal organisations perform better than others and it would progressively accumulate some useful structure about its environment.
The brain is just a really old artefact of billions of years of blind environmental weathering of matter being repeatedly shaped by the requirement that its host continue to exist as a coherent, persistent entity.
In short the brain is not fundamentally special in the principles that make cognition feasible.
What i'm trying to say in response to this article is that he's working at a far higher level of abstraction than is required. It's putting software before physics.
Also this somewhat chimes the same tune from the bitter lesson; it’s extremely non-trivial to reverse-engineer from the top down the residue of eons of evolutionary search. You’re basically trying to infer the process from this one absurdly overfit artefact it left behind.
https://en.wikipedia.org/wiki/Natural_arch
It's doing all that it can.
The problem is in the word "can".
What if its actions are completely determined by the rules of physics? I.e., it cannot do anything other than one thing, whether it applies thinking or not.
I’ve been thinking about this for a few years and I truly believe these cells need to be able to “move” in either a 3d lattice similar to a cellular automata or become nodes in some cyclic graph (or maybe both).
For the genetic representation, I had an insight that DNA/RNA looks strangely similar to SKI calculus (combinators) and have been using a Church encoding to translate the SKI program to an “action” per turn (move, connect, spike, divide, etc.). It has a nice property that the cell’s “program” and the spikes between them, and the input/output to the simulation itself can be the same thing (just a string of combinators).
Either way the author is spot on when it comes to this sort of thing having the properties 1) “always on” (no separate training mode), 2) local interactions, 3) everything stems from a genome, 4) each cell is its own little mini-program with the same genetics as everyone else.
All I can say is that evolutionary program is hard and takes either massive compute or large timescales to run the simulations. Fascinating stuff though
True but this can be a lot larger, maybe even a 1 exabyte instruction size since it depends on he programming language used to express the program. And even accounting for the invariance of runtimes at scale*, the constant factor might be gigantic since we have to build up a lot of cellular machinery first. Reminds me of the Carl Sagan quote: To make a sandwich,you must first construct the universe!
* For example, to translate a program from language A to language B, you can dedicate a constant size to write a language A to B translator. Thus at large program sizes , the kolmogorov complexity (i.e instruction set size) is fairly similar between programs since the size of the program dominates the size of the translator program which is constant. But the constant factor for cellular machinery might be gigantic**
** Or not if we only need a rough simulation
You can make a sandwich with two slices of bread and whatever filling.
If I crumble an assembled taco, I get nachos.
If I crumble an assembled hot dog, I do not get nachos.
If the two were equivalent, performing the same action on each should produce similar results.
I suppose you could say that hot dogs, subs, hoagies, etc are kinds of tacos, but I can't see what element of the taco should give it this priviledge. It would seem, rather, that tacos, hot dogs, and sandwiches are all examples of the ursandwich where some (hot dogs et al) are members of the triwich family (being surrounded on three sides or containing a single continous substrate on three sides) and that some (burgers et al) are members of the duwich family.
This allows us insight into the great pizza debate. To the uninitiated, the thin crust and deep dish appear similar. We can see, however, that they are merely related cousins on seperate sides of the urwich family. One is a rare example of the unwich having bread on one side, while the other is a duwich. This easily explains the preferential division in humans. Some see no familial distinction while others care deeply as to which derivation of the urwich they are eating.
edit: on an entirely unrelated note, I will die on the 'cereal and oatmeal are soups' hill
A taco is a sandwich A hot dog is a sandwich A burger is a sandwich
There is such a thing as an open-faced sandwich, which has one slice of bread.
Asimov’s positronic brain was the conceptual framework with the wrong formula.
Expecting a von-neuman-machine with at most viaion/audio encoded tin binary to replicate a quantum processing machine with multi spectrum inputs audio/visual/tactile/electromagnetic/olfactory encoded in ternary(ACGT) is only going to simulate a very basic image. Not to mention the philosophy of the mind/brain dichotomy (including the gut brain[1]) that connects the physical/enviromental/emotional.
In the instance that the physical (silicon/photonic) recreation is not possible the next best direction would be learing how to program the DNA for wetware devices[2].
[1] https://my.clevelandclinic.org/health/body/the-gut-brain-con...
[2] https://www.impactlab.com/2025/10/18/the-wetware-frontier-wh...
The complexity of the AI itself, the final weights, is vastly higher, because that incorporates all the data that was flowed through the "brain".
In a decent encoding you could probably still fit the initial state of the frontier models comfortably in just a few kilobytes of code-golfed code, including the update functions and every necessary for the actual training. Recall the specification complexity is the smallest program that can create the initial state, not the initial state itself; the specification complexity of "give me a trillion 64-bit numbers generated from this psuedo-random number generator" is on the order of that very English sentence in size, not 8 tebibytes of specification complexity.
Is the result a biological brain? Obviously not. Biological brains do not do what the LLMs are doing. But unlike someone asking this question 20 or 30 years ago, where "yeah, but what if some other architecture could work too?" was still largely hypothetical, and neural nets were still largely toys, now it isn't. We may argue if an LLM is as smart as a human but I would say that at least on its playing field it is quite clearly smarter than most biological brains in existence on most measures we care about. I have to qualify "on its playing field" because it is fundamentally a text completion engine, so for instance even very very tiny biological systems still beat it on things like "ability to drive an ant body around to do useful tasks". That's a separate field of AI right now. Stay tuned on that matter, but it's not how things work today for sure.
There is still something to what biological brains do that we are not matching; as I like to say, humans do what they do without the entire contents of the internet being poured through their head multiple times over. But the idea that maybe we don't have to exactly match biology to get something useful is no longer just a theory.
Short answer yes. We can build a brain.
As for quantum effects. It's just the machine. Not the mind itself. It's sort of like asking how a CPU works and then expecting that to inform your science on cognition.
Reacting to stimulus is a behavioral response. Consciousness is subjective experience. We know humans have subjective experiences (and suspect many other animals do), but not every stimulus reaction is conscious, and consciousness is not limited to perception (feelings, empathy, dreams, imagination, visualization, earworms, inner dialog, memories).
No, we don't. The humans claim that they do, but they may as well be philosophical zombies who claim to have such experiences without actually experiencing them (just as animals don't actually experience pain even if they behave as if they do).
The tick is at the top of an async jobs queue. The jobs provide the tick with constant updates. And this is not in real time. If you are injured say a sudden injury. You won't even notice until 2-3 ticks after.
Behavior is an emergent property of this. Not it's base.
What I'm actually talking about is decision making. This applies to higher and lower life forms. Even a bacteria will decide things from moment to moment. Well okay how does this happen? Well something in the background must provide new data to the "tick". Ticks don't rebuild on each turn. They get re-used. But that re-use is updated from sensory inputs, memory, and then is filtered to limit how many tokens each congitive tick is forced to process. Anyway consciousness is an emergent property of this system. We're just state machines ultimately.
which would make it impossible to emulate via current technology
excellent PBS Space Time on that
* https://www.youtube.com/watch?v=xa2Kpkksf3k
- LLMs suggest that human-like "cognition" can be somewhat mimicked by purely mechanical means.
- Exposing the brain to an MRI would affect cognition if there was a quantum element to it.
This is not a strike against that hypothesis. Quantum systems can be simulated using non-quantum computers but that doesn't make non-quantum computers into quantum computers. Similarly, mimicking some aspects of human cognition doesn't mean you can replicate all of human cognition with the same method.
when mice were exposed to cosmic ray bursts (to test for effects on past and future astronauts traveling to moon and mars)
the mice learned much much more slowly and forgot things much much faster
how's that for "cognitive corruption" ?
(btw this is also why I laugh when Musk thinks he's going to Mars ever, it's not just a one-way trip, he wouldn't even survive the journey itself with current technology and unlike earth there's no magnetic field or enough atmosphere on Mars to deflect them)
These are solvable with known methods.
They make the whole thing rather less romantic.
If you're going to the trouble of e.g. burying your base under regolith or wrapping a planet in a big coil and lots of PV to make an artificial magnetic field, it very quickly becomes apparent that for essentially all possible disasters short of "oops we made a black hole and it caused a magnitude 15 earthquake and all the crust everywhere liquified and mixed with all the water so even ocean bases weren't safe", it's much, much easier to survive disasters on Earth without actually leaving Earth, than to go to Mars and have a normal day.
If someone made a robot factory that went mad and converted all the nitrogen in our atmosphere into ammonia (killing all the plants that no longer had access to the atmospheric nitrogen), while also producing enough smog to halve how much sunlight we got, and dumped a lot of endocrine disruptors into our water cycle, the top of Mount Everest and the Sahara desert would each still be much more hospitable than Mars is today.
I do like the theory of using a massive amount of water to surround a ship as a shield, it solves a lot of radiation problems and something humans desperately need for long travel (with super-recycling) but not a solution once you get there
and we'll definitely have super-advanced space-drones, eventually
Enough humans seem to be repelled by the idea of transhumanism that they may well insist on doing it the hard way.
It's hard to tell given your phrasing but it almost sounds like you're insinuating that traveling to the moon is technically impossible, which would be quite a bold claim given the success of the Apollo missions. Apologies if I misinterpret.
LLM emulate cognition, there is a huge difference, it's a mimick process not genuine
if you read about psychopaths/sociopaths, they do not comprehend emotions but they can learn to emulate them by watching others and mimicking them
that's exactly what LLM is doing, it is taught to assemble words in a way that mimics all the humans in its huge datapool, so it comes off as "cognition" because that's what the humans it's emulating were doing
LLMs don't have conscious processes that perform that kind of reasoning.
If anything the fact that they are blackboxes to their own behavior and are able to non-verbally empathise, is the opposite of what you describe.
Your requirment for something supernatural/quantum to decide between emulation and real thing is self referential.
If only we could tell either way, we'd have made a huge breakthrough in the philosophy of consciousness.
If it's not in the inner COT it isn't there.
An inner monologue isn't something all humans have[0]. It is unclear to me how well (or poorly) Chain of Thought mimics intrapersonal communication. J-space probes (and other things) suggest there are, in fact, other aspects to LLM "thought" besides a CoT scratchpad[1].
However, regardless of how those questions pan out, qualia (of emotions in this case, but the problem is in general) is its own mystery, one for which we are still, so far as I can tell, unable to make progress with.
We kinda know what emotions are for, what function they serve in our genetic fitness, but why do they feel like anything? We do not know, they just do.
[0] https://science.howstuffworks.com/life/inside-the-mind/human... and https://en.wikipedia.org/wiki/Intrapersonal_communication
[1] https://www.anthropic.com/research/global-workspace
https://en.wikipedia.org/wiki/Orchestrated_objective_reducti...
0. https://en.wikipedia.org/wiki/Orchestrated_objective_reducti...
1. https://en.wikipedia.org/wiki/Developmental_bioelectricity