Can intelligence be separated from the body? | Hacker News Reader

Can intelligence be separated from the body?

nytimes.com

53 pointslxm78 comments

> Boyuan Chen, a roboticist at Duke University who is working on developing intelligent robots, pointed out that the human mind — or any other animal mind, for that matter — is inextricable from the body’s actions in and reactions to the real world, shaped over millions of years of evolution.

I'm a little surprised to learn that some people believe intelligence can be separated from the body. Intelligence didn't develop in a vacuum, then get poured into the body — it developed both evolutionary and in the case of you personally as a result of being a body in the world. It isn't something you can separate from our senses, or the physical world we inhabit. Anything else feels like mind-body dualism to me. That, or a science fiction model of the brain as being functionally like a squishy computer chip, just waiting to have programs uploaded into it.

Maybe not intelligence with conceptual capabilities but what about direct experiencing or witnessing? For every one of us, animal or human, we start primarily as an experiencer of reality, then as we grow we build a model of the physical world and how we relate to it. But that "subjective" experience is primary, and our model of reality, materialist or otherwise, is conceptual and inferred and secondary, including the idea that our firsthand experience and bodies are one in the same.

Things like AutoGPT and BabyAGI seem intelligent to me.

Intelligence is defined as the ability to acquire and apply knowledge and skills.

How could it be denied that things like AutoGPT can acquire knowledge and apply skills?

Not sure why you’re talking about side projects like AutoGPT when you could be referencing chatGPT/GPT4 with the same arguments you’re making.

I don’t think the same is as true of LLMs unless they’re put into a system (that can acquire knowledge and skills).

The problem is that it's absolutely something you can sever from specific senses or specific parts of your body. If I were to go blind, I suspect I would be the same person afterward (or at least no more different than I would be had I retained my sight). If I lost an arm, I'd still be me, just a me controlling one less arm.

I guess my point is that if intelligence is really inseparable from the body, you'd expect that it would be impossible to keep a head alive without a body- not for any biological reasons, but because the patient would become abruptly...well I'd like to say braindead but apparently there's some other source of intelligence. And you can debate about whether a sci-fi "upload" of your brain is still you, but if you're going to say it wouldn't be able to think at all then you're the one proposing mind-body dualism.

edit: Of course the obvious counterargument is that plenty of people get minor to moderate brain damage while plausibly remaining the same person. I don't think I'm willing to bite that bullet, so my best pitch is that "me" is as made up a category as every other, so we should expect some weirdness.

> you'd expect that it would be impossible to keep a head alive without a body

Is the head not part of the body?

Here's an interesting reality -- there is no hard dividing line between where the brain stops and the rest of the body starts. It's a bit like trying to say where the atmosphere stops and space begins. And we have neural masses around the heart and stomach that engage in independent activity. Some neuroscientists argue that it's reasonable to say we have three brains.

I'm afraid there's actually a fairly firm line. The neural masses around the gut (and to a much lesser extent heart) are entirely different in structure from the cerebral grey matter we know we think with. It is however consistent with the rest of the peripheral nervous system, because that's what it's a part of. Many of the neurons in the gut are simply sensory neurons doing no processing, and the rest aren't far from what you'd expect to coordinate the various systems.

That's why even though every area of the brain has a cognitive deficit associated with its removal, there is nothing similar for the gut or heart (besides the obvious issues). Mere neurons do not a brain make.

Objects are not characterized by unchanging material substance. Objects are instead the holding together of causes and conditions, meaning that they’re characterized by duration / time.

Assuming I understand what you're saying here, granted. But that doesn't make lead gold, and it doesn't make the large amount of neural mass in the gut a brain.

I actually agree with you mostly. You probably don't need the neurons in your gut to be your whole self.

But the point was more to explore what "whole self" really even means. Technically your brain is composed of many modular components. What decides the line for things that have to be included to be you and things that can be excluded? Is it just the close proximity?

IIRC, the neural tissue around the gut is near identical to the brain tissue.

Which bits? I don't have my old notes on hand (I've since switched to less useful and more fun fields), but as I recall the particular structure characteristic of the cortex (really the telencephalon overall) isn't found anywhere else. The gut isn't just sensory stuff, but neither is the eye. There's nontrivial processing done pre-brain in a number of places, but there's no reason to think that cognition is happening elsewhere.

I'm not knocking the gut, and I totally get why some neuroscientists talk about its complexity. I actually did some in depth research on it a while back (my advisor had done some related work and I was considering writing a survey paper), and there's a lot of neat stuff going on. It reminded me a lot of early cybernetics peoples' dreams of future factories run by vats of bacteria or something. If you imagine all the various steps of the digestive process, and the production of various chemicals in carefully controlled amounts, you can see why you'd want some serious hardware. Not to mention that there are a large number of muscles which aren't consciously controlled and need to respond quickly to changing circumstances.

I do not think cognition is sufficiently understood, to locate it solely in the cortex. The word is even not precisely defined.

I guess what I'm trying to say is that you can pick anything to fall under the umbrella that you want and one of two things will be true; it will happen almost entirely in the brain, or it will also apply to a thermostat/ChatGPT, which means it certainly can't be used to agree with the article above.

If we're making specific and unqualified claims about the possibility of "intellegence" without a "body", which is what this thread is about, then there is some working definition at play. I think that under the working definition the article implicitly uses, their argument is extremely dumb. You can pick any definition you want- I certainly can't stop you! But as long as we're arguing about what is and isn't possible without being "embodied", it really doesn't matter. If your definition of cognition is even broader than mine, you should be even less confident that the random researchers the NYT interviewed have correctly identified the exact conditions under which it is possible.

> If I were to go blind, I suspect I would be the same person afterward (or at least no more different than I would be had I retained my sight). If I lost an arm, I'd still be me, just a me controlling one less arm.

At risk of getting philosophical, you would not be the same person afterwards. You'd be You Without Sight, but you were You With Sight before you went blind. I wouldn't expect a drastic change (your taste in music or political views may not change), but I think is fair to say you would be changed somehow.

I believe that who we are, as a person, is an emergent property of our entire body as well as our environment.

If a person's identity is a property, is it measurable?

Perhaps as measurable as the quantum state of the universe itself. After all, “environment” includes the universe.

So realistically, probably not. That’s why I felt it necessary to say “I believe.”

Look, I love getting philosophical. Really, I do. My friends are so used to it that when I enter a conversation with "Well you might say..." they immediately start throwing objects at my head. Let none say that Ravi has ever backed down from an opportunity to get jiggy with definitions.

But.

Are you 100% confident that the difference between me having a brilliant idea for a fun game with my laser pointer and the me 5 minutes later experiencing a clarity of hindsight my natural vision will never again supply me is remotely comparable to the difference between, say, me and you? Or even me two years in the future? There is obviously a working definition of "same person" here!

Like listen, you can absolutely say that we aren't the same person in the morning that we were before going to bed the previous night. Or going through anesthesia. Or forming new memories, or existing at all as the persisting pattern of our bodies churns through the atoms momentarily instantiating it. No man steps in the same river twice, yada yada yada.

But the article is not making a philosophical argument about what the self truly is. It is positing (or raising the question in an annoyingly leading way, whatever) that intelligence is embodied definitionally. Like literally impossible without specifically the meat grubbers evolution endowed you and I with. There is absolutely no way to go from "we are all one universe" to "no machine learning system will ever be truly intelligent until it's within my chosen specialty of robotics", which is what 14 roboticists said in unison when Times interviewed them. So even if we need to invent new words for the sense in which I am the xchnople as a blind me, so you can keep "same" for whatever you're using it for, there's really no way in which it can swing the question of whether a body is required for specific, empirical aspects of cognition.

I’m about to go to bed so I won’t (just yet) give your comment the response it deserves, but as I read one question struck me: what are the empirical aspects of cognition? Is cognition not a function of life itself?

I guess again that there are senses in which everything alive is doing cognition. Some of these are also senses in which a thermostat is doing cognition. But it remains somewhat orthogonal to the point being made in the article! You are arguing for a definition of intelligence far more expansive than I am- it would be even more surprising under your definition if it was restricted to agents that have what we would recognize as bodies.

Even losing a finger can result in heavy phantom pain.

I have a friend who losed a part of his leg, and he complained for many years about pain.

I think if you separate all peripheral nerve connections from the brain, it will shut down immediately.

But that's all the more evidence of cognition in the absence of physicality. Phantom pain is a result of signals leaking from adjacent areas of the brain to the ones which once handled senses originating from that region of the body. And you sure enough feel something, because you don't feel with your fingers. You feel, and indeed experience every other sense, only in your brain.

Yes. There are two different questions:

Can a brain be constructed in a rather isolated state?

Can a natural brain survive when separated from its periphery?

I think the second is not possible. The first may be possible, but not in the near future.

Oh granted. Maybe head transplants are on the table, but certainly a brain without a peripheral nervous system filling it with data would go into shock immediately. Even if not, there isn't a less pleasant experience I could imagine.

I don't belive weren't closer to sentient intelligence than we were when babylonians started counting on tablets.

But it seems pretty intuitive that intelligence can exist outside the body. If you suddenly lost your body and had your brain in a computer that could receive audio and video input, couldn't you still act? What if you had no senses and just had info fed into your mind for reflection? I understand the idea of "embodied intelligence" and that sensory awareness is a big part of our intelligence, but I don't see why it would need to be mandatory to be intelligent.

Three thoughts: first people in hospitals constraint to mostly their bed loose IQ points by the day lacking physical interactions, second I'd assume it's about developing intelligence and third I can easily imagine the Internet and its technical restrictions plus its users and their behaviour as some kind of body, useful to let a AI train itself on and achieve goals.

But actually I fear the possible outcomes.

> What if you had no senses and just had info fed into your mind for reflection?

Metallica - One.mp3

Serious answer: you'd hallucinate like crazy. A brain deprived of sensory information starts generating its own. Over long enough periods of time, like the protagonist of "One" you might go mad and wish for someone to end it all.

The key is to figure out what features of our history are critical and which are accidental when it comes to intelligence. Seeing as our access to the outside world, and even our own body, is mediated by way of electrical impulses that conceptually come apart from our body suggests that embodiment is not crucial.

The most interesting phenomena tied to this to me are various unconscious/semi-conscious states like sleep or anesthesia. Clearly we have the ability to induce these using simple chemicals. Also from medical data we know what happens when brains are damaged or altered.

Counting also also didn't develop in a vacuum, then get poured into the body; counting was developed equally evolutionarily and equally in response to physical interactions of a person to the world. But as photosynthesis works too outside the cell, provided exterior force and the motive to produce it there when the bio-evolutionary one is taken away, so too are algorithms a real thing that have real properties distinct and separable from the mysterious vital energies that housed their evolution.

I remember a Hackernews wondering if people who were "born paralyzed" might have intellectual disabilities as well, and the answer appears to be yes. One of the things that came out of Papert's Logo research was the benefits to kids with severe motor disabilities: by setting things up so they could drive the turtle around, they had much better grasp of spatial relations between objects than they otherwise would. Being able to manipulate a proxy in the real world (the turtle) allowed them to explore those spatial relations in ways they couldn't with their physical bodies and that's critical to our cognitive development.

There's this fun little article here proposing the spider web as an external extension of intelligence for the spider: https://www.quantamagazine.org/the-thoughts-of-a-spiderweb-2...

I think it's worth distinguishing intelligence from consciousness (and sentience).

Well that’s like saying a model needs all kinds of inputs and training data but once it is trained it’s not portable. That’s just not true. Post training the intelligence is separable.

> Anything else feels like mind-body dualism to me.

It seems you misunderstand what "mind-body dualism" means then. Mind-body dualism is basically a supernatural belief that the mind exists independent of physical reality. For example, dualists believe that an AI, which is based on physical computing systems, can never possess a mind. Believing that intelligence can arise from a physical process (e.g. a computer chip) is the opposite of that.

It's unclear to me whether you actually subscribe to the dualist view but someone who believes that intelligence can arise from a computer chip does not.

Not necessarily, because functionalism where the substrate doesn’t matter can be viewed as a form of dualism. Particularly if we’re talking about animal intelligence and mind uploads. You can see the more extreme forms of this sort of dualism in sci-fi when minds are swapped between bodies, or someone’s mind is transferred to a computer or robot.

Also, dualism doesn’t require supernaturalism. It just requires there be more than one kind of fundamental stuff. There’s no logical principle saying nature has to be only one kind of substance or process. Many modern philosophers who are dualists are also naturalists. Chalmers for example, is a property dualist regarding conscious states (qualia).

Also, one can be a neutral monist, believing that mind and matter emerge for a neutral substance that is neither. Maybe a Kantian would defend something like that? Anyway, philosophy has many options regarding dualism, mind and matter which aren't supernatural. You can even be an idealist and a naturalist. You just think the physical world is material ideas in minds.

It seems you are using "dualism" to mean something else than what it means in the "mind-body dualism" context. The "dualism" in mind-body dualism is supernatural by definition. It's a claim that there exists something beyond nature ("supernatural") that exists and explains the mind. It's essentially a religious belief with no scientific basis. Dualists do not generally believe that AI is possible because they claim it would be missing this "supernatural" property which human minds possess.

That's religious dualism, but you're not taking into account the full range of philosophical debate. There certainly are dualists who do not invoke the supernatural, particularly with regards to consciousness. But in this case, the claim is that functionalism can serve as dualism, when it treats the body as irrelevant to the functioning of the mind. I gave you examples from science fiction. Even the idea of mind-uploading could be considered functional dualism, as it assumes there is something mental which can be copied from the neural/biological, implying the mind is separate.

Science consists of predictions based on measurements of our perception as mediated by our senses or tools that augment them. Dualism (or idealism for that matter) doesn't require religious beliefs. You can arrive at them analytically. Science is great for explaining the conventional reality of things as they appear though. But until I read a convincing explanation for consciousness and subjective experience that doesn't reduce to "neural correlates" I can't get on board with physicalism.

The fact that we're not all unconscious automatons that just evolved from one long cause-effect chain stemming from the big bang to now continues to bewilder me. "Mind" may play an "of the gaps" game here but it's the only reasonable explanation that currently makes some semblance of sense to me.

So currently I agree with the religious on your last point. Machines or AI cannot be conscious. Maybe if they were infused with some sort of biological material though I would say maybe. If our subjective experience is a localization of a thing called "consciousness" and certain forms (humans, animals) localize it to produce a sense of self then why not.

> I'm a little surprised to learn that some people believe intelligence can be separated from the body.

Just look at politicians for examples. /s

As technology progresses, it will probably be possible to scan human brains and make digital copies. With the digital copies, and further study, at some point humans will likely fully decode and know exactly how brains work.

At that point, there isn't any magic or mystery. You can create androids who are as equally intelligent as humans. Maybe we will place some limits or restrictions on them, as safeguards, but no telling how that will work out as time goes by. We might not be too far away from such a day, relatively speaking. Maybe it will take another 25 to 50 years. But it looks like we will get there, if we don't self destruct before then via nuclear war or some other catastrophe.

I don't know if intelligence can be separated from the physical body, or the other way around. Maybe the body emerged out of intelligence. Like nails emerges form the skin, so is the physical body. Maybe it took millions of year of evolution for that intelligence to build a physical body capable of sustaining and expressing itself. Maybe it is not yet done.

Just a thought experiment.

Do you have any proof that it can't be done? Because otherwise you are making an opinion.

It's like saying that people can't fly. Maybe we cannot do it with wings, but we do it nonetheless.

But we have only started this age of more and more knowledge over the last 100-150 years, and it is logarithmic and the curve is getting steeper and steeper with each year. Who knows what will be possible in another 150 years? As far as I'm concerned, I'm 100% convinced that by 2050, you'll be able to put your brain on a chip and we will use robots and inhabit robot bodies to travel around, just download our program into the robot's body.

To what extent is our intelligence a consequence of our bodies, evolution? Would a computer have intuition? Fight or flight responses? As a species, we're still debating which mammals are "intelligent" and they share most of our circuitry and evolutionary history. Will we be able to recognize "intelligence" in another form? Or flipping it around, will it recognize our intelligence? Or will it see us much like we see a chimpanzee colony?

Why shouldn't it be possible to simulate a full human mind, say, at atomic level? What does the intelligence/consciousness need that's impossible to simulate in software? I can't think of many reasons, besides perhaps true randomness?

> Why shouldn't it be possible to simulate a full human mind, say, at atomic level?

Who's designing and running the simulation, on what infrastructure?

Its a thought experiment, so whatever is required. Need a dyson sphere using a stars total energetic output for calculations? Need to network that? Fine, thats an engineering problem.

But in this case "whatever is required" to build an atomic level simulation is the ability to observe and understand the simultaneous state changes of an octillion atoms well enough to write an emulation of them in some other medium (which is arguably a "body" of sorts anyway). That seems like less of an engineering problem and more of a "do you believe in gods?" problem.

Yes, being able to observe a brain to the same depth as the simulation is assumed.

Not just non-destructive real time observation, but also understanding patterns in observed state changes sufficiently well to be able to create lossless abstract representations of them in another medium. All seven octillion of them, operating in continuous time.

There are information theoretic reasons to suspect the human mind can't parse itself in sufficient detail to write that program, but even without these I feel the rhetorical framing "why should it be possible?" makes more sense than "why shouldn't it". At least with "why shouldn't my dog be capable of understanding this HN exchange?" we've got evidence that mammals can learn that in our favour, and the step change between understanding "walkies" and thought experiments about hypothetical brain emulations is probably smaller than that between spotting which brain excites the EEG scanner more and atomic-level brain emulation.

This is not the point though.

I'm not discussing whether we, humans can create such a simulation. It is a thought experiment of what we think such a simulated mind would be like and whether it would have intelligence/self-awareness/consciousness/experience qualia/etc.

The topic is whether there is anything fundamentally important about the human wetware which couldn't be substituted by a computer.

The sticking point for me here is that you are pointing out that it could be very difficult but have not come out right and said you suspect it to be impossible - that would be a fine response to the top level question of what is stopping you from simulating a human in software; "the fundamental error in simulating those atoms is great enough that we would no longer see expected macro behaviors like predictable chemistry". I wouldn't expect you to say it with the same level of conviction, but assailing a question like that with practical concerns instead of the principle of the matter feels like it misses the point a little.

Asking "why should it be possible" is very reasonable - but you gotta ask it. My take on it is simple; the interactions between things can be reliably predicted - if not exactly then with a model whose outputs are indistinguishable from reality.

But the rough implementation details are pretty fundamental to what is being asked!

I don't think there are any logical reasons why a different set of atoms cannot be an equivalent mind with identical conscious experience (my mind constituted different atoms yesterday, yet perceives continuity!) though I'm unconvinced of the existence of an entity capable of engineering it (and agnostic on what permutations of atoms are and aren't capable of representing brain state). But that's a very different question to "is it possible for humans to create equivalent consciousness by writing high level code which symbolically represents the molecular behaviour we think is pertinent to thought as electrical signals as a binary code state machine" which is often implied by such a question, the answer to which I think is a fairly confident no. It's commonly suggested that scepticism about the ability of the human mind to be replicated in silicon suggests "something special" about it; when rephrased as a question about actually building it it becomes more of a question about whether there's "something special" about our theories of computation and i/o pattern matching that transcends fundamentally different chemistry, as well as the mother of all engineering problems.

That doesnt follow from anything that has been said. Do you have some reason to think that a simulation at that resolution and scale is fundamentally impossible? For purposes of the thought experiment you could go bigger or smaller - the practical considerations aren't important for that.

I think you would end up simulating a childhood, society, cultural history, etc, - basically a human world - if you want to develop a human intelligence. I feel like this is an even harder problem than simulating a brain by itself.

>To what extent is our intelligence a consequence of our bodies, evolution?

Entirely? We're meat machines.

>Would a computer have intuition? Fight or flight responses?

If your grandmother had wheels would she be a bicycle?

Yeah, if it walks like a duck ...

Cars have wheels. They aren't bicycles. If all grandma does is function like a bicycle, then she is essentially a bicycle made of meat.

When we have the technology to extend or replace our own neurons with artificial ones, then we can start to answer these questions.

Rebuild your ship and find out if you're still Theseus.

You mean like this: https://neuralink.com/approach/ ?

It doesn't replace any neurons, merely interface with it.

https://archive.is/v5LjK

Intelligence is the capacity to direct our behavior in the face of external circumstances and events. The more it integrates rational thinking, emotions, intuitions and acquired knowledge, the more we are intelligent. This is all contained into the workings of the brain. Even the "extended" cognition is only using previously created artifacts to enhance our memory, the same as we use books. As for identity, it is perceived from inside as we build a model of ourselves. We test our thoughts with feedback we gather. Our identity evolves with experience and incidents. The systemic identity of an individual is its uniqueness in the collective, and is observed from the behavior of the individual in selective circumstances. Nobody ever discovers the entirety of the identity, which only exists at the boundary between inside and outside, that is not observable from either points of view.

I don't think intelligence can be separated from a body. Intelligent beings exist to fight entropy. They have a body that feels things like pain, pleasure, and hunger which lead to actions that minimize or maximize those feelings.

ChatGPT and the like, might have senses like sight or hearing, but those systems are not acting on any kind of self preservation basis. They arent fighting entropy to exist.

Once the machines feel hunger, fear, pain and so on and can use their learned knowledge to act in their own self interest then I think we can call them intelligent. I feel like a "body" of somekind is a requirement for pain and pleasure, but who knows? Im certainly not willing to make any predictions.

The question doesn't make much sense. A body is a ton of inputs and a ton of outputs, and some processing in the middle. You can separate the processing, but that would be as pointless as raising a brain in a tank from infancy.

The process of training modern AI is basically simulating a body for an algorithm, except the feedback models get isn't as consistent as physics is, so legacy intelligence has an advantage.

Models learning how to draw from physical metaphors would strengthen their language ability through analogy; imo understanding a thing moving through space and time is the primary thing public examples haven't completely figured out.

Intelligent intelligence is encoded in DNA ("so many knobs"): becoming, experiencing, learning, seeing, for some walking just from birth. Being sentient and conscious for sure help a lot, it took quite time too - but why the medium (the best I can imagine BTW) shall be the limit ? Good luck!

Awareness of self and of your own mind doesn't require a body, nor feelings for that matter. Human intelligence is actually limited by evolution and our meat bodies. It's quite evident that it's possible to engineer a mind more intelligent than evolution can ever hope to produce. Evolutionary intelligence is limited by biology.

I prefer subtractive models when discussing this topic as it leads to faster conclusions.

What's the minimum number of sensory disablements a person could undergo before they were unable to be intelligent?

If the question is absurd, it's precisely because the framing it operates - the embodiment dilemma - is ill-posed.

I feel like intelligent beings can be raised regardless of language, sight, hearing, voice or a combination thereof. The guardrails to not die in the first few development years(epochs) is what carries us humans into adulthood with job specific specialist education

A Neal Stephenson book named “Fall; or Dodge in Hell” explores this idea a bit.

It’s a worthwhile read.

Well, if politics is any measure, then clearly the answer is yes.

AI has a body, it's just not the type of body we think of when we think of "body."

AI can have many sensors that allow it to "see/hear/feel" outside of itself.

AI is generally limited to its own system and requires a great deal of processing power, so it cannot easily "jump to a new body" as many people worry that a rogue AI will "find its way on to the internet." Fairly sure it would be useless for it to try to shack up on small low-power slow-CPU devices.

AI has a body, and physical limitations, because computers are not fucking ethereal. They have physical locations, even when they are "virtual machines," they still live on physical hardware.

When you take a drill to a hard drive, you are physically destroying data. Data is real.

Is the AI in the room with us now?

The AI is always looking down on us from the cloud.

Not down on those in the air gapped rooms...

Huh, you've got an interesting observation there - I hadn't considered notional agents to be embodied because the specific location the code runs just isnt that important, but if we were to posit self aware artificial agents they might have a unifying perspective rooted in the practicalities of computer and network hardware. "If you flip my bits such that error correction cannot detect, does my data not corrupt?" sort of thing.

But the AI can transmit from good enough computer A to good enough computer B.

We will make it that way eventually. But we will make us that way eventually too, via full body transplant or something like Altered Carbon.

So no big difference indeed. Its just technology level for different entities. People are scared of AI, I would be basically equally scared if some psycopath can achive such transendence.

Ask Stephen Hawking

Reply on news.ycombinator.com