Cognition Without Computation | Hacker News Reader

Cognition Without Computation

spectrum.ieee.org

I was skimming past the examples of other theories which turned out to be wrong, in order to get to the meat of the argument, but before I could detect any such meat the article had already ended.

Maybe instead these conscious experiences come from some kind of self-organization. The computation we associate with these sensations could be simply an invention of our own to explain the mechanism of sentience, not the primary cause of it.

What is meant with "some kind of self-organization", and, crucially, why couldn't it be considered a form of computation?

I was about to post the same thing. Self-organisation maps seem a classical computational model to me. If the author’s point was that computational models should be biologically plausible, there are many other examples as well. I’ve never really understood what neuroscientists are talking about…

Yeah, a proper refutation of the idea that cognition is computation is going to have a VERY difficult time addressing the fact that computation and information theory are VERY fundamental physics, right down there with entropy, thermodynamics, and causality.

I just don't see how you can possibly get around that without figuratively introducing fairies. More susinctly: if it's not that, I think it's going to involve a proposal that involves something outside of observed physics which puts it on very dubious ground.

If you don't accept that, you have some really fundamental truths you're going to have to explain around.

1. By all measures, cognition is intimately associated with the physical brain (and the fundamental forces, chemistry, and emergent phenomenon of a functioning brain). This is basically irrefutable by every measurement and understanding of physics that we have. We can even use drugs, strong electric or magnetic fields, or physical stimulation to interrupt and alter that cognition.

2. Computation itself is such a fundamental notion involving information and operations on that information that most emergent phenomenon you could propose will almost necessarily meet the definition of computation.

Now... exactly what TYPE of computation is an interesting question: is it equivalent to a turing machine? To this, I don't know, but I suspect so. There are many other viable computational models though that likely fit the bill if this fails.

Something I feel more and more is that a lot of our brain capability is related to whatever form of memory exists in it. Things are hard if you can't remember them fully, you're smart if you can absorb a lot.

Space/Density/Regularity (self similarity helping here) are somehow universal I believe.

Equating brain capacity to memory cannot be further from the truth. The joy of being intelligent is that one can pull out things from thin air without remembering anything.

Equating is stronger than what I said. I simply claimed that space/memory is a huge part of what brain function is.

And those who can come up with totally new, orthogonal ideas may just have a better compressor :p

> Yeah, a proper refutation of the idea that cognition is computation is going to have a VERY difficult time addressing the fact that computation and information theory are VERY fundamental physics, right down there with entropy, thermodynamics, and causality.

That’s not at all a given. Our models tell us nothing about the true nature of what they describe, they merely let us make predictions whose accuracy is limited by the congruence of the model with observation up to available computing capacity. Quantum Electrodynamics, which is arguably science’s greatest triumph for agreement between model and observation, doesn’t mean that photons are actually members of the computable subset of C.

It’s not surprising to see man with a hammer syndrome about computation among programmers. It’s an easy trap to fall into and I try hard not to.

Given that Steven Wolfram wants to model the entire universe as a computation, the idea that one would need to refute the idea that one could form a computational model of cognition seems a little silly. You can build a computational model of more or less anything at this point, and "this is not computational" is a red herring.

The question, as other commenters have pointed out, is not whether cognition IS computation, but whether or not a computational model of cognition is the best way to understand it.

The author is Rodney Brooks [1] so I assume he's aware that it can be considered a form of computation, and probably would agree that it can be simulated using a computer (just as Newtonian mechanics can). But my guess is that he's asking whether that level of abstraction will turn out to be the most scientifically useful way of understanding how cognition actually works. After all, cognition could also be described as interactions of atoms - but trying to understand it in those terms is unlikely to yield much insight.

But it is certainly true that the article is pretty uninformative and superficial!

[1] https://en.wikipedia.org/wiki/Rodney_Brooks

At the 20th century, science evolved in a way that now every scientific theory must be described computationally. (It's actually the most flexible way we ever had, so there's no loss there.)

So, yes, the entire exercise looks meaningless to me too. If we get a theory it will be computational. And that tells absolutely nothing.

"the article had already ended"

I had this very same experience.

I actually laughed out loud at the end of the article.

"This long held theory that is certainly vague and fuzzy but is as concrete as we currently know how to make it could be wrong! What if, instead, bargle fluzzle emergent momble self-organization bazamble magrile?"

What if, indeed.

You can consider the gradient-following of water when acted upon by gravity a "gradient calculation" but I think it would be deeply misleading.

This article is crap. It's impossible to even say what the author means by computation. I kept reading expecting to find some kind of there there (after all its in IEEE), but nope. Just abruptly ends with him calling himself a kook. Wow, did Rodney Brooks really write this?

This is a much deeper, better take on a similar rejection of simplistic "computational" models of cognition. It might have been on HN a few weeks ago:

https://aeon.co/essays/the-study-of-the-mind-needs-a-coperni...

In my opinion computationalism is the only non-mysterious explanation of how the mind/brain works in principle. By "computationalism" I mean either any process that can be expressed by reducing formulas in the untyped lambda calculus, or, alternatively, all processes that can be expressed in some parallel approach to computation like the pi calculus. (I've never seen a conclusive argument why the "software" running on the mind/brain must be parallel, though. But there could be such an argument.)

What alternatives are there? Other versions of functionalism seem to boil down to computationalism in the end - although the final verdict on this matter is open. Penrose's quantum approach is falsifiable, which is great, but there is not much evidence for it. Dualism is not compatible with modern physics, suffers from a variety of homunculus problems, and even if it was true, computationalism of the mind would still be the best explanation. (Computationalism does not imply physicalism, although most computationalists are physicalists.) Hypercomputation presumes ordinary computation, most hypercomputers are physically impossible, and there is no evidence that the mind/brain is a hypercomputer anyway.

In any case the article does not define computationalism, so it's pointless.

One alternative theory is that the mind is immaterial.

>>(I've never seen a conclusive argument why the "software" running on the mind/brain must be parallel, though. But there could be such an argument.)

I don't think there is, but the way the brain is connected implies a great deal of structural parallelism.

If things on ieee are going to get philosophical then its wise to start with philosophy! David Chalmers is great at showing the complexity of these kinds of qs: https://philpapers.org/rec/CHAOIA https://philpapers.org/rec/CHATVO-8

I assume he means cognition without classical computation, but that's non-controversial already-- we mostly agree (minus some holdouts) that processing units are also memory units. Whatever else he could mean is not clear to me.

Also, there is very good indirect evidence that cognition is computational: https://direct.mit.edu/jocn/article-abstract/3/2/190/3022/De...

You can multiply these sort of examples dozens of times.

The the brain is the most complex object we know of. Therefore, throughout history, it’s always been compared to the most complex technology of each historical era. “ Descartes thought that the brain was a kind of hydraulic pump, propelling the spirits of the nervous system through the body. Freud compared the brain to a steam engine.” And today we have computational neuroscience. That doesn’t mean that these models aren’t useful. Many discoveries came about from thinking about the brain as an engine, and now as a computer. But these models are just models. They are refined and become closer to the reality that we observe in the brain, but we still have massive gaps in our understanding of the brain. Like the misconception with amyloid plaques in Alzheimer’s that was only realized in the past few years. And now with glial cells, which likely make up 80% of the brain (but we’re not even sure). https://blogs.scientificamerican.com/brainwaves/know-your-ne...

Glial cells were largely ignored, and thought to only protect neurons with myelin, but we’re now realizing how important they actually are for cognition and memory. And computational models and neural networks don’t even take glial cells into account when forming models.

The mind may not work anything like the electronic computers we're familiar with (in particular, the hardware vs software distinction may not be useful for understanding the mind, which is implemented in "wetware" that's both and neither), but that doesn't mean that what it does isn't some form of computation. Indeed, information processing === computation, and there can be little doubt that what the mind does is process information.

What if the mind instead generates information from the noisy chaos of sensations? This goes back to Kant's idea that the mind creates the phenomenal world from the manifold of sensations via categories of thought like space, time and causality.

But it all depends on how seriously we take the idea that the world is actually information (or systems computing information) which is transmitted to the brain via the senses.

The article would be better served by exploring the “explanatory gap” (https://en.wikipedia.org/wiki/Explanatory_gap?wprov=sfti1) — how is it that neurons firing in a parallel asynchronous manner leads to sequential synchronous experience and thought? How is it that matter leads to what many people throughout history have understood as “spirit”.

The question can be restated more clearly: why aren’t we all philosophical zombies? Or more generally: why is there something rather than nothing?

It may never be possible to answer this question, but does an answer matter? We know that it is possible to arrange a physical system (the brain) in such a way that a mind emerges. At some point we will be able to develop such systems in silicon. We don’t need to know why there is something rather than nothing to do so. We simply have to follow the patterns that we know work

I like Rodney Brooks, but this just seems like a nothing idea/article.

Indeed. "Many widely-accepted theories ultimately get rejected. Perhaps this one will be, too." What use is an article like this to anyone?

My current best model of cognition involves no computation. Though arises through the amplification and dampening of information storage patterns as they are translated between different pattern matching regions.

Maybe the best way to describe it is like that kids game telephone. One region takes the sensory data and first listens, then speaks some partial thought. The thought comes back out changed slightly. A fixed point occurs when the idea comes back unchanged. A resonance occurs when a fixed point is shared between different regions.

This has some relation to wave interference, and holographic information storage.

I believe this model has parts that could be useful when modeling information transfer at all size scales, from brain regions to whole brains, and even to communication between individuals and groups.

The key is to find a resonant idea that can be repeated back almost unchanged, and which can therefore be repeated to others.

Memes in the original meaning of the word have this characteristic.

Video: Wolfram on "What is space (and time)". Does compute underly everything? @37:10 https://youtu.be/qoDZKlcdPNM

One day we are all going to come full circle and realize that the ancients were right: the mind is immaterial.

Then why do injuries to the brain directly impact the parts of selfhood we hold most sacred, like feelings memories and attitudes?

You can still say that many of those things (memories, feelings) are material impressions or tendencies in the brain. But cognition itself must have an immaterial component, because it deals with immaterial realities (justice, truth, abstracted quantities and shapes, etc.) That's the Aristotelian and Thomistic theory.

That doesn't follow from your premise. You could justify any number of absurd theories with similar reasoning.

What do you mean? We couldn't know anything immaterial if all we are is material. That's all I'm saying.

Then why posit the existence of the immaterial?

What I'm getting at is I could use your reasoning to posit insane things that you would heartily disagree with, using your reasoning without twisting or bending it. Because your reasoning is baseless, or rather it rests on its own conclusion.

Even to just stick to the realm you're suggesting - why would anything about us need to be immaterial to have concepts like justice? I don't see the necessity of an unprovable, nebulous, unnecessary immaterial reality to explain anything.

I grant that I'm assuming you also think justice exists. But if you do, I think that you'd have to grant that justice itself is not material. The fact that you can know it, even though it's not material, suggests that your mind is also immaterial.

Perhaps a stretch, but maybe like a color filter (injury) affects the perception of a rainbow (immaterial).

What about a rainbow is immaterial?

Or we will realise the conclusions of the 17th century philosophers, that there is no coherent notion of physical/material. Anything we can reason about is a construction of the mind on the occasion of sense.

Yes it starts with the senses, but at a certain point the mind abstracts immaterial notions from physical reality, such as justice, truth, point, line, and so on. So the mind must be immaterial.

If there is no coherent notion of physical or material, there is no notion of immaterial either. Chomsky talks about this at length in "The Machine, The Ghost and the Limits of Understanding" (previously discussed on HN here [1]).

[1] https://news.ycombinator.com/item?id=26448901

If there were no coherent notion of physical & material, then we wouldn't have words for them.

Reply on news.ycombinator.com