Neurologist outlines why machines can’t think

[devolved]

Even the internal model isn't really a detailed model of reality, except at the focus of visual attention; the rest is a bunch of expectations of detail, based on the understanding that if you looked directly at it, it would appear to be detailed.

I'm not sure if that's entirely true. I think that there is some of that, but how much is really exaggerated.

When you read about all of the visual anomalies that are cited in support of this idea, you have to also consider something like cinema and photography, which gives us certain objective context of reality that we recognize to be consistent, and can actually analyze "blindly" pixel by pixel to see that it's all there, and amazingly-enough it matches our visual perception in context of that perception. There are no interpretive mechanisms that exist between camera lenses and photo-chemical reaction of the film.

Of course, we could debate about color normalization in our brain, and tint you get on film under different lighting conditions, but that's quite different than claiming that details that we see are exaggerated. As humans, just like cameras, we have "resolution thresholds" when it comes to our vision, but it doesn't mean that we don't really see what we see.

 

[expos4ever]

The OP appears to simply assume that a machine cannot have intentionality. I see no reason to believe this.

 

[expos4ever]

Its possible advanced quantuum computers might indeed achieve self-awareness and even experience something like qualia. .

The human brain and body may be more like a radio receiver for consciousness. So even rocks and trees may have a primitive kind of consciousness, but as life advances it becomes more developed until self-awareness is reached.

Are you a fellow David Chalmers fan?

 

[devolved]

This the sort of image our eyes actually see:

That's actually not true, or perhaps only true if you approach it with a "reductionist idea" of how vision works. This is how you would see what you see if you stick a very small sensor into your eye and snap a photo, but that's not how our eyes work. Thus, this is not what eye sees. Our vision is a composite of a "sensorial array" spread over a wider surface area of the retinal wall:

https://image.slidesharecdn.com/bioniceye-130508040727-phpapp02/95/bionic-eye-3-638.jpg

 

[devolved]

The assertion that computation is thought, hence thought is computation, is called computer functionalism. It is the theory that the human mind is to the brain as software is to hardware. The mind is what the brain does; the brain “runs” the mind, as a computer runs a program. However, careful examination of natural intelligence (the human mind) and artificial intelligence (computation) shows that this is a profound misunderstanding.

Nah. It's only a misunderstanding if you have colloquial perception of "dualism" between software and hardware.

Computing is a name we give to a process of reality. A program, is simply a state setting for parameterized process. Physically, a computer program is not very much different than, let's say, pouring water through a maze, and closing some doors while opening other. All we are doing is directing water to a certain end. Human mind isn't much different in that context.

What is the hallmark of human thought, and what distinguishes thoughts from material things? Franz Brentano (1838–1917), a German philosopher in the 19th century, answered this question decisively. All thoughts are about something, whereas no material object is inherently “about” anything. This property of aboutness is called intentionality, and intentionality is the hallmark of the mind. Every thought that I have shares the property of aboutness—I think about my vacation, or about politics, or about my family. But no material object is, in itself, “about” anything. A mountain or a rock or a pen lacks aboutness—they are just objects. Only a mind has intentionality, and intentionality is the hallmark of the mind.....

Both Brentano and you are abstracting a concept of thought as a noun from the continuum of a process thought as a function. Likewise, the concept of "aboutness" is a moot here. A picture is likewise "about" a setting that you take that picture of. The context is a projection of what thought "is", and not how our thinking works as a process.

That's the problem with abstract concepts, you can then run with them and think that these are real things that exist apart from contextual arrangement of matter, or processes of reality. And that's where much of philosophers and philosophy runs into a problem.

For example you could turn "running" into a noun, and say "my run was great". What you likely mean is not that there's some "thing" like "run" that exists in reality that we can begin to reify and talk about separately from running as a complex process.

It's the same thing here. Thinking is a process. A process is a sequence of events. We call a specific "snapshot" pattern of certain sequence of thinking ... a "thought". But in context of software/hardware analogy, it's actually the same and inseparable concept.

Thus, what we conceptually label as "computation" is not much different than a "computer's thinking". It's just a very "low level" of thinking that doesn't take into account a much wider array of context at the same speeds that we do.

....But to believe that machines can think or that human thought is a kind of computation is a profound error. Belief in this fundamental error about AI will lead us away from, not toward, the truth about AI. Machines, for example, will never become malevolent and harm mankind. Men will act with malevolence, using machines, or men will use machines in ways that (unintentionally) harm others. Men can use cars malevolently and carelessly and can thus harm others. But the malevolence and careless is in the man, not in the car.

While I agree that all we are doing with computing is externalizing and optimizing of certain emulation of brain processes, I'm not sure what it has to do with supposed difference between "AI" and human thought. I agree that AI is "reduced and emulated version" of human thought as a process. That's why I think that the distinction between "artificiality" of it is rather "meh".

A processor computational process is similar with our brain/thought process in such that both direct certain "inputs" through a "maze" of open/closed "pathways" that channel process of reality to a certain end and not the other. It's an extremely uncomfortable realization for many, but that's what thought and thinking appears to be from factual understanding of neurophysiology.

 

[devolved]

From the bit of research I've done into the area of AI, it seems that older computational approaches are hopeless, but that deep learning and artificial neural networks modeled on the brain might in theory lead to strong AI.

Well, what do we mean by "strong AI". There's a distinction that should be made between a tool and autonomous process. Much of the "deep learning" AI today is merely "crowdsourced" human thought.

Let's take google translate. The previous approaches to translation were "word-for-word" attempts with some lexical grammar rules,. That proven to be futile, because language rules are not absolute, and there's a vast semantic network of concepts that contextualize words and phrases when it comes to likely meaning of any given phrase.

"I was swimming and I was washed onto a bank" would mean something different than "I went to a bank to deposit money". Humans have no problem differentiating between the two contexts. Machines would struggle, because language is not about rules, but about rules with context.

So, what google did is they crowd-source phrase translation to a vast number of people who translated phrases and chunks, which all exist in an enormous database of these phrases that these people contributed to in order to fine-tune the likely translation of any given phrase or sentence. So, it's not really "deep learning" as much as it is "distributed programming" to contextualize enormous datasets with various possibilities of contextual meaning.

And yet, even after a enormous amount of contribution... if I run some following English through google translate:

He brings with him, abroad, a new sense of strategic vision. Obama's critics should give him a break for having to govern in a time of anarchy and limited choices. But he compounded regional crises by laying down red lines, for instance on the use of chemical weapons,:

I get this in Russian:

Он приносит с собой за границу новое чувство стратегического видения. Критика Обамы должна дать ему перерыв в том, что он должен управлять во времена анархии и ограниченного выбора. Но он усугублял региональные кризисы, укладывая красные линии, например, на использование химического оружия.

Which actually means:

He brings with him across the border a new feeling of strategic sight. Criticism of Obama must give him a pause (really weird word choice) in that he must govern during the times of anarchy and limited choice. But he made the regional crises worse, laying down red lines (makes no idiomatic sense in Russian), for example, for usage of chemical weapons.

So, if AI would be given task of managing US - Russia translation for negotiation, probably wouldn't end too well.




I actually conducted another google translate semantic problem experiment:

На морской косе сидела девушка с косой и говорила что косой парень косил траву с косой.

It's a contextualized test of word "коса", which means different things based on what context we plug it in.

On the sea spit (dune) sat a girl with braided hair, and was talking about a cross-eyed guy who was cutting the grass with a scythe.

We instead get :

On the sea spit sat a girl with a scythe and said that the oblique guy was mowing grass with a scythe.


So, the question is... would you let something like that drive your car with your children in it ?

 

[FrumiousBandersnatch]

In the brain, what is the equivalent of a core and a thread?

Nuerons and axons.

No. Neurons don't execute instruction streams, they're more like complex switches or simple signal processors; and axons are neural output wiring, not instruction streams for execution.

 

[Justatruthseeker]

Yes to the first, and no to the second. The water molecules in water waves follow roughly circular motions vertically, parallel to the wave direction.

It's completely different. For example, in the double-slit experiment, a single photon can be shown to 'interfere' with itself.

A single photon or a quantum wave packet?????? I would expect nothing else but the ability to interfere with itself being that packet consists of multiple waves.... not a single wave or a single particle....

Not so fast - carbon 60 molecules (named 'Buckyballs' after Buckmister Fuller) produce interference patterns in the double-slit experiment. I don't recall anyone calling them waves...

So does water.

From the examples you give, it's clear that you need to learn more about quantum mechanics - and it's worth it, it really is weird and fascinating.

Agreed, quantum electrodynamics which served as the basis for all quantum theories is indeed fascinating. That in their attempt to mechanize the EM force they were forced to conclude reality isn't real until we observe it, instead of correcting the flaws leading to that belief.

It's called an electromagnetic wave, and photons can behave as discrete quantised particles, so it's more complicated than that. In physics there are all kinds of waves - are they all 'true' waves?

Yes it is quantum electrodynamics, not quantum mechanics. You mean that quantum wave packet can be interpreted to behave as discrete quantisized particles, like the electron behaved like a particle even if it was merely a wave????? So particles emit waves and waves emit particles?????

Quite; a pity the example wasn't quite right

Agreed, quite a pity a quantum wave packet is mistakenly called a particle.....

All local hidden variable theories have been ruled out by the violation of the Bell inequalities in Bell test experiments.

Except it's flawed in itself.

"a mathematical description of local realism that placed a statistical limit on the likelihood of that eventuality. All these inequalities, like the original devised by Bell, express the idea that assuming local realism places restrictions on the statistical results of experiments on sets of particles that have taken part in an interaction and then separated."

By limiting the experiment statistically, the results are weighted towards the assumption it was weighted to produce.....

However, there are non-local interpretations of QM with hidden variables (e.g. de Broglie Bohm pilot wave theory), and interpretations where the randomness is an artefact of our limited perspective after observation (Everettian 'Many Worlds'). 'Many Worlds' is the simplest, and increasingly popular among physicists who care about interpretations.

You mean when the statistics and experiments are not weighted to fit the assumption? Agreed.

I'm with you on the von Neumann-Wigner (conscious collapse) version of Copenhagen, but there are various other versions that have plausible explanations for wavefunction collapse. The problem they all have is that wavefunction collapse itself is ad-hoc, it has no mathematical derivation.

Max Tegmark (not a Copenhagenist) has said that even if it's wrong, and the wavefunction doesn't collapse (his view), it's still a useful approximation to do the calculations as if it does (because that's what we see).

Agreed, sometimes we have to simplify things to even do the calculations. It's like plasma scalability. It is almost infinite, but we artificially place a lower value of 100 because otherwise even the best computers would be forever crunching numbers. So even the complexity of 100 does not account for all the variables.

Or for example consider an atom electrically neutral as if the proton and electron are not simply balanced in forces so that neither one or the other is in excess, not that it is electrically inert..... But such is what neutral portray's to the mind. Not a balance of electrical forces, but no electrical forces. What our minds "see" and reality are often two different things....

 

[Justatruthseeker]

No. Neurons don't execute instruction streams, they're more like complex switches or simple signal processors; and axons are neural output wiring, not instruction streams for execution.

Because you wont accept that thought happens in the spark between the gaps, not in the flesh of the brain itself?

 

[Justatruthseeker]

It's even stranger than he describes, because when you open your eyes, not only do you not see 'now', but the crisp, detailed world you perceive is not what your eyes actually see - it's an internal model. The eyes don't have the detail and resolution to present a photorealistic real-time image of the world, so we build up an internal model of what we expect to see based on past experience, and use the information from the eyes to check for conflicts and update the model if necessary.

This the sort of image our eyes actually see:

Even the internal model isn't really a detailed model of reality, except at the focus of visual attention; the rest is a bunch of expectations of detail, based on the understanding that if you looked directly at it, it would appear to be detailed.

Oh agreed, we see what we expect to see. But wouldn't seeing form and texture make "photons" the silly putty of the universe????

 

[Justatruthseeker]

I'm not sure if that's entirely true. I think that there is some of that, but how much is really exaggerated.

When you read about all of the visual anomalies that are cited in support of this idea, you have to also consider something like cinema and photography, which gives us certain objective context of reality that we recognize to be consistent, and can actually analyze "blindly" pixel by pixel to see that it's all there, and amazingly-enough it matches our visual perception in context of that perception. There are no interpretive mechanisms that exist between camera lenses and photo-chemical reaction of the film.

Of course, we could debate about color normalization in our brain, and tint you get on film under different lighting conditions, but that's quite different than claiming that details that we see are exaggerated. As humans, just like cameras, we have "resolution thresholds" when it comes to our vision, but it doesn't mean that we don't really see what we see.

So photons are the "silly putty" of the universe? After all, they must convey the fine details of whatever they reflect off of back to the film plate. Not the details of what emitted the light, but the details of the objects encountered after emission... Which leaves a very deep question to quite a few beliefs and theories.

 

[FrumiousBandersnatch]

But the definition is too simple if it involves conflating multiple different meanings of the word "determine" into one concept. Philippa Foot discusses this particular problem in her article Free Will As Involving Determinism, which you can find part of in the Google preview here.

Yes, this is why I define it (in this context) as I do (events are causal).

The little I've seen of Foot's article takes after Hobart; I broadly agree with her position, except in that it seems to me that wants and desires are implicitly deterministic (causal). Whether Russell's universal determinism is true is another matter (I suspect it is if we accept the level 1 or 3 multiverse).

Sorry, I actually thought this was your position--that our conscious choices are in fact illusory and that all decisions can be reduced to the level of brain processes. That is what I meant by conscious choice being outside of the causal chain--it isn't causal because consciousness is itself epiphenomenal.

No, it's not my position; our conscious choices are not illusory (though they may not occur quite the way they feel), and yet all decisions are the result of causal brain processes. I don't see how non-random choices or decisions could be other than causal, being evaluation and selection with a view to some goal; i.e. information processing.

Given the discoveries of experimental psychology and neuroscience, it seems fairly clear that the majority of decisions we make are the result of unconscious processes (Kahneman's 'System I'), for which, if we become consciously aware of them, we assert retrospective agency (even confabulating plausible reasons where necessary). Even for the few conscious, deliberative (System II) decisions, where we work sequentially through the evaluation of the options, most of the 'grunt' work is delegated to System I.

So, although it's all information processing, choices and decisions are made. It is often said that, under determinism, the outcome of a choice is inevitable, so there isn't really any choice and free will is illusory, but this is mixing levels of description; at a human level of description, despite the surprising predictability of people's choices, we don't (and, I think, can't) know all the parametric and process contributions to our choices - particularly as the majority are below conscious awareness - and subjective - we are the system that is doing the processing, and although, from experience, we may suspect what the outcome is likely to be, we still have to go through the process and produce an outcome.

To me, saying that's 'illusory' free will is like saying that the temperature and pressure of a gas are illusory because it's really just a vast number of molecules moving around with random velocities. Bear in mind I'm using a pragmatic definition of free will - as the experience of making choices without coercion or constraint.

You were arguing for a hard deterministic position before and saying that free will was subjective and illusory, and now you seem to have moved to compatibilism instead. So we are no longer as far apart as we initially seemed.

You misconstrued my argument - I hope the above clarifies things a little.

As far as the two-stage model of free will goes, I think this article is a pretty useful way of conceiving it (amusingly, since it's actually arguing against free will). The brain's ability to generate multiple action plans might seem like the sort of determinism that is outside our control, but indeterminism is also to be found here, as none of these plans are actualized until we make the decision to act upon them. Again, fairly Aristotelian--just turn Many Worlds on its head.

Yes, it looks plausible, in general - although, as above, I think calling free will illusory is as useful as saying the solidity of bricks is illusory (because they're 99.999% empty space); the language and concepts of such different levels of description are incompatible. Free will is subjective, experiential, just as consciousness is, and I think the 'illusion' issue with them both is interpretational; neither are quite what they seem to be, experientially, just as phantom limb pain is real and experiential, but not what it seems to be.

 

[FrumiousBandersnatch]

When you read about all of the visual anomalies that are cited in support of this idea, you have to also consider something like cinema and photography, which gives us certain objective context of reality that we recognize to be consistent, and can actually analyze "blindly" pixel by pixel to see that it's all there, and amazingly-enough it matches our visual perception in context of that perception. There are no interpretive mechanisms that exist between camera lenses and photo-chemical reaction of the film.

Not sure what you're saying here - when we view a movie or photograph, the interpretive processing is much the same.

Of course, we could debate about color normalization in our brain, and tint you get on film under different lighting conditions, but that's quite different than claiming that details that we see are exaggerated. As humans, just like cameras, we have "resolution thresholds" when it comes to our vision, but it doesn't mean that we don't really see what we see.

I'm not saying we exaggerate the detail (the real world is as detailed as you like), but that, outside of the foveal focus, which is relatively tiny, the eye's imaging is inadequate to produce the detailed real-time view of the world we see - what we see is an internal representation of what we expect to see.

 

[FrumiousBandersnatch]

That's actually not true, or perhaps only true if you approach it with a "reductionist idea" of how vision works. This is how you would see what you see if you stick a very small sensor into your eye and snap a photo, but that's not how our eyes work. Thus, this is not what eye sees. Our vision is a composite of a "sensorial array" spread over a wider surface area of the retinal wall:

https://image.slidesharecdn.com/bioniceye-130508040727-phpapp02/95/bionic-eye-3-638.jpg

I know how eyes work; here are the stats for the retina. A considerable amount of processing is done at the retina, and those photoreceptors aren't, ultimately, individual pixels, as some averaging is done between adjacent rods & cones.

 

[FrumiousBandersnatch]

A single photon or a quantum wave packet?????? I would expect nothing else but the ability to interfere with itself being that packet consists of multiple waves.... not a single wave or a single particle....

To a degree it's a matter of semantics, a photon is described as a 'particle' of light, and its evolution over time is described by a wavefunction which is usually a wave packet, a superposition of quantum probability amplitudes.

So does water.

Yes, but water waves are classical. Not sure what, other than distraction, your point is here. Ooh, squirrel!

... in their attempt to mechanize the EM force they were forced to conclude reality isn't real until we observe it, instead of correcting the flaws leading to that belief.

Did you really mean 'mechanize'? doesn't make sense to me; 'quantize' maybe?

As for reality not being real until we observe it, you can model it in various ways; as alternate histories being equally real until an observation is made, or superpositions being equally real until decoherence, etc. There are three shells of reality - external reality, consensus reality, and internal reality. When we discover something new about external reality, consensus and internal realities often have to change, and it's disconcerting, not least because, until we can conceptualise it coherently, we can only describe it in metaphor. This not the kind of reality we're familiar with, so we have no accurate metaphors, but if that's how nature behaves, that's reality.

The measurement 'problem' also remains - what is meant by a measurement or observation depends on your preferred interpretation.

You mean that quantum wave packet can be interpreted to behave as discrete quantisized particles, like the electron behaved like a particle even if it was merely a wave????? So particles emit waves and waves emit particles?????

Yes and no. They are what they are - sometimes a wave description is appropriate, sometimes a particle description.

Except it's flawed in itself.

"a mathematical description of local realism that placed a statistical limit on the likelihood of that eventuality. All these inequalities, like the original devised by Bell, express the idea that assuming local realism places restrictions on the statistical results of experiments on sets of particles that have taken part in an interaction and then separated."

By limiting the experiment statistically, the results are weighted towards the assumption it was weighted to produce.....

GMO! You're right!! The last 50 years of research and experimentation by the brightest in the field have been wasted! - I'll alert the Royal Physical Society and the Nobel Committee! Not

I'll leave you to it...

 

[FrumiousBandersnatch]

Oh agreed, we see what we expect to see. But wouldn't seeing form and texture make "photons" the silly putty of the universe????

Riiiight...

 

[FrumiousBandersnatch]

Because you wont accept that thought happens in the spark between the gaps, not in the flesh of the brain itself?

No - because that's what the words mean

 

[tas8831]

Nuerons and axons.

Ah, so you have the same level of knowledge of neurobiology that you do in genetics.

An axon is part of a neuron. Ask OldWiseGuy, the other resident creationist expert on things that he is not.

 

[Silmarien]

Yes, this is why I define it (in this context) as I do (events are causal).

The little I've seen of Foot's article takes after Hobart; I broadly agree with her position, except in that it seems to me that wants and desires are implicitly deterministic (causal). Whether Russell's universal determinism is true is another matter (I suspect it is if we accept the level 1 or 3 multiverse).

Yes, given that multiverse models tend to be motivated by a desire to save determinism, I would agree that they entail determinism. It's a piece of speculation I'm not particularly interested in, though, since I don't share any of the presuppositions underlying it.

On the topic of Foot and causality, if she did any work in metaphysics, I'm unaware of it (though I'm looking through the book I linked you now, so will see what she has to say about some of the relevant ideas soon enough), but she was an Aristotelian and might have approached causality differently than you do. Beliefs and desires are teleological and would not be "causing" behavior in the same sense that brain chemistry does. Efficient vs. final causality does lead to a different framework for those who favor ancient approaches to causality. (And you can even find theoretical physicists who do, so it is obviously still a valid metaphysical option.)

This is actually an interesting problem which leads quite nicely back to the original post, minus the AI red herring and implicit Creationism. What does it mean for beliefs and desires to be causal? These are intentional states--we could say that they represent attitudes towards information about the world. The traditional problem of Cartesian dualism is obviously explaining how the immaterial mind affects matter, but I would be hesitant to let materialism off the hook here, since it is unclear to me how intentional states can compel physical behavior. A muscle spasm produces motion, but what is the causal link between a desire and a motion? The options seem to be:

1) Reductive materialism. Intentional states such as beliefs and desires can be reduced to material interactions and information processing (granting for the moment that computation can be made sense of on a genuinely materialistic ontology, which I would deny). If intentional states are merely supervening upon specific physical states, then they themselves are superfluous to behavior, and it is unclear why they exist at all if we would function in exactly the same manner without them. It seems to me that you would need to retreat into panpsychism to make sense of this approach.

2) Eliminative materialism. Intentional states don't exist. We only believe that we have beliefs. (Just in case people doubted that materialists were capable of YEC level dogmatism.)

3) Non-reductive materialism. The mental is emergent from the physical but cannot be reduced to it. But if mental states cannot be reduced to physical ones, then we are back to wondering what a desire actually is and how it can play a causal role in physical processes. Non-reductive materialism is actually one of my favorite theories of mind, but I do not see how it avoids the problems of dualism simply by labeling itself differently.

Hopefully I have made myself sufficiently clear, since I think this is a better framework for a discussion of intentionality than the initial question of AI was.

Given the discoveries of experimental psychology and neuroscience, it seems fairly clear that the majority of decisions we make are the result of unconscious processes (Kahneman's 'System I'), for which, if we become consciously aware of them, we assert retrospective agency (even confabulating plausible reasons where necessary). Even for the few conscious, deliberative (System II) decisions, where we work sequentially through the evaluation of the options, most of the 'grunt' work is delegated to System I.

Yes, but the fact that the majority of decisions are unconscious is irrelevant if conscious, deliberative decisions ever take place. It does not matter how much of our behavior is unconscious or automatic--you still need to account for the existence of System II decisions or declare conscious, deliberative thought entirely and always illusory. Which seems like it would be an extreme and anti-empirical position.

To me, saying that's 'illusory' free will is like saying that the temperature and pressure of a gas are illusory because it's really just a vast number of molecules moving around with random velocities. Bear in mind I'm using a pragmatic definition of free will - as the experience of making choices without coercion or constraint.

I would have a problem with anyone except an idealist declaring temperature and pressure illusory. What is from one point of view a vast number of molecules moving around with random velocities is from a different perspective a gas with temperature and pressure. How are molecules real but pressure not real? And if we wish to reduce reality, can we not go even further? I am not at all uncomfortable denying the existence of molecules and particles and taking the relations without relata approach.

Yes, it looks plausible, in general - although, as above, I think calling free will illusory is as useful as saying the solidity of bricks is illusory (because they're 99.999% empty space); the language and concepts of such different levels of description are incompatible. Free will is subjective, experiential, just as consciousness is, and I think the 'illusion' issue with them both is interpretational; neither are quite what they seem to be, experientially, just as phantom limb pain is real and experiential, but not what it seems to be.

I don't think it's coherent to invoke the concept of illusion when discussing aspects of reality that are by their nature subjective and experiential, since the notion of an illusion presupposes the existence of a phenomenal state that matches up to reality. If none do, then the illusory is indistinguishable from the non-illusory. Either that, or we end up in maya territory dancing in the waters of Vedic nondualism. Which I find attractive, but you probably wouldn't.

 

[Silmarien]

Are you a fellow David Chalmers fan?

I am! (Though I disagree with him. For reasons that will become clear below.)

That's the problem with abstract concepts, you can then run with them and think that these are real things that exist apart from contextual arrangement of matter, or processes of reality. And that's where much of philosophers and philosophy runs into a problem.

Speaking as a genuine Platonist (well, Plotinian), I don't see any error here. We don't think that abstract concepts are real "things" out there somewhere in the ether as if they were physical--we just think that form is ontologically prior to matter. Instead of saying that structure can't exist independently of matter, we say that matter cannot exist without context and structure. Assuming matter exists at all. Perhaps all of reality can be reduced to mathematical structures dreaming them/ourselves into phenomenal existence.

In my experience, most attacks on Platonism either beg the question or miss the point of what's being said entirely.

Well, what do we mean by "strong AI". There's a distinction that should be made between a tool and autonomous process. Much of the "deep learning" AI today is merely "crowdsourced" human thought.

Strong AI specifically refers to hypothetical artificial intelligent that is fully conscious in the way that we would understand it. It's distinguishable from the weak AI that can carry out specific tasks but is not self-sufficient. Deep learning isn't there yet, but it at least seems like a more promising approach than previous ones.