Artificial Intelligence / Synthetic Consciousness

[sanscardinality]

We got into this a little in another thread, and while I don't have a ton to say today I thought it deserved a thread for future conversation as well. What inspired me to put this up was a quote I bumped into from Dijkstra:

"The question of whether a computer can think is no more interesting than the question of whether a submarine can swim."

It's clever, but frankly I find both ideas interesting - I want a swimming submarine operated by an AI!

SC

[sanscardinality]

Some cool papers by a guy doing modular neural net research:

http://julian.togelius.com/

SC

[caliban]

a quote I bumped into from Dijkstra:

"The question of whether a computer can think is no more interesting than the question of whether a submarine can swim."

SC

I can appreciate Dikjstra's quote. Much of science, and especially physics, moves forward by deciding that some questions are too difficult to answer, and then looking for a different question that can be answered. The original question is still important, but is left to philosophers. For example, Aristotle identified four causes, but science really focuses on one of them (efficient cause, I think, although it may be material cause).

There are many words out there which are minefields, because they are difficult to define in a rigorous fashion, and so we go around and around. Some of them that come to mind are :

mind
think and thought
consciousness
intelligence

to which one might add

freewill
determinism

and so on. These are valid philosophical questions to dwell on. But some people like me, with attention deficit disorder, get impatient, and so we want something more clearly defined we might be able to make progress on.

I think this is what Dijkstra refers to. Is what a submarine does "swimming?" Is what a computer does "thinking"? An question of great interest to philosophers. But there are other questions that Dijkstra, and others, might pay attention to. Because modern computers, while very fast, are clearly very shallow in how they approach the world. They don't "chunk" their perceptions in the way people do. Can we write algorithms that process the world approximately in the way people do? If so, it would make computers more useful, at least in some circumstances. Is that "real" thinking? I don't know, but it would be useful.

I guess one could restate the problem of AI, and of the Turing Test, like this: Is it possible for machines to respond to the world in a way analogous to how humans respond to the world? This is the essence of the Turing Test. You can argue all you like if it is genuine thought or not. But if we could achieve that, it would be useful.

[Windwalker]

Much of science, and especially physics, moves forward by deciding that some questions are too difficult to answer, and then looking for a different question that can be answered. The original question is still important, but is left to philosophers.

There are many words out there which are minefields, because they are difficult to define in a rigorous fashion, and so we go around and around. Some of them that come to mind are :

mind
think and thought
consciousness
intelligence
freewill
determinism

and so on.

Biologists are starting to tackle these questions at several scales, even though they are very aware that they're minefields -- not only because of definitions, but also because they impinge on religion and people's cherished ideas of what constitutes personality and identity. So they are no longer exclusively the province of philosophers.

[caliban]

Biologists are starting to tackle these questions at several scales, even though they are very aware that they're minefields -- not only because of definitions, but also because they impinge on religion and people's cherished ideas of what constitutes personality and identity. So they are no longer exclusively the province of philosophers.

Yeah, well, good luck on that.

I think the problem is less that they impinge on people's cherished ideas, but the problem of the Ultimate Question of Life, the Universe, and Everything:

Exactly what are these questions?

My general impression is that these terms, although in common use, are too vague to be useful.

But I am willing to be shown wrong--just show me the question!

And excuse me for being a grumpy, cynical bastard.

[Windwalker]

My general impression is that these terms, although in common use, are too vague to be useful.

Not in the biological domain, just as the word "theory" means one thing to scientists and another to non-scientists.

P. S. Why are you grumpy? (*laughs*)

[caliban]

P. S. Why are you grumpy? (*laughs*)

It's a principle with me.

So...really...can you give a definition of any of those terms, in such a way as to allow meaningful investigation? My ignorance is truly boundless.

[caliban]

Biologists are starting to tackle these questions at several scales, even though they are very aware that they're minefields -- not only because of definitions, but also because they impinge on religion and people's cherished ideas of what constitutes personality and identity. So they are no longer exclusively the province of philosophers.

Donna suggests you are referring to fMRI experiments, which certainly are fascinating and illuminating. And they are challenging to people's self perception. And I can also think about what we have learned about how much chemistry affects personality, mood, even processing.

In other words, we certainly have learned an enormous amount about brain function.

But I don't know... I still don't think this really addresses "mind" or "consciousness" or any of those other Big Words. If anything, I submit they show that those concepts are too fuzzy to be well defined. In fact, I suspect, as do others, that they are illusions. But we haven't a clue how that illusion is cast. Do we?

[Windwalker]

Donna suggests you are referring to fMRI experiments, which certainly are fascinating and illuminating. And they are challenging to people's self perception. And I can also think about what we have learned about how much chemistry affects personality, mood, even processing.

Yes, definitely these, plus information gleaned by studying people who had parts of their brains influenced by accident or disease (whether autism or stroke) and by parallel primate/human infant learning studies. At the molecular and cellular level, we know a good deal about brain chemistry and wiring.

If you think (as I do) that the brain creates the mind, all the information is there, from free will to liking bagels. It may be so complex that it takes the same amount of time to describe it as to explain it (which in fact is one definition of a complex system). Nevertheless, the fact that it's complex and still very partially known doesn't make it unknowable in principle.

[caliban]

Nevertheless, the fact that it's complex and still very partially known doesn't make it unknowable in principle.

Forgive me my dissent. I'll agree to your above statement, but I'll add another: just because we have learned a lot does not means that it is knowable in principle, either.

We're fencing over semantics to a large extent. We are learning huge amounts about brain function, and will continue to do so. I even believe, or want to believe, that someday we'll build machines that past the Turing Test or some suitable modification. But I also suspect that terms like consciousness, awareness, free will, and so on, may (note the emphasis) remain too slippery, too ill-defined. At best we'll have operational definitions, such as our current operational definitions of self-awareness, as in tests of dolphins and other intelligent creatures looking at themselves in mirrors.

Look at 'intelligence." We have a hundred years of so-called intelligence tests, and as far as I can tell, they mostly only test how well you do on an intelligence test. We can talk about the survival or operational advantages of intelligence, but really, we can't even define intelligence in any rigorous manner.

I insist on this not because I believe the human mind is a miracle or supernatural or anything, as both C.S. Lewis and Roger Penrose, for opposing reasons, conclude. But my training is in physics, and the history of physics has taught us that some questions simply cannot be satisfactorily answered and you'd better get used to the idea. There is no well-defined trajectory for a subatomic particle. There are true theorems that nevertheless cannot be proved. There is no absolute rest frame. In fact, progress is usually made by jettisoning those old concepts and questions that shackle us. And my guess is that at least some of the terms "thought," "intelligence," "consciousness," "free will" will have to be relegated to the dustbin before we can really make progress. (Maybe not. Maybe only some of them. But I'll willing to hit the eject button on any and all.)

That is my ultimate response to the Dijkstra quote that SC began this thread with. Asking whether or not a computing is "thinking" is probably besides the point and will only hinder you. So let's burn those terms and make some progress! that's what I say.

[sanscardinality]

Very glad the quote got such an interesting dialogue going! I'll throw in my $.02.

For my part, I don't subscribe to any particular (or vague) theory of mind - there's just not enough evidence to prove any given perspective that I've come across, though I'm open to being shown my ignorance on subject. I think windwalker is likely correct, but I'll wait for more solid evidence. I don't think Lewis is correct, but that's a gut instinct. I find the Buddhist view practical given our state of ignorance - it's something other than 1) the self 2) other than the self 3) both or 4) neither.

I agree with caliban that many philosophical terms are used loosely and refer to things that cannot be explained adequately to measure in any useful way (correct me if I've paraphrased poorly). I have a few to add:

1) Matter/energy: Stuff in other words. We cannot reduce it to a truly atomic (the philosophical term - not the scientific one - atoms as we describe them scientifically are not atomic philosophically) level, and therefore we cannot quantify it. We can measure it's effects, but cannot describe it.

2) Potency/force: Again, sensible in a relative context, but devoid of cardinality and therefore a description of effect rather than first principle. I do realize I'm using philosophical terms, but science did begin as an attempt to deal with these sorts of things. This is a relatively uninformed view and I'm very interested to see if caliban has a different view as a physicist and some references to point me towards.

3) Knowledge: By which I mean actual understanding/comprehension of any process or substance to the root. Until we understand forces and things, all knowledge is also relative and therefore only accurate to the degree we can engineer with it. Light is an easy example here.

Dijkstra is correct in my view in that it is premature to try to consider our messy universe in these absolute terms, but at the same time history has shown that a good dose of faith in human's abilities is a most productive delusion.

SC

[Windwalker]

We're fencing over semantics to a large extent. We are learning huge amounts about brain function, and will continue to do so.
//
But my training is in physics, and the history of physics has taught us that some questions simply cannot be satisfactorily answered and you'd better get used to the idea. There is no well-defined trajectory for a subatomic particle. There are true theorems that nevertheless cannot be proved. There is no absolute rest frame. In fact, progress is usually made by jettisoning those old concepts and questions that shackle us.
//
So let's burn those terms and make some progress! that's what I say.

We are partly arguing over semantics, but some of our differences also come form the viewpoints of our respective disciplines. The examples you chose are from the domains of the very small, the very large or the abstract. Cutting-edge physics and mathematics operate in domains which our sensory (and possibly mental) abilities were not designed to handle. Biology, on the other hand, is very much "the middle kingdom" in scale (although not in complexity). Biological phenomena can be understood by the senses, more or less, and nothing at their root is "exotic" -- though the outcome is!

That said, I couldn't agree more that old paradigms always outlive their usefulness and need to be changed or discarded. "Scientific theories start as heresies and end as superstitions."

[caliban]

I agree with caliban that many philosophical terms are used loosely and refer to things that cannot be explained adequately to measure in any useful way (correct me if I've paraphrased poorly). I have a few to add:

1) Matter/energy:
2) Potency/force:

Physicists have exapted these terms from common usage, to where they areprecisely defined--but not necessarily as commonly understood. Which is physics is so troublesome for so many people. They have to redefine "energy" and "force" from everyday usage.

And although I agree entire with Athena that one should be cautious about taking lessons from physics and applying them to a different scale (and vice versa--just because sexist and cultural assumptions frequently creep into sociological and anthropological studies is no reason to assume they also creep into physics and math), I still bet that "mind" "consciousness" etc. are too slippery to be used. Or if they are, they will be exapted for use by biology as have "energy" "force" and so on.

But I can give the precise physics definitions of energy and force, which are probably not satisfactory to the philosopher but have proved to be completely self-consistent for use in physics. Can any one give a rigorous or even consistent definition of mind or consciousness? We perceive something like them inside ourselves--but, as in zeroth order Buddhism, I suspect they are illusions.

[Windwalker]

Physicists have exapted these terms from common usage, to where they are precisely defined--but not necessarily as commonly understood.
//
I still bet that "mind" "consciousness" etc. are too slippery to be used. Or if they are, they will be exapted for use by biology as have "energy" "force" and so on.

Exactly! You posted this just as I was composing a very similar sentence.

Can any one give a rigorous or even consistent definition of mind or consciousness? We perceive something like them inside ourselves--but, as in zeroth order Buddhism, I suspect they are illusions.

It came to me that we're discussing two overlapping but distinct issues. One is the operational (testable) definition of a term. The other is the correspondence of a term to objective reality. The first is more rigorous and, by necessity, narrower. The latter shifts as knowledge accummulates and paradigms shift.

Consciousness as we perceive it may well be illusion -- but it is a real phenomenon, insofar as it arises from electrochemical changes that can be measured and which alter or cease during sleep, coma or death.

[caliban]

Until we understand forces and things, all knowledge is also relative and therefore only accurate to the degree we can engineer with it. Light is an easy example here.

This is an example of where science parts with Aristotle. Aristotle has his four causes, and clearly and not without motivation SC is concerned with more than one. But science is concerned strictly with knowledge we can engineer with--just one cause, the material cause, or maybe efficient cause, I get them confused.

Frankly, I'm pretty darn happy with knowledge we can engineer with. We know an awful lot about light! And SC and anyone is welcome to want something more...as long as they don't insist that I share their desires, that I have to agree with them that something is missing. Maybe something is. Maybe.

Maybe not.