When will machines be as smart as humans?

[T'ai Chi]

Now, why would it take so long, seeing as how our understanding of the human mind and our computer technology have evolved in the last 60 years ?

I don't know of course that it will take that long, it is just my opinion. I think it is a much harder (by many orders of magnitude) problem than some assume.

 

[PixyMisa]

It's been a while since I read The Emperor's New Mind, but wasn't Penrose's conclusion less quantum-dependent than that? I seem to recall that all he was saying was "we don't know enough about science yet to make true AI", and that quantum effects may or may not have a hand in that lack of knowledge.

If it is the case that quantum effects can be shown to not have a hand in it, there might be still other areas of science where we don't know enough to make AI. So Penrose's contention can't be dismissed that easily.

At least that's what I remember of the book. I could be wrong.

Well, Penrose has argued that consciousness is directly tied to quantum gravity effects in the microtubules of the neurons of the brain. This seems to be both impossible and wrong.

But if it's not quantum mechanics - and it isn't - then all that is left is classical mechanics (including relativity for computability purposes). Arguing that Strong AI is false because there may be a yet undiscovered branch of physics is an extraordinarily weak position, since QM is extremely accurate and has already been excluded by application of QM theory.

 

[Beleth]

But if it's not quantum mechanics - and it isn't - then all that is left is classical mechanics (including relativity for computability purposes).

Unless there's physics yet to be discovered.

Arguing that Strong AI is false because there may be a yet undiscovered branch of physics is an extraordinarily weak position, since QM is extremely accurate and has already been excluded by application of QM theory.

But if, as you have pointed out, QM can't explain consciousness, and our current understanding of classical physics can't explain consciousness either, what other choice do we have?

I see us having two choices:
1) There is physics yet to be discovered.
2) Consciousness doesn't really exist.

Once you weed out the impossible...

 

[PixyMisa]

Unless there's physics yet to be discovered.

Doesn't work. Any new physics has to be consistent with QM, because QM has been experimentally verified in an enormous variety of situations. And we know that QM can't be directly involved in the production of consciousness, because brain processes work on classical scales.

But if, as you have pointed out, QM can't explain consciousness, and our current understanding of classical physics can't explain consciousness either, what other choice do we have?

What leads you to the conclusion that classical physics can't explain consciousness? Really, I see no evidence at all that supports this.

 

[Beleth]

Doesn't work. Any new physics has to be consistent with QM, because QM has been experimentally verified in an enormous variety of situations. And we know that QM can't be directly involved in the production of consciousness, because brain processes work on classical scales.

If brains don't work at the QM level then we don't have to consider QM effects. So be it.

In the same sense that QM wasn't consistent with classical physics, any new physics has to be consistent with observation, not with a previous theory. Saying that any new physics has to be consistent with QM comes dangerously close to elevating QM to "infallible" status.

What leads you to the conclusion that classical physics can't explain consciousness? Really, I see no evidence at all that supports this.

I haven't come to that conclusion. Classical physics may well be able to explain consciousness. But what we currently know about classical physics cannot. When I say "new physics" I use it in the sense that it could mean either an entirely new physical discipline, or further development in a discipline we already are aware of. Not just the former.

 

[Belz...]

So, consciousness is a classical phenomenon, classical phenomena are Turing-computable, and it is just a matter of MHz.

I don't think so. Mhz is just SPEED. That doesn't tell you about ability or anything. If you could make a calculator that ran on a 10 terahertz clock, it still could only do what a calculator does (although it'd be really fast.) It wouldn't be conscious.

In order for a computer to be conscious, it would have to, presumably, have a "neural" architecture at least somewhat similar in structure to our own.

 

[PixyMisa]

If brains don't work at the QM level then we don't have to consider QM effects. So be it.

In the same sense that QM wasn't consistent with classical physics, any new physics has to be consistent with observation, not with a previous theory. Saying that any new physics has to be consistent with QM comes dangerously close to elevating QM to "infallible" status.

QM is extremely well-tested and extremely accurate. But it's inconsistent with Relativity... Which is also extremely well-tested and extremely accurate. Where the two intersect - quantum gravity - is where the new theory will arise. That's why Penrose chose quantum gravity for his magic theories; however, this idea founders on the rocks of observed fact.

I haven't come to that conclusion. Classical physics may well be able to explain consciousness. But what we currently know about classical physics cannot.

What difference is there between this claim and your previous one? Why do you think that what we currently know about classical physics can't explain consciousness? What evidence do you have to support this assertion?

 

[PixyMisa]

I don't think so. Mhz is just SPEED. That doesn't tell you about ability or anything. If you could make a calculator that ran on a 10 terahertz clock, it still could only do what a calculator does (although it'd be really fast.) It wouldn't be conscious.

Assuming the calculator is programmable (and thus Turing-complete), I know of no theory or evidence that supports this assertion.

In order for a computer to be conscious, it would have to, presumably, have a "neural" architecture at least somewhat similar in structure to our own.

No.

A regular computer can do anything a neural net can do. Once your computer is Turing-complete, architectural specifics make no difference. That's what the Church-Turing thesis is all about. The only mechanism we know of which might make a difference is quantum computation, and we know that that is not what is going on in the brain.

 

[Beleth]

What difference is there between this claim and your previous one? Why do you think that what we currently know about classical physics can't explain consciousness? What evidence do you have to support this assertion?

Classical physics and QM are different "branches" of physics, right? It might be that the physics behind consciousness isn't an entirely new "branch"; it might be that it's just an extension of a current (probably the classical) branch.

I think that what we currently know about classical physics can't explain consciousness because people are still arguing about it. If it could explain consciousness, it would have, and things like this thread would not go on for as many pages as it has.

And you know as well as I do that one can't prove a negative. If you have evidence that what we currently know about classical physics can explain consciousness, please present it.

 

[Soapy Sam]

All objects which show any degree of awareness, are either biological creatures or the creation of biological creatures. To suppose there are examples in the second category begs the question we are asking- can machines be aware?

I find this significant. I tend to take from it the high probability that consciousness is itself a feature of life- of biochemistry.

One might argue that all flying things are either living or creations of living creatures, but that since aircraft can fly, it is not impossible that machines can be conscious.

There are problems with that argument. First it's hardly logically watertight. Second, flight is a simple process- make something with the right shape , move it and it will fly. This is true of other properties which could be used to argue in this way.
In contrast, where it is developed organically, awareness appears to be the result of great complexity.

My personal suspicion is that machines will never be self aware. I hold this oipinion reluctantly, as I loved robot stories. I do believe intelligence can be emulated though. I have said before that if a machine gun is sufficiently equipped and programmed to defend itself, it would be a formidable foe. The corpse on the ground would have few grounds to claim superiority over a computerised tank .

But intelligence in that sense is not awareness and to me seems to be a quite different type of process. I think intelligence can exist without awareness, as it already does at a primitive level in machines and in simpler organisms.
I also suspect, perhaps more radically- that awareness can exist without intelligence, though not without a brain or some organic structure of similar complexity.
In this sense, Intelligence is a process of manipulation of coded data which is (are) external to a processing engine (organic or inorganic). So what a pc does is intelligence.
Awareness is a process of mutual contact (signal and response) between individual processing units- eg brain cells, cells in a developing foetus, ants in a colony.
Intelligence is a property of a whole organism or machine, arising from processing of external coded data. Awareness is a communal property of a mass of nodes within an individual entity. An anthill is aware but unintelligent. Searle's Chinese Room is intelligent but unaware.

Viewed this way, language and memes are levels of intelligence grafted onto an entity , rather as several levels of programming language may exist at the same time in a machine.

The critical point is that awareness and intelligence need not be intimately linked- as, indeed, they are not in humans. CPolk pointed out earlier that human intelligence has access only to a window on human awareness- that is the bit which is self aware, the conscious mind. This for me defines consciousness- an interaction between awareness and intelligence. Whether intelligence is memetic, verbal, mathematical or something else, it has no access to the temperature of the rectum. Yet the nervous system is aware of it and monitors it at a chemical level. (And because we are conscious we can shove a thermometer in there if we want. At which point our intelligence knows too).

A neural net might be aware, yet have no way to communicate it's awareness to any intelligence it supported, because it is not self aware. It's awareness would be moderated by electron movement- indeed would be electron movement , just as it is (in part) in humans. But in the brain, far more things than electrons are moving around: Blood, peptides, molecules of many kinds carry information and perform other tasks. Any of these may be necessary to self awareness. Or all of them acting in concert.

We don't know. But rejecting (as I suspect most of us do) any elan vital or spiritual essence, we are left with hardware , wetware and software. Computer engineers tend to see the hardware and software and ignore the mushy stuff in the middle. Neurochemists study the gook and neurologists deal with the output intelligence.
I think awareness, intelligence and consciousness are products of chemistry- hellishly complex biochemistry, but chemistry nevertheless.
Of course I can't prove it, but invoking QM - or relativity or fairies seems equally daft to me.
The only things with awareness or self awareness have developed it after phenomenally complex biochemical development. I cannot believe that is coincidence. I suspect that while intelligence and even awareness may not be substrate dependent, consciousness may be. I think if we want conscious computers, we may have to grow them.

 

[elliotfc]

If the technological march doesn't stop, (which it may), at some point, machines will have to become as smart as us, won't they?

Well no. Let's say the technological march happens continually, and it's always marching half the distance to the goalline. And once it does that it marches half the distance again. And then it marches half the distance again.

But it keeps marching half the distance and never reaches the goalline. That's how it can keep marching and never get as smart as us.

Alternatively, as long as machines never stand around and talk to each other about reality TV shows and celebrities, it can be argued that they are already smarter than us.

-Elliot

 

[PixyMisa]

Classical physics and QM are different "branches" of physics, right?

Not really. It's all QM. Classical physics is the statistical application of QM to macro-scale systems.

It might be that the physics behind consciousness isn't an entirely new "branch"; it might be that it's just an extension of a current (probably the classical) branch.

That doesn't make any sense. What extension? How? Where? Why? There simply isn't anything going on in the brain that isn't explained by existing physics.

I think that what we currently know about classical physics can't explain consciousness because people are still arguing about it.

Well, that's not a very good reason. People are still arguing about dowsers and astrology and the psychic pet hotline.

If it could explain consciousness, it would have, and things like this thread would not go on for as many pages as it has.

Have you ever seen an Ian thread? Or a Kumar thread? Or those threads on raging kundalini or whatever it was? Von Neumann's incoherent attacks on evolution?

Just because someone argues about it doesn't make much of a case.

And you know as well as I do that one can't prove a negative.

I'm not asking you to. I'm asking you to present some evidence - anything at all - that actually suggests that something in the brain or in consciousness is not explicable by classical physics.

If you have evidence that what we currently know about classical physics can explain consciousness, please present it.

Consciousness is generated by the brain. There is simply too much evidence to rationally doubt this.

The brain is a biological system.

Biological systems are entirely explicable in terms of chemistry.

Chemical systems are entirely explicable in terms of physics.

QED.

 

[PixyMisa]

In this sense, Intelligence is a process of manipulation of coded data which is (are) external to a processing engine (organic or inorganic). So what a pc does is intelligence.

Okay.

Awareness is a process of mutual contact (signal and response) between individual processing units- eg brain cells, cells in a developing foetus, ants in a colony.

Computers do this all the time.

Intelligence is a property of a whole organism or machine, arising from processing of external coded data. Awareness is a communal property of a mass of nodes within an individual entity. An anthill is aware but unintelligent. Searle's Chinese Room is intelligent but unaware.

Searle's Chinese Room would argue cogently that it is aware. What is the basis of your claim that it is not?

Viewed this way, language and memes are levels of intelligence grafted onto an entity , rather as several levels of programming language may exist at the same time in a machine.

Yes, that is a valid way of looking at it.

A neural net might be aware, yet have no way to communicate it's awareness to any intelligence it supported, because it is not self aware.

Eh? If it's a net with feedback, rather than, um, a directed acyclic graph, then it can't help but be self-aware, on some level.

It's awareness would be moderated by electron movement- indeed would be electron movement , just as it is (in part) in humans. But in the brain, far more things than electrons are moving around: Blood, peptides, molecules of many kinds carry information and perform other tasks. Any of these may be necessary to self awareness. Or all of them acting in concert.

No.

Any chemical process can be modelled by a computer, and the product of any chemical process can be modelled without actually modelling the details of the process itself. It's just information. Saying that the biological process is required is ascribing properties to the biology that are physically impossible.

We don't know.

Yes we do. Phyics, mathematics, information theory, computability; we know that what you are suggesting is wrong.

But rejecting (as I suspect most of us do) any elan vital or spiritual essence, we are left with hardware , wetware and software.

Right.

Computer engineers tend to see the hardware and software and ignore the mushy stuff in the middle. Neurochemists study the gook and neurologists deal with the output intelligence.

Here's the thing: When you get down to it, it's all the same stuff. Hardware, software and wetware are interchangeable. That's what Church and Turing showed.

I think awareness, intelligence and consciousness are products of chemistry- hellishly complex biochemistry, but chemistry nevertheless.
Of course I can't prove it, but invoking QM - or relativity or fairies seems equally daft to me.

Yes. I think so too. The point is, this necessarily implies Hard AI. That consciousness is physically possible means that we can build a conscious computer. Consciousness is informational, and information is substrate-neutral.

There are two alternatives to this: one that appears to be untrue, that consciousness requires quantum computation, in which case we can still do it because we can build quantum computers; and one that is irrational, that consciousness is immaterial.

 

[PixyMisa]

Well no. Let's say the technological march happens continually, and it's always marching half the distance to the goalline. And once it does that it marches half the distance again. And then it marches half the distance again.

But it keeps marching half the distance and never reaches the goalline. That's how it can keep marching and never get as smart as us.

Well, except for the fact that technology has been advancing on an exponential curve, which is the exact opposite of what you are suggesting.

I showed (in another thread) that we could build a computer powerful enough to simulate the chemical activity of the human brain today. The reason we don't is that it would currently cost billions of dollars and take several years to complete the hardware, and longer than that to write the software. Because of the exponential growth in computing power, if we wait a few years before starting, we can achieve the same completion date at a far lower cost.

 

[PixyMisa]

Oh, and also that (a) Xeno's Paradox was solved hundreds of years ago and (b) if you've got half-way there, you can just buy two computers.

 

[69dodge]

I'm asking you to present some evidence - anything at all - that actually suggests that something in the brain or in consciousness is not explicable by classical physics.

Consciousness is generated by the brain. There is simply too much evidence to rationally doubt this.

The brain is a biological system.

Biological systems are entirely explicable in terms of chemistry.

Chemical systems are entirely explicable in terms of physics.

QED.

I agree that there is a great deal of evidence that consciousness is generated by the brain. That is an experimental fact. But I do not agree that current theories of physics predict that particular experimental fact. How can a theory of physics predict, for example, that I will see red if photons of a certain energy enter my eye, when the theory hasn't even the vocabulary to say "see red", but has only the vocabulary to say "photons of a certain energy"? Physics would be just as happy if high energy photons produced a red sensation and low energy photons produced a blue sensation. In what sense, then, can it be said that physics explains the sensations of color that I do experience?

 

[69dodge]

Any chemical process can be modelled by a computer, and the product of any chemical process can be modelled without actually modelling the details of the process itself. It's just information. Saying that the biological process is required is ascribing properties to the biology that are physically impossible.

How do we know what's physically impossible and what isn't? We only know what we observe. If it happens, it's possible. Otherwise, we don't know whether it's possible or not.

You keep saying consciousness is just information. How do you know this? If some computer is conscious, then a computer can be conscious. And if no computer is conscious, then no computer is conscious---in which case, consciousness isn't just information. How could you possibly know which is right?

Yes we do. Phyics, mathematics, information theory, computability; we know that what you are suggesting is wrong.

I don't know he's wrong.

 

[Belz...]

Assuming the calculator is programmable (and thus Turing-complete), I know of no theory or evidence that supports this assertion.

I know of none that supports the reverse.

A regular computer can do anything a neural net can do. Once your computer is Turing-complete, architectural specifics make no difference. That's what the Church-Turing thesis is all about. The only mechanism we know of which might make a difference is quantum computation, and we know that that is not what is going on in the brain.

But we still have no idea HOW consciousness is formed, so although we're certain we can, one day, replicate it, it's a little early to say that ANY computer with enough calculating power IS self-aware.

 

[Belz...]

All objects which show any degree of awareness, are either biological creatures or the creation of biological creatures. To suppose there are examples in the second category begs the question we are asking- can machines be aware? [...]

My personal suspicion is that machines will never be self aware. I hold this oipinion reluctantly, as I loved robot stories. I do believe intelligence can be emulated though. I have said before that if a machine gun is sufficiently equipped and programmed to defend itself, it would be a formidable foe. The corpse on the ground would have few grounds to claim superiority over a computerised tank .

I don't know why you believe that biological systems are so special compared to non-biological ones. Both are, quite simply, composed of the same basic stuff.

 

[Complexity]

I haven't come to that conclusion. Classical physics may well be able to explain consciousness. But what we currently know about classical physics cannot. When I say "new physics" I use it in the sense that it could mean either an entirely new physical discipline, or further development in a discipline we already are aware of. Not just the former.

Nonsense. You are babbling utter nonsense.

You seem to have little understanding of consciousness, physics, or science in general.

What evidence do you have for the statement that I bolded?