Neurologist outlines why machines can’t think

[Justatruthseeker]

Well the statement that God is the fabric of the universe seemed to imply it.

Ok, I can see how that might be unintentionally misleading. but don’t we believe the fabric of the universe is energy?

Granted, I realize people actually think there was no fabric of the universe until the universe was created, but I don’t by that.

 

[Justatruthseeker]

Intentionality doesn't mean an outlook on life. It means that when you see an apple, that image comes attached to conceptual meaning. It isn't just pixels on a screen. I would say that even simpler animals display intentionality--I've got some fish that display somewhat odd feeding habits because of associations that they make. If you've got the genuine ability to learn without supervision and associate objects with concepts or feelings, I would say that you possess intentional states.

I don't know all that much about modern machine learning, but I do know that artificial neural networks draw their inspiration from the brain and work very differently than traditional task-specific algorithms, and it's the latter that critics of Artificial Intelligence tend to cite when talking about its impossibility. We can wonder how a machine might have a unique sense of self, but the same question applies to the biological brain as well.

I'm not sure there's any theory of mind that would rule out AI entirely. Even a theistic substance dualist really ought to grant the possibility that God in his goodness might bestow a transcendent mind upon sufficiently advanced artifical life. Some of the more biologically based naturalistic theories have serious problems allowing for AI, but I don't know if they're completely insurmountable.

I certainly don’t rule it out. I’m just not sure we are even close to the point were we are capable of making a program that is truly capable of independent thought. If it can’t rewrite its own programming as needed then it can never be truly be intelligent.

I don’t mean just learn and associate, but the capability to change its “mind” or even its entire outlook about the world.

It has to decide what programming is best for it, like we decide what we choose to believe or not.

 

[Justatruthseeker]

Well that is true; God could have made humans out of metal instead of flesh and blood if He had chosen to, and then doubtlessly there would be some foolish metal humans arguing that meat could never think like they can.

Lol, well until the meat does , but I don’t rule it out.

 

[Justatruthseeker]

Only if you limit your view to deterministic programs.

It doesn’t matter right now what kind of program. As shown in the video computers just don’t get it, while you and I have no problem with the same data being interpreted in multiple ways.

The line with a circle equals flower and tree or child and parent is simply beyond their ability to grasp. While even a child can grasp the concept.

One day in the future perhaps, and I hope I am still around when it happens.

I wish we could, I’d transfer into an Android body and go visit the stars.... Then you could call me dataseeker

 

[Strathos]

I agree that current technology isn't close to creating AGI, although perhaps that's because we are taking the wrong approach altogether.

 

[FrumiousBandersnatch]

But is it really probabilistic randomness, or just our inability to determine all the variables that lead us to this conclusion?

This depends on the particular interpretation of QM that you prefer. For example, both Bohmian Mechanics (pilot wave theory) and the Everettian 'Many Worlds' interpretations are determinate, and the apparent randomness of quantum outcomes are due to lack of knowledge (for the former, knowledge of the physical state of the whole universe, for the latter, knowledge of the ensemble of quantum universes).

The conventional or most popular interpretation, the Copenhagen interpretation, has it that the properties to be measured don't actually exist until a measurement or observation is made, but their appearance is governed by a probability density function (the wavefunction) that collapses to a definite value at the instant of interaction. This wavefunction 'collapse' is an unexplained ad-hoc addition to the fundamental maths of QM, and the Copenhagen interpretation is decreasing in popularity, as physicists get used to the idea 'Many Worlds', which has no ad-hoc additions or assumptions.

But whether it's 'true' randomness or pseudo-randomness, I still don't see how it usefully contributes to free will. For myself, I want my decisions and choices to be, as far as possible, determined by my inherited genetics & predilections, and the influences of my life experiences, the things that have made me uniquely me. The brain is a 'noisy' system, so there will always be some effectively random contributions, but they only make my responses slightly less reliable or predictable.

 

[FrumiousBandersnatch]

I specified "simple algorithm" because I think it's hard to argue that a piece of traditional computer programming possesses intentional states.

Agreed.

I'm considerably more open-minded about modern learning systems, though I would be interested in hearing why you think some already do have limited intentional states.

If intentional states are mental states that are about other things (in the world or other mental states), they are states with referrents; in this sense, it seems to me that any system that has a representation of some other thing, and uses this representation in lieu of or with respect to that other thing, has intentional states in respect of that thing.

This means, for example, that a system that can represent aspects of its environment, then act in the environment (e.g. navigate) according to its representation, can be said to have intentional states in respect of those aspects of its environment.

I am not sure what you mean by "poorly defined in functional terms," though. If you mean that these issues are hard to approach empirically, I agree, but that's the nature of the beast when dealing with the subjective side of theory of mind.

The problem I see is that if intentional states are to do with representation and/or modeling, then, by a kind of reductio ad absurdum, all computational systems, and even simple mechanical systems, could be said to have intentional states if they manipulate and/or act on internal representations of things. For example, even a thermostat can be said to represent environmental temperature in the bending of a bimetallic strip, and to act on certain states of that representation... The sole purpose of computing systems is to manipulate and act on representations of other things - but where is the line between simple computation and an intentional state?

Clearly, a thermostat doesn't have mental states, but then how do we define a 'mental state' in this context?

The definition of 'intentional state' I used above is obviously inadequate, so what are the functional characteristics of an intentional state? i.e. how do we recognise one? what behaviours characterise an intentional state?

And when attempting to define precisely what constitutes an intentional state, we should be wary of begging the question by requiring irrelevant properties, e.g. "intentional states are representations of things in a human brain, therefore, by definition, other animals can't have intentional states..."

I would be surprised if there isn't a pragmatic definition somewhere, but I don't recall seeing one.

 

[dgiharris]

In the last 150 years we went from riding horses to literally standing on the moon to inventing means for every single person on the planet to talk to and/or send messages to every other person on the planet instantenously.

It is the height of hubris to definitely declare "anything" impossible.

Computer Science and Neuroscience are both technological fields that are in their infancy when compared to many of the other sciences and disciplines: Physics, Chemistry, Math, Electrical Engineering, Mechanical Engineering, etc.

I feel declaring AI impossible now is akin to physicists and mechanical engineers in the 1700s declaring flying machines impossible.

We don't even really truly understand the human brain and consciousness yet, so how can we then go on to declare artificial intelligence and self awareness "impossible".

Sure, maybe it is impossible, but I would argue strongly that we are in the infancy of our knowledge about such things and really shouldn't be making predictions. Of course, we don't "think" we are in our infancy, much like Physicists and engineers back in the 1700s we'd think we know "most" of all there is to know simply because we know more now than any other humans in history (by simple virtue of a linear time).

But I guarantee you that if we could view the world a hundred years from now, they'd laugh at our current understanding of Neuroscience and Computer Science.

 

[Petros2015]

But I guarantee you that if we could view the world a hundred years from now, they'd laugh at our current understanding of Neuroscience and Computer Science.

Either that or they will envy our running water.

 

[Justatruthseeker]

This depends on the particular interpretation of QM that you prefer. For example, both Bohmian Mechanics (pilot wave theory) and the Everettian 'Many Worlds' interpretations are determinate, and the apparent randomness of quantum outcomes are due to lack of knowledge (for the former, knowledge of the physical state of the whole universe, for the latter, knowledge of the ensemble of quantum universes).

Our lack of knowledge is only exceeded by our belief that "this time" we've got it all figured out. Of course that is what everyone thought before too.

The conventional or most popular interpretation, the Copenhagen interpretation, has it that the properties to be measured don't actually exist until a measurement or observation is made, but their appearance is governed by a probability density function (the wavefunction) that collapses to a definite value at the instant of interaction. This wavefunction 'collapse' is an unexplained ad-hoc addition to the fundamental maths of QM, and the Copenhagen interpretation is decreasing in popularity, as physicists get used to the idea 'Many Worlds', which has no ad-hoc additions or assumptions.

Yah well, you are certainly welcome to believe that cats aren't alive or dead until we open the box, but when it stinks up the room, don't blame me.

It's that flawed theory that led to the incorrect belief on the dual nature of light.

But whether it's 'true' randomness or pseudo-randomness, I still don't see how it usefully contributes to free will. For myself, I want my decisions and choices to be, as far as possible, determined by my inherited genetics & predilections, and the influences of my life experiences, the things that have made me uniquely me. The brain is a 'noisy' system, so there will always be some effectively random contributions, but they only make my responses slightly less reliable or predictable.

We have free will. Knowing all the variables would help you determine that the rock didn't fall randomly on any given day, but even knowing all the variable won't ever stop you from changing your mind and doing what you choose to do. Granted the variables may make you want to migrate in the winter, but you can instead choose to put on a coat and stay. We change the variable by our actions, unlike particles which only change due to the effects of other particles.

Yes, you could argue we merely react to the actions of other particles, but you can wade upriver, while the particle is forced downstream.....

 

[Doctor.Sphinx]

I'd like to hear peoples' definitions of "intelligence". Computers, in my mind (and I've worked on them for nearly 40 years) will never be intelligent. They can store vast quantities of information, and they can make super-fast computations, but they can't reason. They only deal with inputs they've been given (or have derived from those they've been given). We have nothing to fear about AI because the computer's "I" is quite a poor substitute for human intelligence.

You could have something similar to Terminator, but it would be a man behind the machines.

 

[FrumiousBandersnatch]

Yah well, you are certainly welcome to believe that cats aren't alive or dead until we open the box, but when it stinks up the room, don't blame me.

It's that flawed theory that led to the incorrect belief on the dual nature of light.

Schrodinger's Cat wasn't a theory, it was a reductio-ad-absurdum critique of a version of the Copenhagen interpretation of QM. There's been a lot of water under the bridge since then. Most of those who still prefer the Copenhagen interpretation think wavefunction collapse doesn't work that way. I prefer interpretations without wavefunction collapse.

We have free will. Knowing all the variables would help you determine that the rock didn't fall randomly on any given day, but even knowing all the variable won't ever stop you from changing your mind and doing what you choose to do. Granted the variables may make you want to migrate in the winter, but you can instead choose to put on a coat and stay. We change the variable by our actions, unlike particles which only change due to the effects of other particles.

Sure, subjectively we have free will, but if your choices are the deterministic result of processes in your brain, your will itself is deterministic. All that means is that there are always reasons for what you do - even changing your mind. The alternative is purely random, meaningless choices.

 

[Silmarien]

Sure, subjectively we have free will, but if your choices are the deterministic result of processes in your brain, your will itself is deterministic. All that means is that there are always reasons for what you do - even changing your mind. The alternative is purely random, meaningless choices.

That's a bit of a false dichotomy, though. There's the popular two stage model of free will, where it's a combination of limited forms of indeterminism and determinism that together allow for self-determination: our decisions are determined by our own motives and deliberations.

I distrust any metaphysical approach that writes off too much of the world as we experience it as illusory--and this includes approaches I find appealing, since I've developed a taste for ontic structural realism and Pythagorean idealism recently. There are too many theories of causality out there to commit to a single paradigm.

 

[FrumiousBandersnatch]

That's a bit of a false dichotomy, though. There's the popular two stage model of free will, where it's a combination of limited forms of indeterminism and determinism that together allow for self-determination...

The possibility of quantum uncertainty influencing the macro-scale activity of the brain, and the effective randomness (pseudo-randomness) of chaotic activity, the complex interactions of molecules, and thermal noise, all mean that some degree of indeterminism is likely to be involved in brain function - but clearly not enough to significantly degrade its function, or we wouldn't survive.

But if free will is just the result of a little indeterminacy in the system, should we consider a computer system prone to give occasional incorrect results because of excess thermal noise in its circuits, to have free will?

... our decisions are determined by our own motives and deliberations.

So, are your motives baseless, or are there reasons for them? are your deliberations based on reasoning?

I distrust any metaphysical approach that writes off too much of the world as we experience it as illusory--and this includes approaches I find appealing, since I've developed a taste for ontic structural realism and Pythagorean idealism recently. There are too many theories of causality out there to commit to a single paradigm.

It's good to be skeptical. I'm particularly sceptical of propositions that seem incoherent.

 

[Silmarien]

But if free will is just the result of a little indeterminacy in the system, should we consider a computer system prone to give occasional incorrect results because of excess thermal noise in its circuits, to have free will?

There's a difference between a bit of indeterminancy in the system being a necessary condition for free will and it being a sufficient one. I think that the key is that a system needs to be able to act upon itself and influence itself in holistic fashion, and if a computer system manages to do that fully unsupervised, I'm unlikely to deny that it has free will.

So, are your motives baseless, or are there reasons for them? are your deliberations based on reasoning?

We're equivocating between different meanings of "deterministic" here. Saying that our decisions are grounded in our desires and reasoning doesn't mean that our behavior is compelled by them--we have options available to us (indeterminism) and we ourselves choose between them (determinism). The difference between libertarianism and determinism is that the former would entail that we as holistic systems are making those determinations in a top-down fashion, not that they can be reduced to physical processes beyond our illusory control.

(I may be halfway between compatibilism and libertarianism, but I think that's fairly common. Free will doesn't mean we can grow wings and fly by wishing for it.)

It's good to be skeptical. I'm particularly sceptical of propositions that seem incoherent.

Me too. But as always, I disagree with you on which propositions are incoherent.

 

[Justatruthseeker]

The possibility of quantum uncertainty influencing the macro-scale activity of the brain, and the effective randomness (pseudo-randomness) of chaotic activity, the complex interactions of molecules, and thermal noise, all mean that some degree of indeterminism is likely to be involved in brain function - but clearly not enough to significantly degrade its function, or we wouldn't survive.

But if free will is just the result of a little indeterminacy in the system, should we consider a computer system prone to give occasional incorrect results because of excess thermal noise in its circuits, to have free will?

Good point.

So, are your motives baseless, or are there reasons for them? are your deliberations based on reasoning?

It's good to be skeptical. I'm particularly sceptical of propositions that seem incoherent.

Agreed, such is my reasoning for not accepting the dual nature if light.

 

[Justatruthseeker]

Me too. But as always, I disagree with you on which propositions are incoherent.

Which rules out anything but freewill.

 

[FrumiousBandersnatch]

There's a difference between a bit of indeterminancy in the system being a necessary condition for free will and it being a sufficient one. I think that the key is that a system needs to be able to act upon itself and influence itself in holistic fashion, and if a computer system manages to do that fully unsupervised, I'm unlikely to deny that it has free will.

OK, that seems reasonable.

We're equivocating between different meanings of "deterministic" here. Saying that our decisions are grounded in our desires and reasoning doesn't mean that our behavior is compelled by them--we have options available to us (indeterminism) and we ourselves choose between them (determinism). The difference between libertarianism and determinism is that the former would entail that we as holistic systems are making those determinations in a top-down fashion, not that they can be reduced to physical processes beyond our illusory control.

Sorry, I was using the philosophical definition of 'determinism' meaning that events are the result of (caused by) prior events, and 'indeterminism' to mean that events are not caused by prior events (not deterministic); indeterministic effectively means random or stochastic; in physics, only certain quantum events are thought to be truly indeterministic (although this may well be a lottery-style artefact of our limited viewpoint), and it's thought that these don't generally propagate to macro-scales except in very specific circumstances. Randomness in the macro world is really the pseudo-randomness of deterministic but unpredictable influences (e.g. chaos, complexity).

My argument is that our decisions are the result of prior events, e.g. the life experiences that have configured our brains and how they process information, and those processes themselves. This processing may include some pseudo-random contributions (e.g. thermal 'noise') and, possibly, quantum indeterminacy. Since our brains function reliably enough for us to survive and prosper, it seems these disturbing influences must, in general, have only small effects.

(I may be halfway between compatibilism and libertarianism, but I think that's fairly common. Free will doesn't mean we can grow wings and fly by wishing for it.)

OK.

Me too. But as always, I disagree with you on which propositions are incoherent.

OK. It would help me understand your position if you could explain precisely what you mean by free will by means of an example where you can explain the deterministic and indeterministic contributions to it.

 

[FrumiousBandersnatch]

Agreed, such is my reasoning for not accepting the dual nature if light.

The so-called dual nature of light is an artefact of our attempts to visualise or describe it; the duality is a duality of aspect; i.e. it's not something that is both a particle and a wave, but something that has aspects of both a wave and a particle according to how you observe it.

A simple analogy is a cylinder, which is neither circular nor rectangular but appears circular when viewed end-on, and rectangular when viewed side-on.

 

[Radrook]

The real question is why are so many humans not intelligent?

Emotional interferences with thought processes. Also, android consciousness is far more difficult to attain than people imagine it to be.

Android Consciuosness
http://writerdreams.freeforums.net/thread/16/androids-consciousness