Philosophical thread of the week: Could robots be conscious?

[John Q. Mayhem]

A friend and I have a recurring conversation about how robots are slaves.



We believe they are, but that's because humans don't yet consider them 'people'.

In related news, cars are quickly growing to hate humankind.

Drive carefully!

 

[Cheiromancer]

Umbran,

Computer chips may be more predictable than neurons, but they are just as deterministic. That is, what they do is determined by their previous physical state plus the laws of physics. Said laws are indeterministic (because of quantum mechanics) but that is why I say the two phenomena are equally deterministic. One thing that would make neurons "more indeterministic" is if there were non-physical "ghosts in the machine" that could affect the operation of the neurons. (i.e. Cartesian Dualism) The neurons would have behavior that showed a degree of independence of their previous state and the laws of physics that could not be matched by a physical system lacking such "ghosts". AFAIK there is no evidence of such "ghosts."

On the other hand, having neurons whose operation is indeterministic does not, for some theorists, to be enough to guarantee freedom, only randomness. If your actions are random then they are not "up to you" in the way that a free action would be. You don't have a choice about what you are doing; your actions are due to the indeterministic functioning of your neurons, not a choice of yours.

For instance, suppose one cluster of neurons contains a minuscule sample of radioactive material. If there is a radioactive decay (50/50 chance) then you will turn left. If there isn't, you will turn right. Whether you turn left or right is indeterminate. But it seems to be better described as a random choice, not a free choice that you can claim ownership of.

 

[Umbran]

Computer chips may be more predictable than neurons, but they are just as deterministic. That is, what they do is determined by their previous physical state plus the laws of physics. Said laws are indeterministic (because of quantum mechanics) but that is why I say the two phenomena are equally deterministic.

Er. Please don't take offense, but that is the most non-functional use of the word "deterministic" that I have ever seen (and in my line of work, I see the word used a lot). By your definition there, non-deterministic things don't exist, because all things in the universe follow the laws of physics. Determinism is not about following laws, it is about the predictability of the result.

A thing is deterministic if, given known inputs, it's resulting behavior can be predicted. Another way to say this - if the initial state and inputs fully determine the result, then the thing is deterministic. This is not the case for neurons. It is the case for transistors.

And, to pour more fuel on the fire - modern electronics is digital - meanign that a signal is "on" or "off", and nothign in between. Biological neurons are not digital, they are analog, and have a range of signal strengths.

For instance, suppose one cluster of neurons contains a minuscule sample of radioactive material. If there is a radioactive decay (50/50 chance) then you will turn left. If there isn't, you will turn right. Whether you turn left or right is indeterminate. But it seems to be better described as a random choice, not a free choice that you can claim ownership of.

Yes, your example is non-deterministic, and not free choice. But, really, what you have there is a deterministic stimulus-response arc hooked up to a random stimulus. That's not the same as a system that enacts logic through non-deterministic means.

We do have to be careful, as in this realm subtle failings of analogy mean quite a lot.

Just in case you missed it - I have already noted the fact that we humans may only have the illusion of free will. I just don't see as there's any point to considering that case. If we don't have free will, then this discussion (in fact, all discussion, and all human endeavor) is pretty much moot.

 

[Numion]

Fascinating stuff.

The proof Gödel used for his incompleteness theorem is pretty fascinating. How to use mathematics to prove that there are statements that can't be proven with mathematics? It's about constructing a mathematical statement that references itself, i.e. a statement that basically says that "this statement is false" (ye olde "this sentence is untrue paradox").

 

[Cheiromancer]

Er. Please don't take offense, but that is the most non-functional use of the word "deterministic" that I have ever seen...

I think my definition of determinism is standard for philosophers. It is the thesis that at any point in time, there is only one physically possible future. Indeterminism is the denial of this; it says that there are "forking paths" in the history of a entity. It used to be that science underwrote a determinist understanding of things, but quantum mechanics changed that (unless you believe in some kind of "hidden variable" theory, which I don't). However, macroscopic things like people might not be able to benefit from this kind of indeterminism.

Compatibilists say that determinism is compatible with free will, and libertarians believe that they are incompatible, but that since there is free will, then indeterminism must be true. The two sides tend to use different definitions of free will, though.

Whether something's future can be predicted, well, that's a different question. And I can see how a chaotic system, one sensitive to initial conditions, might be very unpredictable. And that could be called freedom, but for people, predictability is sometimes just called "character" and is an essential component of freedom.

 

[RedWick]

I'm keen on the idea of restricted/bounded free will myself. You're free to act within a certain range of behaviors, and the range of those behaviors changes in response to how you act.

As for the original topic: It'll probably happen quite by accident and people won't realize it's occured until after it's well established. When consequently studied and reverse engineered, an exact cause probably won't be found and we'll be left to wonder how the hell it happened. I'm eyeing the internet and our proclivity to hook everything we can up to it as the most likely source of machine intelligence.

 

[Umbran]

I think my definition of determinism is standard for philosophers.

*shrug*. It wasn't the one my philosophy profs used. But perhaps they weren't "standard". I find it hard to merit a definition that says "it is deterministic, though the result cannot, even in theory, be determined beforehand".

It is the thesis that at any point in time, there is only one physically possible future.

Geh. That way lies nihilism and a number of other philosopies I find unsavory. If there's only one possible future, there's not a whole lot of point to the exercise. But, of course, if there's only one possible future, whether or not I commit suicide in despair has already been writ, so what I think about it matters little, I guess

It used to be that science underwrote a determinist understanding of things, but quantum mechanics changed that (unless you believe in some kind of "hidden variable" theory, which I don't). However, macroscopic things like people might not be able to benefit from this kind of indeterminism.

We can certainly build macroscopic items that do display QM indeterminism. For the most part, on the macro-scale it all comes out in the statistical wash, but, as you say, if the system is highly sensitive to initial conditions, amplification can happen. If it'll work with a geiger-counter on a computer, it'll work for a brain. That's good enough for me

Whether something's future can be predicted, well, that's a different question. And I can see how a chaotic system, one sensitive to initial conditions, might be very unpredictable.

Where I come from "Determinism" is a matter of "can it be done, in theory" rather than "can I do it now, with the tools at hand". A deterministic system can be predicted, in theory. A system that is mathematically chaotic is (be definition) entirely deterministic. The question is whether we know the conditions and rules to high enough precision. Humans are either non-deterministic, or chaotic, but there's no way to tell at the moment which is the case.

 

[Turanil]

And, to pour more fuel on the fire - modern electronics is digital - meanign that a signal is "on" or "off", and nothign in between. Biological neurons are not digital, they are analog, and have a range of signal strengths.

Hey, there has been many interesting comments, but this one helps me a lot.

My own comment is about Free Will / Freedom of Choice. My idea is that a person who has to choose something will have to evaluate through rationale (intellectual) criteria, but knowing / being aware of only a few of them, not all and their possible implications; then emotional response, personnal history, neurosis, and instinctive behaviour will weigh in, most of the time at an unconscious level. Add to this, that often one doesn't have hours to ponder a decision, so must make a "choice" based on insufficient information. So in the end, a person is unable to make the "good" choice, meaning that if a person could really make the good choice, there would not be any choice at all. Now, "good" must be in relation to something, and thereafter, what if you think about consequences in other domains? etc.

As such, we could imagine a computer being able to make much more rationale decisions, but in the end he won't be able to have access / treat all information. In addition, our thought processes are based both on impressions of the world (via our five senses) which are limited, plus via our thinking which uses abstracts (words have a relative meaning, not the same for everybody). We can imagine that a computer would apprehend the world much better (scanners, etc.), but what about treating abstract concepts?

 

[Cheiromancer]

*shrug*. It wasn't the one my philosophy profs used. But perhaps they weren't "standard". I find it hard to merit a definition that says "it is deterministic, though the result cannot, even in theory, be determined beforehand".

Well, the definition at Stanford Encyclopedia of Philosophy defines determinism as "given a specified way things are at a time t, the way things go thereafter is fixed as a matter of natural law." But it points out that Laplace and Popper both expressed determinism in terms of predictability. That may have been where your profs were coming from. The notions are separable, but are also easy (and often useful) to commingle. It's also easy (although less useful) to see determinism as entailing fatalism, but they are distinguishable too.

Anyways, I don't want to go too far off topic. I think that since computers share the same physical nature as we do, then if we are conscious and exhibit free will, the same could be true of computers. And if it is only an illusion on our part, then computers can suffer from the same illusion. To cover my bases, I sometimes talk about "the experience of free will". Perhaps the computer chips that those computers use will be so small that quantum effects will render their precise operation unpredictable- I don't think this kind of unpredictability is required for them to be free or conscious, but I'm willing to grant that future computers might well be inherently unpredictable.

I'll grant the point, but I really don't think there is a contradiction between being predictable and being free. In fact, the reason why societies employ a system of reward, punishment and persuasion to control behavior is the thesis that free beings are predictable. If you could not predict that people would seek rewards, avoid punishment and live their lives according to rational principles, then a society of free beings would be impossible. People who egregiously flout the social code are judged to be insane, or at least suffering from a compulsion that has diminished their freedom.

Or take the case of when a commitment is freely undertaken. Marital fidelity, say. Your behavior in respect to that commitment should be utterly predictable in the future, but I would regard marital fidelity as a continuous exercise in freedom, not a diminishment of it.

Anyways, I think there is an intimate connection between consciousness and freedom. Between having experiences and having the experience of free will. If a system is capable of both perception and action, then the kind of modeling of the world necessary to have phenomenal consciousness is of the same order of complexity as the kind of causal modeling of the world necessary to have the experience of free will.

 

[Kemrain]

I don't think we'll ever know whether a computer is sentient or not, because we have no way to know whether or not another human being is sentient or not. Heck, we can't even be sure that we are sentient. I've heard arguments to the contrary, but they've never swayed me. I'm confident that because I'm here to observe the world that something exists, but I can't prove that that something is me in the same way that I am.

I'm not good at wording these thoughts, but I'm sure someone else can explain what I mean.

- Not Kemrainé Descartes.