News for hackers. Stuff that matters.
Re:Can AI Become Conscious? by javajosh on Tuesday May 12, 2020 @03:44PM (#23158769)
Am I the only one who thinks the answer is obviously yes, and that it's kind of a waste of time to ponder it? It reminds me of discussions of free will -- completely boring because, even if we are all finite state machines (and we are, in some sense) then random inputs would mean we have free will and random inputs are assured by, for example, chaotic variations in the weather. In the same way, consciousness is pretty clearly characterized by a pattern that is constantly in motion, sensing both itself and the outside world. The only reason we think there's something magical about the phenomena is the residue of traditional religious beliefs.

One point against the traditional belief is that if you've ever read Aristotle you might be shocked to learn that the Greeks of antiquity believed that "soul" was quite physical, expressing itself in blood, semen, etc. The medieval characterization of mental illness as being an "imbalance of the humors" comes from this early idea about how we work. Needless to say, the definition of soul has retreated entirely into the metaphysical, where it's cannot be disproven, by definition.

Re:Can AI Become Conscious? by air7 on Tuesday May 12, 2020 @04:50PM (#23159589)
Off topic, but in my mind the answer to "free will" is obviously, boringly no. If we are, as you've put it, a finite state machine, the fact that random inputs are possible doesn't make the machine any less deterministic.

Come to think of it, this is not off topic at all. It follows that if AI can become conscious, it means that consciousness does not contain (or at least require) free will. That's if we agree that a computer running code has no "free will" no matter how complex the code is.

Re:Can AI Become Conscious? by mcguire on Tuesday May 12, 2020 @07:13PM (#23161262)
Since we've got yes and no, why not get the third answer:

The question is badly posed and therefore meaningless. If the universe is deterministic, then you cannot have free-will as it is normally considered. If it is non-deterministic, then (by Bell's Inequality, if you want to throw around big terms) it is random, which is also not free-will as it is normally considered. If you have free-will, as we all feel like we do, then the universe can be neither deterministic nor non-deterministic. Something is smelly in Denmark; the question itself does not make sense in it's current form.

I am not actually an analytic philosopher, I just emulate them as a hobby.

Re:Can AI Become Conscious? by tim333 on Wednesday May 13, 2020 @08:11AM (#23166007)
I'll give a fourth one - it seems obvious to me the discussions conflate two things. Am I free to choose what to do - obviously yes. Is that choice determined by the state of the particles etc I'm made of - also obviously yes. Whether you call that free will or not is a question of what perspective you take on it. From a common sense legal perspective I'm making decisions of my own free will rather than someone holding a gun to my head. From the weird perspective / definition of some philosophers no. And they argue on and on instead of accepting it's different definitions applied to the same reality.
Re:Can AI Become Conscious? by misja111 on Wednesday May 13, 2020 @02:32AM (#23164090)
No you're not the only one. It amazes me how strong the believe still is that our 'self' or 'soul' is something that can't be captured by machineries or neural networks. This belief has been around for centuries, and every time that science was able to cross some border that was thought to be unique for the human mind, the reaction of the people was to immediately come up with something new that proved that we humans were unique.

It is so obvious that what defines our 'consciousness' is this self-believe in itself and not much else. We are programmed by nature to see ourselves as unique and special, because this way we are more likely to defend our own interest and to reproduce our own genes at the cost of others. This belief is so strong that it makes it near impossible to imagine that it is a mirage and that we are nothing more than a bunch of molecules.

Re:Can AI Become Conscious? by hoseja on Friday May 15, 2020 @09:39AM (#23192001)
Then what is that thing behind your eyes, experiencing itself? It need not be magic but at the same time, there is not even a suggestion of true scientific understanding of the nature of it. There could probably be a robot created that perfectly mimics human input/output. The question remains, would that robot necessarily posses consciousness and what is it's nature?
Re:Can AI Become Conscious? by root_axis on Tuesday May 12, 2020 @04:40PM (#23159446)
> random inputs would mean we have free will and random inputs are assured by, for example, chaotic variations in the weather

Randomness would not equate to free-will, it'd be the opposite, since it would mean one's choices are completely disconnected from reality rather than a reflection of one's mental state and environmental circumstances.

In fact, a deterministic perspective is the only one where the idea of free will approaches coherency since it mandates that the actions of the individual are a function of who they are. If one repeats the scenario exactly, a person with free will must always make the same choice because if they didn't it would mean that their "will" is irrelevant with respect to what actually transpires in reality.

Re:Can AI Become Conscious? by _hexley_ on Tuesday May 12, 2020 @06:09PM (#23160608)
> One point against the traditional belief is that if you've ever read Aristotle you might be shocked to learn that the Greeks of antiquity believed that "soul" was quite physical

Aristotle's view is, to a large extent, the traditional religious belief (at least in the West). Aristotle would not have held that the soul was physical as you say, rather that the soul was the "form", or the animating principle of the physical body. The "magical phenomena" as the "residue of traditional religious beliefs" you mention is not really "traditional" at all. Descartes came along in the 17th century with his "res cogitans" and mucked up the whole issue by explicitly rejecting the traditional Aristotelian view.

Re:Can AI Become Conscious? by downerending on Tuesday May 12, 2020 @04:23PM (#23159225)
Arguably, if the answer was obviously yes, we wouldn't all be arguing about it. And as for "random inputs", I don't think that's the kind of free will most people have in mind--it's arbitrary, but doesn't involve people having control over themselves in any way.

My thought is that the defining characteristic of consciousness is suffering. If you can't torture it, it's not conscious. If you can, it is.

This would seem to rule out the programs that run on our computers. But, since we have little (if any) real understanding of reality, who knows?

Re:Can AI Become Conscious? by mewpmewp2 on Tuesday May 12, 2020 @05:48PM (#23160353)
How do you determine something can be tortured?

What if you remove the ability of a person to feel pain? Modify chemistry just enough to always be high as if she had injected heroin? Does it mean the person is not conscious any more?

If we can simulate all particles of a human on a supercomputer and all the reactions to torture will be the same is it not really tortured?

Re:Can AI Become Conscious? by downerending on Tuesday May 12, 2020 @06:40PM (#23160962)
The heroin example is a fair point. I'm not sure. In practice, it's hard to think of examples of conscious beings that don't suffer.

As to the latter, it doesn't seem that that would be torture. As I said, though, I think we ultimately know nothing about this subject. Nor can we, even in principle.

Re:Can AI Become Conscious? by Gustomaximus on Wednesday May 13, 2020 @12:21AM (#23163440)
I agree its likely but even if obvious, "it's kind of a waste of time to ponder it" seems so wrong.

Humans are incorrect all the time. Even researchers retract the previously 'proven'. Until the answer is clearly proven beyond all questions, pondering and researching is how we discover the unknown for that question, or error, and often so many other we didn't expect to find but come as a sideline of questioning the known. Intellectual curiously has great value for even boring and known of things.

Re:Can AI Become Conscious? by javajosh on Wednesday May 13, 2020 @08:05AM (#23165970)
But the alternative to pondering isn't "do nothing", but rather "do experiments". The philosophy of it all seems rather pointless. I don't have any references, but I'm sure there was a lot of ink spilled about the philosophy of flight, arguing that it was or was not possible a priori, before we actually did it. Was any of it useful? I don't think so!
Re:Can AI Become Conscious? by gfodor on Tuesday May 12, 2020 @03:45PM (#23158788)
It's not obviously yes, since we don't have an explanation for the experience of consciousness. Qualia truly is magical, unless you are in fact a p-zombie I'm talking to.

We also don't have an explanation for what I call 'the harder problem' https://twitter.com/gfodor/status/1225230653932761088

Re:Can AI Become Conscious? by javajosh on Tuesday May 12, 2020 @03:54PM (#23158895)
I don't have an explanation for the experience of digesting food at the cellular level either. But I still do it. You don't have to explain conciousness to have it.

As for qualia, I'm reminded of the people coming to my door asking, "Do you believe in Jesus?" I reply, "Do you?" and when they assure me fervently that they do, I follow up with, "Well, okay, how do you know you do?" I think you could do the same with r/Jesus/qualia/g.

Re:Can AI Become Conscious? by Trasmatta on Tuesday May 12, 2020 @04:14PM (#23159118)
I don't understand where you're going with the analogy of digestion. Nobody is saying you have to be able to explain consciousness for it to exist. And science can easily explain the mechanism of digestion, but consciousness remains a mystery.
Re:Can AI Become Conscious? by javajosh on Tuesday May 12, 2020 @04:30PM (#23159309)
I don't think you need to explain consciousness to have it, or to produce it in an AI. I suspect that an AI consciousness will be grown from a (reasonably complex) seed, more than constructed. And it will probably require a huge amount of slow, human interaction at first. (Ted Chiang has a wonderful novella called The Lifecycle of Software Objects that is the only SF that I know that explores this possibility).
Re:Can AI Become Conscious? by whylie on Tuesday May 12, 2020 @05:36PM (#23160208)
So you have a belief - the same thing as a person who believes in Jesus? The benefit of having an explanation is it doesn't matter at that point what one believes.
Re:Can AI Become Conscious? by gfodor on Tuesday May 12, 2020 @10:34PM (#23162830)
You suspect != it’s obvious.
Re:Can AI Become Conscious? by javajosh on Wednesday May 13, 2020 @08:11AM (#23166008)
A couple of your comments seem to imply you think my claim is that concious AI is objectively obvious. I was only asking if there are others for who it seems obvious -- clearly there are many that don't, and I don't take issue with them or their position! It's just I always read articles like this and it just strikes me as odd. I used an analogy in another comment about philosophizing about the possibility of heavier-than-air flight prior to us actually doing it. Which was a strange thing to get all worked up about, in hindsight.
Re:Can AI Become Conscious? by gfodor on Tuesday May 12, 2020 @10:33PM (#23162822)
The point is that without an explanation, claims of obviousness of the mechanics of consciousness are weak at best.
Re:Can AI Become Conscious? by mettamage on Tuesday May 12, 2020 @03:50PM (#23158843)
There are some siamese twins that are able to use the senses of the other (they share some brain regions, I believe) but it requires great concentration. There might be some insight there.
Re:Can AI Become Conscious? by supergarfield on Tuesday May 12, 2020 @04:18PM (#23159166)
That's true, but I think it's only "not obviously yes" in the sense that it's also not obvious that you or my mom are conscious.

It does seem like accepting that all other physical humans are conscious, and that there are no extra-physical phenomena in a physical human, implies that a simulated human would also be conscious.

Re:Can AI Become Conscious? by quantumsequoia on Wednesday May 13, 2020 @06:43PM (#23172716)
The answer could possibly be "no", not because of something magical about consciousness, but because of the limitations of AI. Our AI systems are all based on Turing machines. It's possible that simulating consciousness requires some hypercomputation not simulatable by a Turing machine.