Ask HN: Do you think machine consciousness is possible?
I'm planning on going to grad school to study AI because I think it is very interesting. For a very long time it has just seemed natural to me that computer scientists would eventually discover a way to make computers appear as intelligent as humans. Since it wasn't done yet, I wanted to work on this problem. I had no thoughts of solving it, but perhaps help it along.
However, I recently had the scary idea that machine consciousness may not be possible. I've thought this before, however this time it really hit me and scared me some. Considering I'd like to devote much of my resources to the problem, I'm now a little concerned that it may all be a waste. I'd prefer not to waste my life on something that turns out like the phlogiston theory.
Therefore, because it may bring good discussion and for my own benefit I'm asking:
Do you think machine consciousness (or at least something that looks like it) is possible? If not on current computer architecture, which "new lead" in computation do you think will allow it?
For extra credit: Do you think the Church-Turing thesis (anything that is computable is computable by a Turing machine) indicates that machine consciousness is possible?
188 comments
[ 1.4 ms ] story [ 176 ms ] threadMy personal take on this is that I'm really not sure.
There are some pretty clever people here that are sure it is possible, 20 years or less.
There are others (of which I'm one) that think it may be possible but either devilishly hard compared to what has been achieved to date or beyond our abilities, figure at least 20 years, probably much more, if ever.
And then there are those that think that it is impossible.
I'm an absolute nobody when it comes to stuff like this but it interests me greatly. When I was 15 or so I envisioned a world about 2 decades away where computers could be taught. We're 30 years down the line from that point and we're still programming computers more or less the same as back then.
But that does not mean that things can't change overnight, and who knows, maybe you're just the guy for the ticket and you will be the one to crack this nut.
What we call consciousness is probably very closely tied with having a physical body perceiving things it the outside world. So I think for a long time there will be differences between machine and man, but machines will eventually win out and become vastly superior to people. I just wouldn't count on it in your lifetime.
The really interesting question is: if we can quantify consciousness, what happens when we create something that's more aware and conscious than we are? Would we be considered sentient by a being that thinks a thousand times faster and in hundreds of thought-trains, lives for a million years, and can converse millions of ways simultaneously at bandwidths millions of times greater than speech?
We would be like insects to something like that, and it's not such a far-fetched idea or that far off.
Agree with your general thrust but it's the qualitative change that's the more interesting, as it's perhaps unknowable (in the way that your dog Fido will never understand most of your thoughts, no matter how patiently you explain them).
My personal belief is that the reason progress in AI has been so slow is that AI optimists vastly underestimated the computing power needed to produce good AI, and when our computers are finally fast enough (20-30 years perhaps, and the most important metric is probably not FLOPS but memory bandwidth) AI will actually not be that hard to achieve.
Having said that, consciousness is not a requirement for intelligence. It would interesting, and plausible enough for self-improving AI as smart as a human to be developed without addressing the question of consciousness.
For extra credit: No there is not enough evidence.
Here is my opinion on this matter: It is possible to simulate the universe, but it is not possible to be the universe.
Simulation is different from being, just as predicting the weather is different from manipulating the weather. And simulating consciousness on a computer is different from being conscious.
On the other hand, I know if I ever had to debate this question with Daniel Dennett or Eliezer Yudkowsky or any other capable person who believes there's nothing special about "consciousness" I would lose the debate.
Attributing a special meaning to consciousness, on the surface, is just as illogical as believing in religion- Neither seems to have a valid defense, as far as I can tell. Therefore, I find it disquieting that I DON'T believe in religion but DO feel so certain that computers can't be conscious.
Once you figure out the reason for this apparent inconsistency in my belief system, please let me know :)
Whatever it is that distinguishes consciousness for what computers can possible do is also the thing that makes you realise that computers can't be conscious. If you could logically write the reason down to 'prove it' then it would also be something that you could implement in a computer. So the reason, whatever it is, has lie somewhat outside of specifiable logic.
Just in case this is somehow original. I'm designating this 'thing' as a 'dejb' and calling the whole thing "dejb's theory/proposition/whatever". Although it actually probably just a restatement of Godel's incompleteness theorem. Also I actually believe that computers can be concsious.
However, he's just playing with your intuition, the same way you are, and his logic (though not his certainty, either, from the bit I've read) is flawed. You should read up on his "Chinese room" thought experiment yo learn why.
Now that someone brought up the S-word, I can't help but paste in one of my favorite Searle quotes (http://sss.stanford.edu/others/johnsearle/), which to me very clearly indicates how much of a mental and logical struggle it is (and how much flubbery is required!) to attempt to justify his position:
'Could a machine think?' My own view is that only a machine could think, and indeed only very special kinds of machines, namely brains and machines that had the same causal powers as brains. And that is the main reason strong AI has had little to tell us about thinking, since it has nothing to tell us about machines. By its own definition, it is about programs, and programs are not machines. Whatever else intentionality is, it is a biological phenomenon, and it is as likely to be as causally dependent on the specific biochemistry of its origins as lactation, photosynthesis, or any other biological phenomena. No one would suppose that we could produce milk and sugar by running a computer simulation of the formal sequences in lactation and photosynthesis, but where the mind is concerned many people are willing to believe in such a miracle because of a deep and abiding dualism: the mind they suppose is a matter of formal processes and is independent of quite specific material causes in the way that milk and sugar are not.
Even if the guy's a freaking idiot (which he is), you've gotta appreciate the way he tries to turn the "dualist" label against the people that are most opposed to dualism, while that's almost exactly what he's arguing for. He would have made a great politician; unfortunately, he's made his career arguing about something that he doesn't really understand, which is a damn shame, I hate to see so much mental energy spent fighting a lost battle...
This need for all beliefs to have a scientific basis is a bit pathological in my opinion. You shouldn't feel bad for believing something "illogical" if the only reason it is illogical is Occam's Razor or scientific intractability.
My reasoning for this is that it should be obvious to everyone that our world and experiences are increasingly shaped by ideas in modern times. Even though physics can explain the mechanics of the universe in great detail, the way ideas affect our individual experiences and society as a whole has very little to do with their ability to be objectively classified as true or false. By forcing a scientific perspective you are limiting yourself from the benefits of philosophies and ideas that are beyond the realm of science.
I'm not saying that a belief about artificial consciousness is beneficial one way or the other, but just that you shouldn't be chided into dismissing your own opinion just because it does not rest on firm logical conclusions based on established scientific evidence. The other side is quite wobbly as well.
Even with consciousness being so ill defined I'd say that there are a few points that we can agree on. For one, even if peoples definitions and expectations of consciousness differ, we can all find examples of life that we agree are conscious.
The boundary may be really fuzzy. We might disagree whether animals are conscious or brain-dead/comatose are conscious. But we could probably find a case that we agreed on.
And once we assume that I would ask you whether a computer program would ever be able to pass the Turing test and convince us that it was conscious. If a computer program would be able to pass I'd argue that it was conscious, since consciousness seems to me to be totally subjectively defined.
If you say that a program would not be able to pass, I might agree with you intuitively, but I would question it intellectually. Because I've always believed that there are cases that are definitely perceived as conscious by other people. I would be shocked if someone truly believed that the person next to me is somehow not conscious. So the next question I would ask is whether consciousness resides and arises from something that can be explained scientifically. Or to put it simply: Can we describe,now or in the future, the workings of the brain? If so, I really don't think we would need a scientific explanation for any epiphenomenon such as consciousness, all we would need to is recreate those initial conditions and we should be able to mimic consciousness.
If you say that we won't be able to mimic the workings of the human brain I'd simply say that you are making the statement that there is physical phenomena that can never ever be explained by science.
If you say that even if we recreate a human brain, we will never mimic consciousness, then I'd reply that you are saying that my personal definition of consciousness not affected and defined by other peoples definition of consciousness.
I think that might be why you see an inconsistency between attributing a special meaning to consciousness and and religion. We instinctively act like the human race can find common ground on the definition of consciousness. We don't instinctively act like we can find common ground on religious beliefs.
I define consciousness as having an internal model of the world that includes yourself, as well as your own though processes (at a lower degree of fidelity). This says how you compute things, not what you are computing, so it is orthogonal to Church-Turing.
Whether it is achieved will depend on economic forces. I don't see much economic value in making conscious computers, or things which seem to be down that line. So I expect consciousness will come out of pure research (perhaps within a corporation, like IBM), well after computers have exceeded the raw processing power needed.
Because they will be so different from humans, it will have to demonstrate a significantly higher degree of consciousness than a human needs to in order for most people to be comfortable with the term.
I realize it is not a defensible argument to say that the human experience is just to "weird" to be computable- I wish I had a better argument for defending my position...
Nonetheless, I find it hard to understand people, such as yourself, that are probably exposed to this same "weirdness" in their heads as I am and yet are so confident it is merely an illusion created by a sufficiently complicated computation.
I agree, though, Hofstadter and Dennett have done an extraordinary job of devising such an answer. There is lots more work to be done, of course.
But what if the answer to the "weirdness" question is, in fact, that it's just an ill-formed question? What if it's just an illusion, and your brain is tricked into seeing something magical about itself where there is nothing there?
Apart from a mystical explanation, I cannot imagine any satisfying answer to the question (i.e. one that doesn't leave you feeling uneasy like Hofstadter's answer does to most people), which usually means that there's something wrong with what we're asking, not with how we're trying to answer it.
I really don't think the question is ill-formed, though. There is a big explanatory gap. Denying that is simply dishonest. It would be like, if you don't know why a bicycle is easier to balance when it's moving, pretending that there is nothing to explain. "Oh, it's just bicycle parts."
From my perspective, being a computer would feel a lot weirder.
I think drastically spoken, feelings are just firing neurons. I don't think there is really anything special about (apart from the degree of complexity of the human body). If a computer has an algorithm that says
if(user hasn't typed anything for 14 days) lonely = true
then it has feelings, too. It might sound absurd, but only because it looks so simple. But imagine an enourmously complex program, and the information "lonely" trickles through it. There might be a routine somewhere
if lonely = true connect_irc_channel(#depression)
and so on and so on. From a certain level of complexity onwards, it won't be so obvious anymore, and we won't be able to prevent feeling that the computer really feels.
No - it's just normal. I think it would be weirder if it wasn't some kind of computation, if I was a unique spirit out of several tens of billions that appeared from some unknowable place and origin, or whatever else 'i' might be.
Doesn't it seem to be beyond what can be expressed by mere computer algorithms?
I have no reason to think it is - I haven't felt what it's like to be a quicksort or a face recognition algorithm for comparison, but I certainly feel like a heap of evolved feedback loops at times - when I fear the dark in a room I know, when I catch sight of living shapes where there are none, when I feel fight-or-flight at certain caller ID numbers, when I desire things that I also don't want, when I feel judgements of people based on some trivial detail.
Nonetheless, I find it hard to understand people, such as yourself, that are probably exposed to this same "weirdness" in their heads as I am and yet are so confident it is merely an illusion created by a sufficiently complicated computation.
What else could it be? There isn't anywhere else it could reasonably be that we know of. You're either suggesting something unknowable (e.g. magic) or some kind of cruel joke (i.e. like a quine is a program that prints it's own source code, we could be a consciousness that sees the world except for the blind spot around the part that allows us to see how we work - and a cover over where that would be).
I'm not confident that we are 'just' a computation, but I'm fairly confident that I am in accordance with the laws of physics (including any we don't know yet) and that places limits on what's possible with the amount of matter in my head, the energy input and output, the sensory input and output bandwidths, the timescales involved, the known behavioural results of varying localities of brain damage, etc.
Besides, what do you mean illusion? Consciousness is not 'fake'.
Of course, this only applies to intentional creation of consciousness. It says nothing about the distinct possibility that we could create a consciousness by accident.
We can already generate big enough NNs that we can understand them in general (some areas get "specialised" in some way and more active in some conditions), and we know that the result is what we expect - but there's no way someone will take a look at the weights / resulting model and tell you what it does. I think we can create consciousness intentionally, but not understanding the process completely.
Selectively breeding cows for increased milk production allows evolution to find a way or ways of making that happen. The breeder doesn't need to worry about hormones or glands. Evolution doesn't need to worry about the breeder's motives, why are only the high milk producers producing offspring?
Sometimes, when I'm getting overlogical, I see machine intelligence as being pretty much our destiny (i.e. that singularity thing) and that organic life may be very much obsolete once it gets into its own.
As organic life was to the universe before it, machine life will be to us.
ADDED: but yes, extended phenotype and all that jazz.
The design of machines evolves all the time.
It may very well be the case that this is either true or false. We simply don't have enough evidence.
And given the fact that we are discovering new properties of matter and organic reactions all the time, there is a bias towards this being false.
From the biological point of view, the simplest species is AS SUCCESSFUL as the most complex species, since both are still alive and procreating.
Why do humans have eternal significance? To whom?
Known to who? I don't like this very self-referencing way of thinking. Very similar to the discovery of America... people were already in America way before Columbus came and "discovered" it. (although apparently he didn't even realize he hadn't arrived to India)
He/She's allowed to have such an opinion, but this discussion is trying for a particular feel and that isn't contributing helpfully to it.
The poster you are responding did not claim that humans are the shit. They merely pointed out the fact that we have so far proven incapable of proving that we aren`t said shit. Now, you could argue that such a hypothesis could only be falsified by the construction of a machine consciousness. That, however, is orthogonal to the fact that it remains possible that there is in fact some bizarre quality of the universe or the human race that makes machine consciousness impossible. Not a terribly scientific position to take, perhaps, but a perfectly sound philosophical one.
No, human intelligence is an emergent quality, and I fathom that even our massive representations of humanity's information (akin to what Google is compiling) will soon begin to exhibit interesting qualities of its own once it becomes complex enough to exhibit perhaps interesting, unanticipated emergent qualities (in fact, if it didn't I would be absolutely shocked). Many strange and unpredictable (or at the very least unanticipated) things arise from even the simplest of "complex systems" (Conway's game of life and some of Wolfram's automata), much less the wonderful systems detailed by our individual neural mappings and our individual genomes. (q.v. http://en.wikipedia.org/wiki/Emergent_behavior)
People always think that some human will "write" an AI like HAL or the like, but it is much more likely that nature will roll its own AI once we have made a comfy enough nest for it to germinate. After all, isn't that how we got here? (from an evolutionary biologists standpoint anyhow...)
And again, sorry for being contentious. I just despise the "religious" argument (even if there is no "named religion" being expounded... Let's keep Ockham's Razor at the ready here....)
We're machines. That we may not have made a machine of the same class ourselves is an implementation detail. Maybe a large one, but still not really an argument against making conscious intelligent machines. Biology can do it, sooner or later so can we.
Perhaps the disagreement is about when do machines (keeping in mind they're mechanical by definition) cross over to becoming more and more like biological entities (replication). And once they do, is it still coherent to think of them as machines as we use the term today. Or would we then think machines have been elevated to the biological level.
The human has a hideous complicated algorithm in there, involving a live-long history, internal feedback and reflection and arbitrary side constraints.
A spider for example is much simpler. 'If the net shakes, walk where it shakes and eat.' (Certainly, + a batch of regulations to be able to walk and sense, but the point stands).
However, imo at a certain point, a computer program passes the complexity of, say, a spider. Just consider modern compilers. These things are of baffling complexity and do things inside no human can imagine in details :) Or, imagine data mining software, or even just very complicated, security aware network guards. All of these softwares are very, very complicated in their input-output-behaviour, and even though they are not as complicated as a human, they certainly can compete with a spider, at least for me.
And exactly this view is what caused some pretty nasty discussions for me, since some people are just not crossing the line of 'everything is an input-output device of different forms' and they stand hard on the ground that machines and animals are different, because they are machines and animals (and some go ridiculous ways, 'god made animals, humans made machines', and whatsoever, not even some 'but the spider might be more complicated than FOO, because, which whould be a nice discussion :) ).
So, overall, I, as a person who is working hard on being a tolerant, non-racist person (which is really hard), don't see a reason to exclude the possibility of machines and robots being conscious, just because 'they are electrical and not organic' (which has a ring of "he is black, he CANNOT do science" to me. Sorry if I just offended a lot of people).
"Consciousness is not computable" just means we need to build a new class of machine. Whether it's a blob of organic jelly in a box, or an IC with currently unknown structures on it to exploit certain handwavy quantum processes, or whatever; it just makes the task more complex, it doesn't make it impossible.
The only two physical process that we know of that are fundamentally impossible to compute with our computing model are quantum computations and chaotic systems. And for chaotic systems we most certainly can simulate them in a way that the output has all the correct properties as far as we know. It's more that we can't reliably do prediction in such systems due to finite precision.
The simple fact is that computers are already our superiors on many tasks. And those tasks are not simple either.
It may very well be the case that this is either true or false. We simply don't have enough evidence.
It's true, we can't prove it one way or the other at the moment, but that usually just means that Occam's razor should guide our speculations and assumptions. Since the brain seems like all it's doing is performing some computations, why would we ever assume that its true function is to do something more than that?
And given the fact that we are discovering new properties of matter and organic reactions all the time, there is a bias towards this being false.
There's a good chance that the way the brain achieves its computations is indeed a bit more complex than, for instance, an artificial neural network, absolutely. But that's an implementation detail, and there's certainly no reason to assume that there's no computational model that can account for it - biologists are very close to having full working models of individual neurons already. Further, it's highly likely that the brain does its job in a biologically convenient way, not a logically convenient one, and I'd give good odds that there are a lot of logical simplifications that could be made to end up with a cleaner architecture that performs the exact same tasks.
Do you have any proof or any testing that can be done to validate that?
P.S. I am an atheist, therefore I do not believe there is anything metaphysical about the human brain. Maybe there is something special about our mushy carbon structure -- doesn't matter. In that case, we'll just build our artificial brain with wetware. Hardware is hardware is hardware (and physics is physics). Eventually we'll get an artificial neural net as complex as our biological one. The real question is software...
Machines are just an expression of humans' extended phenotypical effects on their environment.
Building a car by hand versus building one with a modern robotic assembly line both yield the same output.
Evolved entities (read: complex entities) are not designed.
One could follow all the steps involved in getting a human from his evolutionary ancestor to his current state by using a pair of magic tweezers to make each mutation in the genome happen at the right time and in the right way. Could you tell the difference based on the result?
(BTW, mutation is random, but evolution is not a random process).
Two example quantitative differences (although, I'm not sure why the specific exclusion of qualitative differences):
1) The human eye, for example, has flaws (that are solved in other similarly evolved eyes) that an engineer laughs at - if you were building a human eye, you wouldn't make the same "mistakes". The deviation from the "better" version is measurable.
2) Evolved entities exist because of procreation. And their only reason for existence is to assist genetic material to replicate. Constructed objects reproduce exactly 0 times and have 0 genetic (or other replicator material).
What are those flaws, if you happen to know? I'm curious.
An engineer would probably try a different ordering.
Any engineer would naturally assume that the photocells would point towards the light, with their wires leading backwards towards the brain. He would laugh at any suggestion that the photocells might point away, from the light, with their wires departing on the side nearest the light.
Yet this is exactly what happens in all vertebrate retinas. Each photocell is, in effect, wired in backwards, with its wire sticking out on the side nearest the light. The wire has to travel over the surface of the retina to a point where it dives through a hole in the retina (the so-called "blind spot") to join the optic nerve.
This means that the light, instead of being granted an unrestricted passage to the photocells, has to pass through a forest of connecting wires, presumably suffering at least some attenuation and distortion (actually, probably not much but, still, it is the principle of the thing that would offend any tidy-minded engineer). I don't know the exact explanation for this strange state of affairs. The relevant period of evolution is so long ago.
That said, this isn't, for me, a strong refutation of design. Perhaps the designer's purpose was to create creatures with imperfections such as these, with the higher purpose of communicating something deeper.
Or, is there any way of knowing that we won't some day discover there is a very good reason for this?
Therefore, I don't consider designing vs evolving as a good way to separate machines capable of consciousness from machines not capable of consciousness.
Organisms (and by extension, or by reason) are alive because their ancestors have been lucky enough to survive long enough to have offspring.
Note, I'm not arguing that machines cannot be capable of consciousness.
But I am saying that constructed machines are necessarily distinguishable from organisms primarily because they're designed with a goal in mind (unlike organisms - these aren't designed but are the result of the co-operation of genes into higher levels of complexity, team work that has happened to be useful to the survival of the replicating matter - DNA, RNA, or possibly other such material elsewhere in the universe).
I would posit that because of the evolved nature of organisms, there may be flaws in their consciousness (e.g. through chemical imbalance, irrational deduction etc.). It is more likely that machines that are designed for intelligence & consciousness would be gifted (or cursed, depending on your convictions) with perfection, rationality, normalcy, as attributes, or at least their designers would attempt that.
So the the nature of their consciousness would be qualitatively different from that of organisms.
From this I think there are two potentially interesting products:
- Evolution by natural selection may be usable as an engine of machine consciousness. Some variant of evolution may be at the core of our own consciousness.
- Evolution by natural selection is a mechanical process that we can "take apart" and understand relatively easily. It also happens to be the engine of the process that created species. Happens to be. Even if the evolution of species had never happened, evolution would still exist in abstract. This put it in very good position to be discovered. Perhaps there are equally powerful concepts waiting to be discovered. Perhaps one of them is at the core of our own creativity. Perhaps one of them can be at the core of machine consciousness.
a) The phenomena of qualia remains unexplained. This has no bearing on making something that can mimic consciousness, but does become significant if one wants the real thing. The last argument on this: http://news.ycombinator.com/item?id=766462
b) QM may be intrinsically linked to consciousness. (see Roger Penrose, _The Emperor's New Mind_) Of course, this a mute point given the QM computing seems to be around the corner.
I discount Penrose's argument because he offers no phenomena that actually require his QM effects to explain and offers no plausible evolutionary path by which human minds could have evolved the mechanisms relying on QM. Do other primates have QM effects in their brains? What do these explain, given that the current models of the brain that we have are capable of explaining everything that we see? (Though we don't yet have the practical capability to simulate an entire complex brain and suitable io.)
If we assume that humans are conscious & material, without bothering to define either, we at least know it is possible to have consciousness embodied in something material. At least in theory we could figure out how human consciousness works and replicate it. There may be easier ways, but this is at least one theoretical possibility.
You could in the same way have inferred that humans could figure out a way of creating flight from the fact that it exists in nature.
The ethical issues will rather be the moment when AI's are clever enough that they can be teached. What will humans teach them? Personally I believe the best protection from abuse will be to get that learning process in the public (as compared to AI's learning from companies or military for specific purposes). So my own long term target are distributed virtual worlds in which AI's improve by getting passed around. A single computer user might still teach them bad stuff, but I hope some sort of selection based on many people watching and exchanging bots will get the best possible results even though completely preventing abuse won't be possible.
Erasing knowledge and memories of such AI would be analog to killing, and it will become problem, because throwing out data is easier, and writing you enemies to DVD after each game of Quake 10 becouse of moral issues will be problematic after some time.
If true machine consciousness is possible, it's a lot further off than we would like to think.
I don't know if hardware is holding back progress in AI, especially with distributed computing; it is more likely a question of software and programming. I think in order for a computer to exhibit a reasonable facsimile of human intelligence, it will have to do more than just run programs written by humans. It would have to have the ability to write and rewrite its own code.
Just to give you one example. Think of bots in shooters. Humans have no problem browsing the whole screen in realtime and evaluating the data. But try even just doing the simplest image processing in realtime with a typical resolution and you will already find out how hard that is. So bots can't do that. Instead they cheat and shoot a few single visibility lines and even those are already targeted because bots can know where opponents are and so they only do for example linechecks to certain areas like head, body, arm, legs and try to find out if that line is blocked. And even that might already be too slow with many opponents so you have reduce it further and shoot only a certain number of lines each frame. And there is no processing yet - this is only about getting the simple information of - is there something or not. This is how slow computers still are. You could certainly do a little more with giving a single bot the whole processing power of your computer, but with the increasing complexity of game worlds the needed processing power grows rather fast (current addition of simple solid-body physics in games makes AI programming already way harder). For AI tasks computers are still very slow.
Don't worry that AI will turn out like phlogiston. The journey will yield its own rewards, and plenty of partial success will also be extremely valuable.
Do you think the Church-Turing thesis (anything that is computable is computable by a Turing machine) indicates that machine consciousness is possible?
Only if you assume that consciousness is a computation, which is assuming everything.
I normally try to resist this topic, but what you're saying here tugs at my heart-strings:
I'm now a little concerned that it may all be a waste. I'd prefer not to waste my life on something that turns out like the phlogiston theory.
Consider the total failure of algorithmic approaches to even begin, even pathetically, to replicate anything recognizable as consciousness. Were the people working on it dumber than you?
Try to find some way to control for the geek fantasy factor, in yourself and others, before deciding what to do.
(By the way, now I'm curious: what was it that led you to "the scary idea that machine consciousness may not be possible"?)
It probably scared me this time and not others because I'm very stressed over the GRE, and grad school applications (namely: where the heck should I apply), and how me telling grad schools "Oh, I'd like to study machine consciousness" will go over. I could be wrong, but I'm concerned they wont take me seriously...so I've been trying to think of something that sounds more acceptable.
(In the same way that I have to believe you when you say you have consciousness).
We shouldn't assume that we humans are the best at everything. We shouldn't assume that we humans are the best at anything, really.
We can substitute pretty much anything for the first noun phrase: " X is as explanatory as magical pixie dust in the brain "
My dictionary defines consciousness thus:
consciousness |ˈkän ch əsnəs| noun the state of being awake and aware of one's surroundings : she failed to regain consciousness and died two days later. • the awareness or perception of something by a person : her acute consciousness of Mike's presence. • the fact of awareness by the mind of itself and the world : consciousness emerges from the operations of the brain.
====
We may not know HOW consciousness comes about, but we do have a definition of what consciousness is.
And now we're in an infinite loop.
If there is no such model or we can't find it for some reason we won't develop an AI.
I seem to have the problem also that we are using the thing we are trying to simulate to work out how to simulate it. From our perspective the problem could be almost impossible.
I believe consciousness is an emergent property of a complex system; it is simply the nature of the universe that complex systems exhibit consciousness. I'm defining "complex system" as any system whose outputs/results affect the inputs/possible states. If I think something, the possibility space for my next thought is dependent upon my previous thought, and so on.
I think artificial consciousness research will progress, through computer simulation, to the point where a "real" consciousness emerges from a sufficiently complex simulation. The hard part will be mapping its inputs/outputs into human-compatible form; success will probably occur by accident at first. But when we can do this we'll be able to talk to a totally simulated consciousness through the prism of it being another "person".
At this point more research/thought will be put into the nature of consciousness itself, and how to connect with other-than-human consciousnesses. We'll use the experience of bridging communication with artificial consciousness to successfully communicate with naturally-occuring consciousnesses associated with other complex systems (the earth, a tree, the galaxy, etc). It sounds a little insane, but I totally think this is within the realm of possible in our lifetimes.
Of course, that's all based on the notion that consciousness is an emergent property of a complex system, and not something entirely unique or bestowed by higher powers or whatever.
There are some problems simulating the human brain that would also have to be addressed even once we can create a working system, such as the AI being a bit of a blank slate, like a infant or a coma patient.
I see the whole process as having to follow a path similar to this:
1. A breakthrough in computing power, something capable of simulating very accurately small areas of space, this means perfectly representing ridiculously complicated chemical reactions and some natural laws.
2. Succeeding in creating a software environment to execute these simulations within.
3. A breakthrough in mapping an existing person, some sort of scan that creates a mathematically provable perfect (or close enough) representation of an area in space. Like some humans mind or possibly their entire body until the subject of the scan can be simplified on the computer. Sort of like taking a photo and then cropping off the body. The above simulation environment may be what is used to provide the simulated inputs and outputs to the head, like the CNS and cardiovascular system. Not to mention the inputs to the eyes and other senses. Sort of like a virtual head in a jar.
4. So far we would have a conscious system, but it would be a copy of a pre-existing being. The next step would to be to somehow, ethically, re-write this being. This would provide a learning challenge with the goal of simplifying and modularising the human brain. Such as hacking language areas, input nerves, the reliance on virtual blood and sustenance and most importantly the memory.
The final product of this important stage is the most simple and easily tweak-able simulation of the human brain that could be used by all researchers and eventually commercial applications. If all these virtual brains are the same or comparable, this isolates the memory as a way to load in or edit what is essentially... people. The creepiest analogy may be the best, they will be like swappable save game files, executing in virtual machines (the hacked brains) that operate within another virtual machine (dare I say it, a super-simple matrix of sorts)
5. We may never reach anywhere near this far along the process due mainly to ethical reasons that cannot be overcome with mere ingenuity. But if we do, the next step is compressing all this down further and further until we have the most simple possible (perhaps provable somehow) implementation of a mind that does not require all the layers of virtualization.
God it's easy to get caught up in this stuff. I hold this prediction on my fingertips in hopes that any developments may blow it away so I can re-evaluate and make a new one.
What is consciousness? Well, for me a big part of it is the "movie screen" of my vision. Consciousness is sort of like this movie I can watch, with the understanding that the hands that go into the field of vision are my own- "I" can control them and my body's position in the field. If I want the field to change, I can make the movie look somewhere else. My consciousness also includes the full mapping of the body, including its pain and pleasure states, temperature, hunger, etc etc.
Language is a large part of consciousness, in particular, English. Integrating language into my consciousness was pretty catastrophic- it was like formatting my conscsiouness with a new files system. Apparently it erased all memories previous to that because I can't remember anything earlier than that. English both enables and limits the scope of what I can think. It has a lot of ambiguities and is only moderately good at feelings, and emotions. But hey, it has tenses which vaguely relate to this sensation of passing time that is relevant to my consciousness and a significant portion of reality can successfully be communicated to me by any other consciousness running the Englihs module. If I wanted to be more precise I suppose I could convery my consciousness to LOGLAN or some other language built on predicate calculus rather than germanic and latin grammar base.
In addition to all this, my consciousness stores a lot of memories, skills, ideas, and memes. I have models for basic mathematical concepts, as well as skill sets such as how to operate a motor vehicle. I also have prejudices and preferences. My consciousness is limited by this- if I were not familiar with certain philosophical memes, such as "social contracts", my consciousness would not be able to understand advantages of a national healthcare plan.
Emotions are also part of my consciousness. There is happiness, a "feeling" of well-being when dopmanine, electricity, magnetism, radio waves, and more are flowing rather well. I am likely to be happy when the majority of my body and psyche's needs are met and I am pleasantly engaged and using enough, but not too much, of my energy bandwidth. Anger is a territorial response I would feel in my consciousness when I feel my phsyical or psychological territories are being threatened- it is a territory defense response. Sadness is a tension I feel when there is cognitive dissonance between an idealized vision I have of how things should be. (Tears of sadness if things are worse than I feel they should be, happiness if they are better.) Sadness may also be felt due to various social rejections as my organic ape-mind has a great deal of hardware devoted to social interactions. Pain is experienced when my nerves detect entropy is increasing in my wetware.
As for communicating with other sentiences, this seems inevitable as more species approach and cross Threshold. If humans are at 100 billion+ (or whatever Threshold requires), there are many speces which must be very close. Some of the more advanced marine mammals such as dolphins and whales are apparently very close, let's say 98 billion interconnections, and they even have some of the advanced cerebrocortical structures we think of as giving us our higher civilized thought structures. Elephants, chimps and other greater apes, and maybe dogs, cats, pigs, and some birds seem to be climbing above the 70 billion mark.
If we could figure out how to artificially enhance brains with more interconnections, both human and animal, we could probably "boost" several near-sentient animals across the line by nano-creating more neural interconnections. Energy would flow across the new connections and they would "burst" into their own version of movie-screen of consciousness. It would be fun having other sentiences to talk to. I think we as humans are fairly l...
As for the question, I am sure that it is possible (except I take issue with the word "consciousness" - what is it supposed to mean?).
I think it's painfully clear that sometime in the coming years humanity will reach the pinnacle of its scientific achievement with the advent of an artificially intelligent machine: one that is able to think and reason and is self-aware. Technology is moving at such a rapid pace and in the right direction that this is just the next logical step.
In order for this to occur, however, significant advances must be made in fields outside of technology; e.g. quantum computing will probably be a huge stepping stone, and for that to come to fruition we must first fully prove quantum mechanics.
I recently heard about the term "yak shaving". This seems to be a good example.
Human's consciousness is but one form, in my opinion. We are a complicated biological machine seeking the fulfillment of our existence; how is that any different than a machine seeking the fulfillment of its existence?
I believe the money is in self-evolving circuits and programming; allowing the machine to mold what defines its existence based on external parameters - and overtime, based on internal parameters (will to change).
Now here's the thing - you would still feel like you, looking out of your eyes. The other versions of you would be distinctly separate to you - not you.
That bit that makes you feel you are you - that's what some would call consciousness and others would call a soul.
Christof Koch at CalTech (hi, virgil) http://www.klab.caltech.edu/~koch/
Larry Yaeger & John Beggs http://www.indiana.edu/~rcapub/v30n2/mindmade.shtml
Of course, Douglas Hofstadter http://www.cogsci.indiana.edu/
Koch, Churchland, etc. speak on consciousness: http://mitworld.mit.edu/video/342
Thanks very much for the links. I am stressing over where to apply, hopefully these will help some.
Nick Bostrom's Simulation Argument details the most obvious probabilistic implications of substrate independence in consciousness: http://www.simulation-argument.com/
The most blatantly obvious indicator that consciousness is substrate-independent: We are DNA-based life forms. DNA stores information. It's program code stored in molecules. You are the product of the code of your parents. If for some bizarre reason we find out that we HAVE to use DNA to create other conscious systems, we will still have the ability to do exactly that. Not "machine" in the sense of being composed of metal, but certainly "machine" in the sense of not being the immediate product of natural selection.
David Chalmers' work should be particularly relevant to you, and you will find him at the Singularity Summit this year. http://en.wikipedia.org/wiki/David_Chalmers
Even if you don't live to see machine consciousness as a reality, the only other pursuits that might compare are anti-aging and intelligence enhancement research. If you're not going to create something that can figure out how to give you an indefinite self-contiguous narrative, you have to support its creation or face certain death.
I'm guessing you already have your CS undergrad or will have it soon, and you're interested in AI, so that seems the natural choice. I'd say you're overdecided if that's what you want to study.
Hopefully they have videos posted online of the event...right now two months rent is not available to spend on a conference, very unfortunately.
Current theories say consciousness is caused by a specific mode of representation/ computation. Like good scientists and rational thinkers we should submit everything to rigorous testing/ proof. Let us not please take things at face value.
[Or we can take the easy route and live in a bubble :) ]
But I think the question you should be asking yourself, if you're wondering how to spend your life, isn't if machine consciousness is probable, possible, whatever. You should be asking, what do I want to do with MY consciousness? If you want to work on these problems, then do it. Doesn't matter if it happens in your lifetime or not. If you enjoy it then give it a try. Worst case is you spend your time doing what you enjoy. Best case is... well, machine consciousness I suppose.
Besides, you should always be thinking big. The person I admire the most has a small metal plaque that sits on her desk. It says "what would you attempt if you knew you couldn't fail". I happen to think this is the most powerful sentiment I've ever encountered.