More deep thinking for a Friday. The four past ponders are here, here, here and here. James S has joined Jennie with a later comment at the second so you might want to look at that too.This is arguable based on what we understand by the verb 'be'. But how far could the notion 'be' be a useful device and thought a self-sustaining thing, a closed loop willing itself into existence?
17 comments:
Is "be" what we need to look at, or "think"?
You put enough "if this then that" statements together and you might trick an AI into thinking it can think, but is it really thinking?
Good starting point. If we accept Descartes' statement, all that thinks 'is', but all that 'is' doesn't necessarily think. The two could be brought together if we accept the subjectivity of perceived reality, that what 'is' is actually a thing existing for each of us only through our thinking.
If the AI thinks it's thinking, it is thinking. There could be judgements made about whose thinking is 'better', but if we're making the judgements using our thinking as a model, they're surely shaky, at the very least for having a very narrow basis.
I am I think?
By the time I see these it's Saturday down here in Australia, and Saturday is no day for deep thought. I'll give it a bash though :)
The cogito doesn't suggest that thinking is a particularly human activity. I suppose it couldn't, seeing as the possibility of AI was so far away as to be unimaginable to Descartes. At any rate, why shouldn't an AI's existence be justified by thinking as far as a human being's is? Are we suggesting that "am" in this context means something like "possess sentience?" If so, then the statement is indeed circular: I am thinking therefore I can think. It's irrefutable but it's not reasoning.
Personally I have no problem with the idea that anything that thinks it's thinking is a sentient being. It's been a few years since I've read old Descartes but I reckon he was trying to say that since he was aware of his own thinking, he had a self and so his body was also real. I'm not sure that follows.
A more interesting issue to me is that Descartes' statement doesn't seem to be able to avoid solipsism. If we take self-awareness to be evidence of physical existence (which is dubious, as Dave's point about AI shows), then all we can ever really know is that we exist. We can't know if anyone else does.
Philip K. Dick wrote a great story about this. I forget what it's called right now but the main guy one day realized that there were no objects around him, only little notes with "table", "hot-dog man" etc, written on them. Once he was able to see the notes through the illusion, aliens came down and told him that he was one of their children and his education was now complete :D
I be therefore I self aware therefore I think am think not that was I aware of self
Heh, I used that terminology on purpose. If statements does not equal AI as far as I'm concerned, but they can certainly blur the line. (or do they?) The problem is to define intelligence.
But let me give you an example to walk you through my train of thought:
Say I write a computer program, and I'll use pseudo-code for simplicity sake:
get response from user:
If x = 2 then print "Hello"
Is that awareness? Not really. But if you put enough of those statements together, it starts to get confusing. I wrote a chat script once that monitored a channel, and had a list of words to watch for. Depending on those words, it would randomly put together sentences made from other lists. All "IF" statements and random rolls. It's not awareness, but someone talked to it for 10 minutes once. However, it's just reacting, it doesn't have any sort of will of it's own. Even if I had programmed in something to say "Every 5 minutes, say something random or wait another 5 minutes" it's still just reacting to commands.
Each year there's a contest to create an "AI" and judges talk with them and try to guess which are programs and which are humans. This too is just reacting to judges input.
So, is AI just a complicated series of reactions? Maybe.
Here's the ugliness of "I think therefor I am":
If a robot ever gained TRUE conciousness, is that AI? I'd say no. Carbon based, robots, and others we have yet to discover, who's to say that one is more concious than the other? However, is it really thinking? Now, does that matter if it is actually thinking? If you strung together enough IF reactions to make a computer appear real, how do you test it's allowance to life? If it screamed and begged as you attempted to delete it, should it still be deleted? It is, after all, just a series of IF statements - "If user tries to delete me, then turn on water works - monitor user's face, heart, etc to see what pulls their heart strings more, emphasize that type of cry for help"
Conciousness like ours is more than just a series of reactions... or, really, is it? How can you be sure? Does that make a series of IF statements a child? Are we merely a series of reactions in a larger process?
So we're alive. So what? We're carbon based. My laptop is computer based. Is it already a slave, bent to my will, a series of reactions, millions of lines of code, all designed to receive input from outside sources and create visual and auditory output? Google search preempts me as I type, guessing what I'm going to type next.
Next will we begin defining our own humanity by the role each of us fit into society? Janitors aren't really "alive". They don't have to think, they just push a mop all day, right? They see garbage, they throw it out, right?
To summarize:
What exactly constitutes "thinking"?
Can something be considered alive if it's just a series of reactions?
Is it enough to be able to appear alive?
Machines are already complex and anticipate - just because they don't mimic human behaviour, does that mean they don't fall into definitions of "I am"?
Something a little more right-brained from me today.
Something you might want to read, if you haven't already.
While machines can anticipate and respond this does not constitute thinking. They are still lacking the ability to imagine. When the point comes where a machine goes beyond even an infinite set of responses and predictions and creates its own frames of reference, then it could be said to be alive.
We ask the computer, it answers within the terms of its programming and to the best of its ability. Does it ever decide to lie or be economical with the truth? Does it withhold information that would go against its own self interest? Does it create possibilities within which to frame its responses? When the answers to these questions is yes, then we may have something to talk about.
I can't accept the premise that machines can never be alive because they are artificial constructs. If you are of a religious bent, then you have to consider that we too are artificial constructs, created by God (or the superior being or beings of your choice).
Are we therefore God to the machine? Will machines question our existence one day? Will they imagine that millions of years ago a toaster descended from the trees and walked upright on its rubber pads?
We are programmed to think by our nurture and nature, while machines are asked to select one or more options from a list of probabilities. Machines will think when they start adding to their list of responses themselves.
On a side note I'm always bemused by the way we use philosophy etc to justify our place as the superior being on this planet. We cry out that we are special because we can reason and imagine, resulting in a communal pat on the back that spreads round the globe.
Unfortunately the one thing we crave, recognition of this, doesn't happen. We get no round of applause from the other creatures we share the planet with, and a hearty 'Well done guys!' from the heavens is not forthcoming either.
I need to lie down now...
and while I think about it... Jennie's Time article deserves a discussion of its very own... nice find!
@ Jim Hale - Exactly.
@ James S - I'm not sure if the existence of a physical body was Descartes point, but it seems quite possible it was. In the case of an AI in the form we imagine it now, self-awareness - or awareness of thinking - would correspond to a physical existence too, in the software at least, even if not a body. Good point about the statement only suggesting the existence of the one perceiving, not the rest of the world. That said, if 'I think and therefore I am' means 'I think therefore I perceive existence', the question is still whether we can know we exist beyond that closed loop.
@ ArmChairGeneral - Language is definitely an issue here!
@ Dave G _ Nplusplus - It's hard to build on that. You've put and answered key arguments.
I would make one point re definition. We often assume that one characteristic of an AI will be the ability to interact with us. This needn't be true, although it seems to be a factor in attempts to create AI. Again, maybe we're the problem, that our judgements are shaky, on too narrow a base.
The thought I have at the end of your comment is this. If the program cries not to be deleted, and we delete anyway, perhaps it's a sign we don't make the grade? If we can't be sure the program isn't life, surely we shouldn't be deleting it, in case it is? As your comment shows, we're not really in a position to judge what deserves to survive and what not. This rather forces our hand: perhaps everything should survive because we could well be wrong? This would place a great burden on the creator, if the creator could be a killer too every time the memory is wiped.
@ Jennie - That's good music. I hadn't read the article, but the idea is familiar of course. The concept seems to break down into three general elements: the continuing development; the reaching the point of being comparable; the passing this point and everything changing. The continuing development is hard to argue with, although arguments of sustainability can surely be made. The reaching the point is hard to discuss, for the reasons already given here, by Dave especially, and in my reply. The question is really of what that point is. What is it in terms of structure and interaction with the world around it? It almost certainly isn't a point in fact, but a vast and hazy space we might have been in for a very long time now. If the point doesn't exist, then the change might be only a change in our consciousness, of where we are. As for the worst case scenarios, I'm not yet persuaded an AI could or would have access to the levers to bring them about, although I do accept it's well within the realms of possibility. In general it all seems very vague, despite there being a surprisingly specific date. I'd bet a profit could be made on prophecies like this too, and if so, that could influence the forms they took. There is of course also the danger the prophecies might be in part self-fulfilling.
@ Jim Hale - For me you expand very well on Dave's comment. The entity being able to add to the list of responses seems a measure for a step on the way, as does creating its own frames of reference. As for the speculation on origins, it shows how much fun this can be, and also how deadly serious. I agree on our excellent ability to justify ourselves and what we do. The again, we are also able to move beyond that, and do so of course, if too rarely. That round of applause will be a long time coming.
And after all of that I think I might be needing a lie down too. Good discussion.
I think we tend to be too hard on composing criteria for AI in any case. If we look at our distant ancestors can we honestly say that they had intelligence? If you go far enough back surely there will be one ancestor who had no more reasoning capability than most other creatures.
Our capacity for reason and self-awareness developed over an extended period of time and from very humble beginnings. Would it be safe to say then that early man was not a sentient being? As a species we didn't begin with 'I am' or 'be'... we 'became'. Therefore from similar humble beginnings is it not possible that machines will also 'become'?
This becoming will however be very rapid. In fifty or so years we have progressed from 'thinking' machines that have required human action to allow them to function, to machines that require only cursory maintenance to allow them to function.
I would also add a proviso to 'the ability to interact with us', as this already exists in a simple form, and change it to 'chooses to interact with us'.
I like the analogy with our development. The word 'choose' too.
I'd also like to add re the singularity idea that I completely accept the possibility a program will one day be able to do much that seems unlikely or impossible now, for example write poetry or a great novel.
Many good points being discussed here.
Although, I have to say, I'm tired of people pointing at Kurzeil with his singularity as the next big thing. I used to follow him and his updates. He's someone who had a lot of great ideas and his theories are still very strong. But he's also one of those people who's so very definitive of his predictions ("It WILL happen then") and isn't right all that often, just ignoring when he's wrong and repredicting.
2045 now, I believe it predicted for years we've already passed. There was even a conference where someone from Google couldn't make it, so they sent a wheeled robot instead and teleconferenced, and that day he was shouting about how the Singularity is here... except that's not AI, that's just an high end RC car. He can't even get his own terminology straight.
I dunno, I was innundated with his propaganda a while ago when the singularity was around the corner, and he just couldn't keep things straight. There's little that separates him from say Ray Bradbury for thinking up Data.
Lol, woops.. that would be gene roddenberry.
If I can mix possessive and contraction, you can certainly mix these..! You had me for a while though - I was looking for the link between Bradbury and Data and just couldn't find it. I'm very glad you came back with Roddenberry!
Porky said:
Good point about the statement only suggesting the existence of the one perceiving, not the rest of the world. That said, if 'I think and therefore I am' means 'I think therefore I perceive existence', the question is still whether we can know we exist beyond that closed loop.
Exactly, yes. Descarte's statement cannot avoid solipsism.
Post a Comment