First, the web is only a system of hyperlinked documents which can be browsed by a man or a machine. This means that the web is only information, and as far as I can tell information is not intelligence. A system capable of using this information may be intelligent, but not the information per se. So, maybe the internet will evolve to a "superintelligence". But this leads us to my second point.
From the verb used in the question, emerge, it looks to me that this "superintelligence" is reached without human intervention. But as spontaneous generation does not exist, we need a system capable of evolve, without external help. And as far as I know, nowadays, neither the Internet (network) nor any of each nodes can evolve without human help.
So, until somebody invents an evolving device and connects it to the Internet, my answer is no.
In the everlasting words of Pat Cadigan ("Death in the Promised Land"), what you are describing is akin to putting a dirty shirt and a pile of straw in a wooden box for the spontaneous generation of mice.
It took an enormous amount of cumulative optimization pressure to turn self-replicating RNA strings into humans. The Web doesn't have anything remotely on that order, and no, the "reproduction" of corporations and startups doesn't even begin to approach the quantitative selection pressures involved.
I read Kurzweil's book "the age of intelligent machines" where he set quite a few predictions for the next 10-15 years. This was 15 years ago - and he got almost all of them right.
I just don't want to dismiss him - besides it seems arrogant that we should forever be the best evolution can offer.
People who think computers of sufficient power will just become self aware, don't really understand computers. If intelligent machines arise (and I think they will), it'll be because someone worked very very hard to make it so. Btw, it's arrogant to assume we're the best evolution has offered since you're defining best in human terms.
This is exactly the kind of philosophical question that pg talked about in "How to do Philosophy" that breaks down because of the near impossibility of objectively defining the terms involved, or even coming to a solid consensus on their definition.
Obviously, the key term here is "awareness". We "know" we are aware because . . . uh, because we think therefore we are, I suppose. Do our cells know about our awareness beyond their genetic programming to do what they do to contribute to it? Do we go out of our way to make our cells 'aware' of our super-awareness?
My point is that way down the road when there are exponentially more sensors, programs, humans, and other 'resources' connected to the web, how would we even know that it had become . . . something more? It's probably never, ever going to take over some terminal and say "Hello there, Dave". And we may never know, because just as you can't find consciousness in the brain, you won't be able to find it in the Net. It would be the ghost in the machine, that had little interest in its component pieces.
For those, myself included, who say this or that will 'never' happen, I'd like to add a quote from Freemon Dyson, talking about a multi-disciplinary conference where the future of human evolution was discussed: "When the physicists started talking about the really long term, [the others'] eyes glazed over."
I don't know that a superintelligence will emerge from the web, per se, but I think it is becoming a sufficiently large body of textual information that an appropriate AI program could use to bootstrap itself to amazing levels of intelligence. AI programs have a problem that they have to infer all their patterns just from text, which is fairly low-bandwidth. Humans (and robots) get much more bandwidth through vision, hearing, touch, and body feedback. This might not seem important, but it provides context around every piece of information that comes in, making it much easier to make sense of things and giving richer patterns to try to match against. Without that context, an AI would need to multiply the amount of text needed by 1M (1B? 1T? X?) to train itself the way a child does.
The problem with technology people making predictions about intelligence is they don't understand the most relevant field: epistemology. In particular, you can't make very good predictions about AI before you have a very good understanding of how learning and knowledge creation works.
You can tell someone doesn't know these things when they never bring it up, despite it being critical to their claims.
One way this is relevant is that people talk a lot about raw computing power. As if that was the bottleneck in humans. But the more critical issue, by far, is irrationalities. And they don't address whether AIs will have those, or not. Nor do they address how to "parent" an AI.
This particular article doesn't even mention traditions or memes.
heh. that is another reason it's important to be an expert on the subject: the mainstream view is very confused.
the whole knowledge is justified, true belief thing is such a mess. they've known induction doesn't work for ages, but haven't replaced it. good answers to these things, on the other hand, are largely ignored. (hint: Popper)
Let me begin by extending a large, rigid middle finger to those questioning the eventual validity of such an amazing article. Having said that, I agree with pchristensen in that the web itself will not simply become "self-aware", but when an AI-capable system (or thing) gets a hold of such a pool of information, it will feed itself a quantity of information that the human brain could never understand. When this day comes, it will be the last step towards limitless innovation and human/machinist achievement. We are so close!
It seems like most of you are defining the web to be separate from the people using it. Doesn't the Machine include the hamsters making the wheels spin?
If it did, how would anyone know? Could it be already out there, desperately trying to reach out to us but blocked out by our spam filters? Or would it even want to? When was the last time you tried to have a meaningful conversation with mitochondria?
I think it will... and I think it will blind-side the general populace when we get there.
I actually think our hardware is already sufficient for this (obviously not one computer, but a large network of them). I think it's mostly a software problem at this point. I think a new generation of database products in the works (with more flexible, dynamic objects) and the recent (last decade) appreciation for statistical pattern recognition will take us a long way. I think we are not much more than inefficient, redundant, statistical processing systems.
25 comments
[ 2.7 ms ] story [ 62.2 ms ] threadFrom the verb used in the question, emerge, it looks to me that this "superintelligence" is reached without human intervention. But as spontaneous generation does not exist, we need a system capable of evolve, without external help. And as far as I know, nowadays, neither the Internet (network) nor any of each nodes can evolve without human help.
So, until somebody invents an evolving device and connects it to the Internet, my answer is no.
It took an enormous amount of cumulative optimization pressure to turn self-replicating RNA strings into humans. The Web doesn't have anything remotely on that order, and no, the "reproduction" of corporations and startups doesn't even begin to approach the quantitative selection pressures involved.
See: http://www.overcomingbias.com/2007/11/no-evolution-fo.html
I just don't want to dismiss him - besides it seems arrogant that we should forever be the best evolution can offer.
Sharks have outlived us by a million years, so in evolutionary terms they are definitely more successful than us.
Obviously, the key term here is "awareness". We "know" we are aware because . . . uh, because we think therefore we are, I suppose. Do our cells know about our awareness beyond their genetic programming to do what they do to contribute to it? Do we go out of our way to make our cells 'aware' of our super-awareness?
My point is that way down the road when there are exponentially more sensors, programs, humans, and other 'resources' connected to the web, how would we even know that it had become . . . something more? It's probably never, ever going to take over some terminal and say "Hello there, Dave". And we may never know, because just as you can't find consciousness in the brain, you won't be able to find it in the Net. It would be the ghost in the machine, that had little interest in its component pieces.
For those, myself included, who say this or that will 'never' happen, I'd like to add a quote from Freemon Dyson, talking about a multi-disciplinary conference where the future of human evolution was discussed: "When the physicists started talking about the really long term, [the others'] eyes glazed over."
You can tell someone doesn't know these things when they never bring it up, despite it being critical to their claims.
One way this is relevant is that people talk a lot about raw computing power. As if that was the bottleneck in humans. But the more critical issue, by far, is irrationalities. And they don't address whether AIs will have those, or not. Nor do they address how to "parent" an AI.
This particular article doesn't even mention traditions or memes.
How do you know that you know that? ;-)
the whole knowledge is justified, true belief thing is such a mess. they've known induction doesn't work for ages, but haven't replaced it. good answers to these things, on the other hand, are largely ignored. (hint: Popper)
I actually think our hardware is already sufficient for this (obviously not one computer, but a large network of them). I think it's mostly a software problem at this point. I think a new generation of database products in the works (with more flexible, dynamic objects) and the recent (last decade) appreciation for statistical pattern recognition will take us a long way. I think we are not much more than inefficient, redundant, statistical processing systems.