Before starting writing about Why AI is dead, I’d like to invite a few comments on “Is AI really dead?”. Today, in my view, I don’t see any breakthrough research going on. I no more sumble across programs which can create poetry, can chat effectively or can demonstrate novel applications. So I consider AI dead today.
The reason AI is dead because researches, in the past, have taken a wrong decision that the goal of AI is to *imitate* humans. By imitation I mean programming a computer to behave like a human. For example, the turing test involves the testing of a computer to be able to communicate like humans.
This is nuts!! Do we design air-planes to fly like birds. No! Then why do we aim the machines to behave like us. Intelligence itself is an entity which can be approached via n number of paths. Then, why do we aim to approach it through the most difficult of paths which is human-like intelligence. Evolution has taken millions of years to evolve the intelligence. We aim to do achieve the intelligence via same path in relatively lesser period of time.
I know there is a contradiction of terms. First, I say AI is dead because there are no more development into human-like intelligence. Secondly, I refuse to say that human-like intelligence is AI. See, AI itself is a very broad concept. It is more that how you pursue AI rather than what it is. Human-like intelligence is a form of AI, not vice versa.
So, this death of human-like intelligence is, indeed, very welcome. It will now let a few budding enthusiasts (myself included) to experiment alternative routes to AI. By alternative routes, I mean letting computers themselves decide what intelligence (according to them) really is. Then use that intelligence for really cool applications.
So, at last, I would like to conclude this post by saying that we are really lucky that finally the attempts for programming human-like intelligence are dead. It had to be someday, but luckily that day is today.
Good luck!!
I don’t thinks that AI is really dead, it only slowed. The first goal of the AI in the past 19(60~70) was only to get pattern matching as we call it now. I read few books of hofstdater (that i suggest you if you haven’t read it) that suppose and AI based on analogies. It’s a solid point for a good AI. Btw i think that before invent a new intelligence we must fist understand us.
1. Yes, we did design airplanes like birds, just simplified.
2. The paths evolution takes are the EASIEST ones, not the hardest – these are the paths with the smallest gradients.
3. Imitation is the easiest way to go – if you can’t make a functioning human-like brain, you can’t even *define* intelligence, much less *re-define* it.
Otherwise, yeah, AI is pretty much dead, at least until dedicated hardware is available. Simulating massively parallel systems on an 8-threaded CPU is madness. The good news is this hardware is already being developed.