Is AI Dead? | Hacker News Reader

Is AI Dead?

5 pointsjoe_the_user7 comments

I've been reading "The Singularity Is Near" by Ray Kurzweil and scanning his sources. One thing that struck me was that things like Neural Nets hadn't made very much progress since I was an MA grad student 17 years ago. Back then, Kohonen networks were the rage and as far as I can tell, there isn't anything terribly more interesting today.

Now, I actually wish, I hope, that this isn't so. Perhaps I've missed something but it seems like rather than advance any prospects of honestly intelligent machines, both academic and industry research now, in 2009, concentrates on extending what already exists - better languages, incrementally better algorithms, better simulations, better databases and better hardware.

To my mind, this doesn't stand in the way of something like "the singularity" but it means that if such a thing occurs, it will be with something like the Blue Brain Project - simulating the brain without understanding it. This may indeed work but, strangely enough, it means experimenting with processes which we really don't understand and thus makes something ... dum dum da ... Skynet actually seem disturbingly plausible...

But that's a side issue. Big topic - "Is AI, Artificial Intelligence, Dead?" Yea or Nea?

I think the whole field of AI research lost credibility back when ANNs couldn't do XOR and therefore where not turing-complete. That is now solved and AI research is slowly gaining momentum again.

Also I think Kurzweil (and others like him) are a huge distraction. They mystify AI as something powerful, unpredictable and maybe harmful. But it's just a way of applying stochastics to computational problems. In other words: using guessing to solve complex problems.

Thats all just my personal point of view of course.

I don't think that the idea that once computers reach a certain level, they can be used to build more powerful computers should be discounted. If you define "the singularity" in this fashion, you don't need any concept of consciousness but you have to admit that there will be issues.

Similarly, once there are robots that, say, guided by a few people, can take on many, many people in a conventional or irregular army, on a real battle field, then you'll have a situation which will influence real social and political decisions. You could call that "the singularity" too.

And since these and other issues are approach, I Kurzweil does us a service point it out even if many of his formulations are indeed crude and simplistic.

Hmm,

Could you give some references on this?

Not at all. The research continues, but instead the the application focus has shifted.

What you are describing is Artificial Consciousness, this is great and all, but the practical applications aren't many compared to regular AI related research. Things like optimizing various designs, automated <insert some activity>, natural language processing, image recognition, etc.

Mother Nature has been working on this problems for millions of years and has only come up with one solution. It's located between your ears.

There are plenty of things that mn has been working on for millions of years and we still manage to build things that do the same job - is it just a matter of time before we build something to go between the ears?

Problem is there are two AIs - strong and weak. Weak AI is just about replicating stuff humans can do - chess playing, diagnosis, vison, talking etc. Strong AI is all about building something that thinks and is consious - like a human. Weak AI never dies - strong AI goes through cycles as it hits a high and then a low. See: http://www.i-programmer.info/babbages-bag/297-artificial-int...

Reply on news.ycombinator.com