Hi guys....

I think we can build AI which can drive a car manage a factory or control the whole defence system....

but we can not build an AI which is as flexible and able to learn as we do i.e universal Ai(which is shown in movies)

cuz i think we can not code an Ai,we have to evolve them as we did....

what do u guys think...

vj

I see where your coming from.

But I'm not sure, we've concluded that everything is mathmatical right? and pattern theories have helped support that.

So why can't universal AI, be also re-constructed, using these patterns and equations that we can find out.

I agree with you. Despite teh enormous power of formalization, universal AI, just like our own minds, is not a formal system. Therefore, it cannot be reproduced via patterns or equations, that are formal systems. However, artificial evolution may be the only way of approaching to a real, semi-functional universal AI.

OLA man the thing is u can not manually recognize those patterns and equation and add them to the AI,it has to generate them automatically,for that evolution is the best way....

Thst's right, Vijay. Evolution leads to paths that are mostly almost-impossible to achieve trough mathemtahical formalization.

Our (AI researchers) main problem is time. Evolution in nature has an advantage of millions of years. Compared to that, we're only just beginning (with stunning results, I might admit).

As for the universal AI debate, I'm not sure if that's even what we should aim for. What's the advantages of having an entity that thinks and acts like a human being? We've got tons of that already.

Yeah Rudi, we have but they don't interface with a machine. The goal is to give the machine the proprieties that they lack and humans have, not to make a machine that thinks like a human.

So we either connect a microprocessor to a human brain, or put a "brain" on a machine!

AI would have to evolve a lot of human judgment is based on personal experiences. A machine that can beat a human or at least think like one can be made. If i go back a few years one of the greatest chess players ever was beaten by a computer program called Deep Blue.

hey buddy, the whole point of AI is to learn. else it wouldnt be called AI, but computer programming.

I work in processing human language ("natural" language).
The discussion of whether humans have all their language directly from the environment or are "wired" to assimilate certain communication patterns predates the computational linguistics field.
In a sense, you want to learn (evolve) from data as much as you can but you want to model (program) this innate abilities as much as you can. Starting completely from scratch requires tons and tons of data (take a look at Hidden Markov Models, for some interesting example).