Thread: How smart is Intelegent?

I was wondering what you people think of artificial intelligence and what it takes to be living organism.

I noticed how people are advancing artificial intelligence software. You can even have small conversations with them. Even though most of this is all scripted dialogue. But isn't what we say and do scripted in a small way. When someone says hello your first responce is to respond in some way. And you do this because you know you should respond.

This is then scripted in someway. It may be millions of time more complex then how its done in your brain but its the same idea.

So with this how far do you think they will need to be, to be living objects that can think for its self and react on its own and be living. I also think that if these mechanical organisms could reproduce they could be even more living.

To think, it must learn. Like learn to understand language, and speak. Then we can't have intelligence "out of the box". We'd have to engineer a newborn mind...

I agree with pong about the fact that true intelligence is the ability to learn from the environment.

there is an interesting video on T.E.D. talks link below that is interesting.

http://www.ted.com/index.php/talks/h...re_robots.html

although i must say pong i disagree about having true A.I. from a box. The first Artificial life form could be created and then its learned experience could then be transfered to a replicated body and therefore instant A.I.

Originally Posted by SlugMan

I was wondering what you people think of artificial intelligence and what it takes to be living organism.

I noticed how people are advancing artificial intelligence software. You can even have small conversations with them...

There's limitations as to what we can do with AI, I think partly because there's limitations on computational theory and solvability. Some problems, which humans can figure out intuitively, are impossible to make algorithms for and are impossible for computer to mimic.

Intuition, imo, is like a super-logic. There are some conclusions, or solutions that a person just happens to come to without any logical thought behind it, so you could formulate it into a logical problem or design a program which understands it.

There are two exclusive approaches to AI. One is faking it. For example if you mention the keyword "dream" to a chatbot it responds with scripted paraphrase of Freud. And that may be convincing. The other approach is genuinely adaptive, most simply with a neural network. This simulates a blank slate of gray matter, that fills up with patterns from experience and restructures on some neural reward scheme. These seem to do intuition pretty well and in my opinion are the real way to go. But the trouble is they start ignorant, like babies, and don't always learn what we wish them to.

We tried to teach a neural net to spot Soviet tanks in satellite images. So we showed the net all these example pictures: this is a tank, this is not a tank, this is a tank... Finally it seemed to catch on so we tested it with pictures it had never seen. Is this a tank? Wrong. Turns out all our example "tank" images had been taken on clear days, and many "not tank" images showed random crummy weather. The AI learned fair weather is a tank.

But isn't that just how real intelligence works?