The Turing Test

[jimbaloo]

So, according to this article I may actually be lying when I tell my wife that I have a bigger brain

On the subject of AI. Could an AI grow more intelligent than it was originally designed to be over time? And if it does would that intelligence be an evolution of existing traits or could it evolve into something different that humans as the creators of the machine could no longer understand?

Actually, now that I think about it, there would have to be a finite limit to how 'intelligent' any given AI could be given hardware limitations.

[fingal]

So, according to this article I may actually be lying when I tell my wife that I have a bigger brain

On the subject of AI. Could an AI grow more intelligent than it was originally designed to be over time? And if it does would that intelligence be an evolution of existing traits or could it evolve into something different that humans as the creators of the machine could no longer understand?

Actually, now that I think about it, there would have to be a finite limit to how 'intelligent' any given AI could be given hardware limitations.

Well, my understanding is that AI as it currently stands is not about a 'self aware' machine (I think therefore I am) but something skilled at performing an intellectual task usually performed by a human.

Take a checkers playing programme. The first really successful AI checkers game 'learned' over time until its creator couldn't beat it any more. In a way it seemed to be more 'intelligent'.

You touch on an ancient fear: the idea that a machine could 'take over' and possibly destroy its maker. Think of the Dr. Frankenstein story, or even HAL in 2001. I think there's something in that: true intelligence implies free will, and free will means that if something wants to destroy you, it possibly can. Then we can talk about Asimov's laws of robotics. One of these states that 'No robot can harm a human being', but he was writing fiction!

A complex subject!

[geese]

I would like to have a little say on this.

What is true AI ?
Does it mean they think like humans but thousands of times faster?
Would data of star trek be considered a true AI?

Wouldn’t a true AI have to inherit many human qualities? They must have the abilities to learn and adapt.
If this is so the potential to do good is great, but the potential to do bad is just as great.
You want AI to make decisions for us on a political level? Who then would be the masters of whom? Think of the most hated man in the world. Think of WWII. Do we really want an AI version of him?
What would happen if our AI masters wanted to do something for our own good, lets say no more fish & chips or coffee . . . . no more sex :o ok i may be going a little way out, lol

may be i'm just paranoid.

[gorgeous lucky luke]

The Question

Could an AI grow more intelligent than it was originally designed to be over time?

In Science fiction AI's are usually machavellian creatitions as in Dune, or more usually sponteaneously evolving entities, that misteriously appear somewhere out there in the nodes of the net. If the latter, I suppose there would be no limmit to how far it could evolve.

Another potentialy more interesting thread, is the migration of the human mind into a kind of virtual world.

[jimbaloo]

If I remember the plot correctly the AI's in Gibson's Neuromancer were designed systems that exceeded their original programming. The idea's presented put Artificial Intelligence in the same context of human existence in virtual worlds where most people regarded AI's as merely another tool.