Wednesday, May 14, 2008

Don't Swat the A.I.

Not that I am in any way qualified to speak on the subject, but I saw this very enlightening article today and it gave me reason to think about mankind’s efforts to create artificial intelligence.

What an overwhelming proposition. The human brain has been described as one of the most complicated organizations of matter in universe. How much we understand about the brain is minuscule compared to what we still find completely baffling. In fact, we have every reason to expect that there are many things about cognition that we don't even know that we don't know, such is our benighted condition. What a daunting task, then, to create artificial intelligence based on the mind of humans.

It’s tempting to make the leap based on the fact that we have created super-computers that are, in many ways, superior to the human mind. But in many fundamental ways, they haven’t even scratched the surface of human intelligence, and might never if we simply try to throw more and more computational power at the problem.

So what if we just go at it like you would in solving any other problem? We didn’t start building aircraft with 747’s or B2's. The Wright brothers put together a few pieces of wood and cloth and flew it a short distance. What if we start by creating artificial intelligence at the level of a cockroach, instead of at the level of Einstein? I’m not sure that’s the purpose of the research talked about in this article. But it could certainly lead to that. Today the cockroach, tomorrow the lizard?

8 comments:

The Exterminator said...

In the field of very artificial intelligence, you might like to consider this.

John Evo said...

Because of shit like that - build an insect!

the chaplain said...

Evo's right: it's better to start with something simpler than the complexity of the human mind.

John Evo said...

Yeah. If they want to start with a brain, they should use mine. :)

Lifeguard said...

I think it goes beyond brain and computational powers. It's the fact that our brains are just one part of a gigantic system that gathers information in the form of touch, taste, smell, vision, and sound and integrates it all. Dennett's "Consciousness Explained" is a dizzying tour into this territory. Interesting stuff.

PhillyChief said...

I believe most robot endeavors have been focused on mimicking insects for both their simplicity as well as their stability in traversing complicated terrain, then along came Asimo to blow everyone's minds.

There are also two different routes looked at for creating AI, hardware or software AI. Probably a third as well, a hybrid of the two. It'll happen eventually, and some asshat theist will protest the "soulless abomination" or something. Some things are inevitable.

Ordinary Girl said...

Why an insect? Why not bacteria or some other single cell life-form if you want to get simple.

The fact is that there are many different kinds of AI. It depends on the purpose of the AI. Not every researcher is trying to re-create the human brain. Some AI are used to mimic evolution, some for prediction, others for navigation. Neural networks is actually a pretty rare area for programmers working in AI as it hasn't yet become practical for use.

I also don't think the human brain is the most complex thing in the universe. Just because we don't understand it yet doesn't mean it's of the highest order. Maybe it's more fitting to state that it's the most complex thing we're aware of at the moment? I still don't know if that's true though.

grumpylion said...

Ordinary Girl -

The Enterprise is the most complex thing in the known universe. And neural nets aren't so rare anymore - hell, they play chess and backgammon and beat the pants off human experts.

Evo -

The cockroach has already been built, but it's not all that smart. Does have a name - George Bush.