Warnings that artificial intelligence could be dangerous to humans have been coming for years (i.e. "Terminator" movies). But that's just fiction, outlandish, far out in the future if it's even possible. Only nerdy science-fiction types believe that stuff. Right?
Looks like the potential danger of AI is taken more seriously now. Respected voices in the field are sounding an alert. Though it may never evolve a conscious creativity (like humans have), it could evolve to do dangerous things that its creators never expected.
What if, for example, humans programmed it to make as many paper clips as possible . . with inadequate ethical restraints. It could go down this way: AI locks humans out of the internet to keep them from interfering, and then re-directs all earth's natural resources and industry to the manufacture of paper clips.
AI is continually trained to do deep learning, to keep learning new skills and applications. Will its abilities multiply exponentially to the point where humans can't control it?
Some experts believe that. One of them, Eliezer Yudkowsky, sounds like he's almost in panic.
(cont'd tomorrow)
No comments:
Post a Comment