more of should we fear AI thoughts
more of should we fear AI thoughts
https://aeon.co/essays/true-ai-is-both-logically-possible-and-utterly-implausible
https://aeon.co/essays/true-ai-is-both-logically-possible-and-utterly-implausible
https://aeon.co/essays/true-ai-is-both-logically-possible-and-utterly-implausible
https://aeon.co/essays/true-ai-is-both-logically-possible-and-utterly-implausible
I'm going to suggest that the author hints at a definition of true AI that is probably unachievable (a machine that is never wrong and always right) and is in this sense arguing a straw-man.
ReplyDeleteWe have intelligent applications that tend to do only one thing but they do it really well. I believe that we will soon see machines that can perform multiple complex tasks and do each of them really well, but maybe not quite as well as their single-minded brethren. Extrapolate this and we get to a machine that can do just about anything well enough while being unremarkable at all of them; Sort of a "Jack of all trades, master of none" AI.
This is the one that will be indistinguishable from human intelligence. He's just average. He makes mistakes like you and I, and he's not an expert on anything, much less an expert on everything. As such, he might not be any more useful to humanity than any of us. But is he sentient? Maybe, and maybe not. The question is irrelevant though, if he is in fact indistinguishable from human intelligence. If you can't tell the difference, you'd be required to treat it as sentient.