AI and Superintelligence - some links

[andrewducker]

A friend linked to Wait But Why's really readable link on AI and Superintelligence.

Another friend countered with another link, which points out a few errors in the article.

Charlie Stross is sceptical that it's coming within his lifetime.

Personally, I felt like 2067 was a reasonable date about 13 years ago. I've not seen any good reason to revise that since :-)

(I may or may not make it that far, depending on life-extension technology and the amazing advances in healthcare over the next few decades.)

Apparently the best book on the subject is Nick Bostrom's "Superintelligence", which I haven't read, but when my brain starts working again...

Comments

[birguslatro]

I was a bit interested in neural networks in the 80s, so I'm not surprised they're at the core of the current AI explosion. It seemed obvious to me AI wouldn't be real if we understood how it worked, and I doubted large neural networks would be understandable. This seems to have proved be the case.

AI language translation may not quite match a good human translator yet - as long as speed isn't factored into the test. "You have ten seconds to translate this novel - starting from now!"

And that's the rub. A speedy cheap robot is going to win against expensive, slow humans. And in domains where trained humans are right 80% of the time but the robots are right 90% of the time, the robots are also going to win (the human's jobs.)

Wait but Why's fun, isn't it? I'd planned to say something about their Elon Musk three-parter, but never found the time! :) It begins here...

http://waitbutwhy.com/2015/05/elon-musk-the-worlds-raddest-man.html