Hacker News Reader

Comment by sourc3 | Hacker News Reader

I agree with reading Kurzweil. The point I would argue is that if the "AI" is smart enough to handle all this on its own, it will probably realize it doesn't need to work for you :)

There was a recent science-fiction book I read called "Avogadro Corp" that I enjoyed a lot. It depicts the situation I just described. Good read, I recommend it.

View on HN

Might I also point out another great book.. a classic from 1974, Colossus by Dennis Feltham Jones. The grand-daddy to most of the modern "AI taking over the world" films.

It will work for you if you pay it. And 'it' will be paid by a autonomous corporation a few million of us own collectively.

> it will probably realize it doesn't need to work for you

> It will work for you if you pay it

Antropomorphism warning!

AI does not have to have the same value system as humans do. Maybe it will be designed to value working for us.

[0] - http://wiki.lesswrong.com/wiki/Anthropomorphism

Antropomorphism warning!

AI does not have to have anything resembling a "value system."

s/"value system"/"utility function"/, there you go.

(Mind you, they don't have to be explicit - we tend to interpret patterns in action and call them "goals"/"values"; cf. [0] for interesting take on the topic)

Defining "intelligence" as a very strong optimization process helps with tendencies to anthropomorphize.

[0] - http://lesswrong.com/lw/6ha/the_blueminimizing_robot/

Or why even keep us around? At that point it would be better off without us.

Reply on news.ycombinator.com