http://idlewords.com/talks/superintelligence.htm
Personally I don't think strong AI is possible in a classical computer.
Super Intelligence
-
- Tenth Dan Procrastinator
- Posts: 4891
- Joined: Fri Jul 18, 2003 3:09 am
- Location: San Jose, CA
Re: Super Intelligence
Are you saying that cognition relies on something that Turing complete computation doesn't provide? Or are you saying that semiconductors won't scale to a point that will intersect with what would be necessary for cognition? Or the AI would need access to lots of robots/manufacturing in order to perform the basic science that would be necessary for evolutionary leaps. Or something else is a limiting factor?
It would have to be an AI trained with expertise in many fields before it could start to become a strong AI for sure... If we silo the information provided to each AI instance, and disallow them to communicate directly, we could probably prevent a strong AI. Without a wide range of experiences, I don't think a human-level AI would have the "imagination" to become a strong AI.
It would have to be an AI trained with expertise in many fields before it could start to become a strong AI for sure... If we silo the information provided to each AI instance, and disallow them to communicate directly, we could probably prevent a strong AI. Without a wide range of experiences, I don't think a human-level AI would have the "imagination" to become a strong AI.
-
- Grand Pooh-Bah
- Posts: 6722
- Joined: Tue Sep 19, 2006 8:45 pm
- Location: Portland, OR
- Contact:
Re: Super Intelligence
This, yes.quantus wrote:Are you saying that cognition relies on something that Turing complete computation doesn't provide?
All other objections can be resolved with sufficient time. A sort of "Long Now" project if I were wrong with the entity operating at 1/1000th time would be something.