(I decided to reply to this comment with a post.)
Been Lurkin’: “What exactly (i.e. which books, websites, etc.) have you been reading on these topics? I used to be somewhat into this kind of thing like five years ago – I read Kurzweil’s book and a lot of the transhumanist stuff on the web – but it doesn’t seem like much has changed conceptually since then, although processors are faster and all that.”
I used to follow AI closely. I did work in image recognition. I evaluated expert systems applied to automatic configuration and system tuning. I studied the blackboard planning systems used for navigation systems and neural net systems applied to feature detection in photos. Most methods showed early promise but failed to scale when applied to tough problems or were too slow when interacting with real world events or were too brittle when handling unexpected events. At that time a 10 MIP processor with 256 Meg of RAM was considered a powerful AI platform.
Things have changed.
Hans Moravec: “When will computer hardware match the human brain?”
“matching overall human behavior will take about 100 million MIPS of computer power”
Moravec estimated that such computing power would be commonly available for AI use by 2020. I believe we will reach that point in the next few years.
The Cell processor that will be used in game consoles has:
· Peak performance (single precision): > 256 GFlops
· Peak performance (double precision): >26 GFlops
It is designed to support large multiprocessor architectures. I expect such cheap, powerful processors to significantly enhance AI application performance.
Been Lurkin’, you asked for web sites or books on this topic. I’m just starting my search but I found these sites interesting. I don’t yet have a good source for new articles on this topic.
Posted by fly at 06:48 AM