MrMind comments on AIFoom Debate - conclusion? - Less Wrong Discussion
You are viewing a comment permalink. View the original post to see all comments and the full post content.
You are viewing a comment permalink. View the original post to see all comments and the full post content.
Comments (48)
Don't waste your time with the AI foom stuff. The commonly held opinion by experts that work with actual AI technology is that any sort of hard takeoffs will have timelines that are not small on human scales, at least with technology that exists today.
You need only do a few back of the envelope calculations using various AI techniques and AGI architectures to see that learning cycles are measured in hours or days, not milliseconds.
I would just advise caution: we do not have examples of AI improving their own architecture, so it is dangerous to assume one way or another.
We most certainly do. The field of AGI research is a thing.
We do? I wasn't aware of them... do you know of any example?
SOAR, ACT-R, MIcrobePsi, OpenCog, LIDA, Numenta, NARS, ... you might find this list helpful:
http://linas.org/agi.html