I was wondering if Eliezer could post some details on his current progress towards the problem of FAI? Specifically details as to where he is in the process of designing and building FAI. Also maybe some detailed technical work on TDT would be cool.
This email by Eliezer from 2006 addresses your question about FAI. I’m extremely skeptical that he has accomplished or will accomplish anything at all in that direction, but if he does, we shouldn’t expect the intermediate results to be openly published, because half of a friendly AI is a complete unfriendly AI.
I was wondering if Eliezer could post some details on his current progress towards the problem of FAI? Specifically details as to where he is in the process of designing and building FAI. Also maybe some detailed technical work on TDT would be cool.
This email by Eliezer from 2006 addresses your question about FAI. I’m extremely skeptical that he has accomplished or will accomplish anything at all in that direction, but if he does, we shouldn’t expect the intermediate results to be openly published, because half of a friendly AI is a complete unfriendly AI.