MIRI publishes a lot of research on ‘neat’ systems like first order logic reasoners, and not on ‘scruffy’ systems like neural networks. I heard Eliezer Yudkowsky allude to the idea that this is for convenience or budgetary reasons, and that they will do more research on neural networks (etc) in the future.
Does anyone have any more information about what MIRI thinks and intends to research about ‘scruffy’ AI systems?
MIRI publishes a lot of research on ‘neat’ systems like first order logic reasoners, and not on ‘scruffy’ systems like neural networks. I heard Eliezer Yudkowsky allude to the idea that this is for convenience or budgetary reasons, and that they will do more research on neural networks (etc) in the future.
Does anyone have any more information about what MIRI thinks and intends to research about ‘scruffy’ AI systems?
Basically, this:
https://intelligence.org/2016/07/27/alignment-machine-learning/
It’s now MIRI’s official 2nd agenda, with the previous agenda going under the name “agent foundations”.
Okay, thanks.