I felt a lot more hopeful about the future when SIAI was doing CFAI than I do now that MIRI is doing agent foundations.
because your estimate of problem difficulty has gone up or because your estimate of strategy quality has gone down?
-
-
They're doing more formal work, which is one kind of quality, but it's directed more at recursive self-improvement (which is bad ceteris paribus) than at value learning. Corrigibility stuff was interesting, but comparatively small.
-
Also, providing other groups with advice on how to not mess up AI instead of getting AI right themselves is worrisome because people in industry don't care, absent a few high profile endorsements
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.