Landmark paper "Logical Inductors" shows how to coherently reason about your own reasoning. Resolves several paradoxes satisfactorily. These ideas will be useful for higher-order approximate inference. https://arxiv.org/abs/1609.03543 @critch_ai @So8res
i was not impressed by the paper bc of merely pointwise convergence / impossibility of finite-time guarantees (wrote about this more at http://nostalgebraist.tumblr.com/post/160975105374/on-miris-logical-induction-paper … ). are my concerns misplaced / did the paper do something cool that i missed?
-
-
I agree with your points, but I consider it a big contribution to give an algorithm even with only asymptotically nice properties. Like Solomonoff induction or AIXI, It advances the conversation. Many still object to putting probabilities on theorems or computable quantities.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
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.