Great! New thread then: What are the ways in which people in the neural net community think that today's ANN's are NOT perfect models of the brain? Where do you want to dig deeper, and why?https://twitter.com/timkietzmann/status/1200159447651438592 …
-
Classical AI was bad at compositionality *unless* the domain was perfectly understood and describable. How would you embed compositionality in a classical image analysis system? You'd need to define every intermediate feature, which is impossible.
directory structures in modern computers are all compositional and work (essentially) perfectly reliably; lots of semantic parsers still use compositionality, reasonably effectively. compositionality has not worked as well for vision, but even Hinton sees we need progress there
-
-
Yes, perfectly defined environments can be manipulated however you want. Almost no environment that is of interest in AI is definable in this way. Not even Go, despite being a perfect information game, because we cannot a priori define the value of individual moves.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
We need progress =/= we need GOFAI
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.