somewhat misleading but wildly retweeted characterization of new @OpenAI result. the system is (very impressively) trained to do the perception and manual dexterity parts, but cube solving algorithm itself is innate, and symbolic, not acquired via training.https://twitter.com/OpenAI/status/1184135128869527552 …
-
-
Just gets more and more underwhelming as you dig in. Somehow these kids with lego and an iphone managed to get the state of the cube purely from vision...
-
Boston Dynamics would just put Aruco markers on that cube LOL. I should make one just for shits and giggles. Computer vision friendly Rubics cube... LOL
End of conversation
New conversation -
-
-
The authors suggest they could overcome the stickers on the cube / Giiker sensors with some kind of memory-augmented vision model such as an attention augmented CNN or CNN features passed into LSTM
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Wow guys, thanks for digging into this. Unfortunately the ‘OMG’, ‘Skynet’, and ‘Deep state will use this’-people will never read this. The trough of disillusionment will be deep and equipped with concrete flooring with spikes.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
AFAIK, Bluetooth sensing was only for debugging. I think you guys under evaluate the effort here.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Yes. But how do you propose to solve this problem using a hybrid solution? You think that adding GOFAI to DL will do any good?
- 2 more replies
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.