Yes, manipulating is a hard problem and that’s impressive by itself. Science question for friends: can Rubik’s cube be solved by fixed feed forward networks? https://twitter.com/halhod/status/1184395411546505217 …
-
This Tweet is unavailable.Show this thread
-
Replying to @deliprao
I *think*
@halhod was doing some A+ nudge-wink trolling of@GaryMarcus with the CNN comment. But yes, NNs are not at all suited to many-step-until-reward puzzles like this without very strong hints (like carefully engineered reward functions or watching lots of successful runs).1 reply 0 retweets 2 likes -
Replying to @_brohrer_ @deliprao and
And I absolutely agree with the rest of Hal’s assessment. Handling things is hard, and sim-to-real progress is the most important contribution of this paper.
1 reply 0 retweets 2 likes -
um, wasn’t that close to my assessment, too? i wrote “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.”
0 replies 0 retweets 5 likes
because i think it is completely misleading and advanced two agendas that are bad for AI: rampant blank-slatism that implied a form of learning that did not take place, and a neural network-centrism that implied symbol-manipulation was absent when it was essential.
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.