I will say again that the work itself is impressive, but mischaracterized, and that a better title would have been "manipulating a Rubik's cube using reinforcement learning" or "progress in manipulation with dextrous robotic hands" or similar lines.
-
-
Show this thread
-
New conversation -
-
-
But dexterity is the hard/new part. When humans say they can solve a Rubik's cube, they usually mean they can mechanically follow the steps to a solution. The only non-trivial parts of making a robot do that are the dexterity and the vision, and the vision part is already solved.
-
actually they had to instrument they cube and the hand because vision was not adequately solved; that was on my list of points.
- 3 more replies
New conversation -
-
-
This criticism is a mix of misleading and incorrect comments, though I agree the use of Kociemba's algorithm should have been stated more prominently in less technical terms.https://www.reddit.com/r/MachineLearning/comments/dkd4vz/d_gary_marcus_tweet_on_openai_still_has_not/f4d8lmq/ …
-
i have replied to you on reddit.
- 3 more replies
New conversation -
-
-
Ah....so this was not an off-the-shelf cube, but rather one that had been modified to contain electronic elements from which cube orientation could be detected?
-
Yep, a lot of the work was instrumenting the cube:pic.twitter.com/6ko5rRblg7
- 3 more replies
New conversation -
-
-
What's your take on OpenAI's GPT-2 that's been covered by the New Yorker? https://www.newyorker.com/magazine/2019/10/14/can-a-machine-learn-to-write-for-the-new-yorker … .
-
my (somewhat implicit) take on gpt-2 was here, a bit earlier:https://www.wired.com/story/adaptation-if-computers-are-so-smart-how-come-they-cant-read/ …
- 1 more reply
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.