But a few hours is 13,000 yearsin simulation time
-
-
Kiitos. Käytämme tätä aikajanasi parantamiseen. KumoaKumoa
-
-
-
Not sure how the task difficulty compares. but we were able to do Baoding balls on the same robotic hand within 4 hours.https://twitter.com/svlevine/status/1177434279795515392 …
Kiitos. Käytämme tätä aikajanasi parantamiseen. KumoaKumoa
-
-
-
Just to be clear, they didn't solve Rubik's cubes. They manipulated Rubik's cubes to appropriate states given a solution sequence. That this is hard to tell from a quick read has a lot to do with deceptive marketing.
-
That's not to take away from the accomplishment, this is really exciting.
Keskustelun loppu
Uusi keskustelu -
-
-
Actually, in their "defense" it took them a few months
-
But yeah, 13000 years simulation time certainly. The answer is adequate learning priors
Keskustelun loppu
Uusi keskustelu -
-
-
Also, they didn't learn to solve it completely even. It's like 20% success rate for hard perturbations of the Rubik's cube and 60% for normal ones. After months of training!
Kiitos. Käytämme tätä aikajanasi parantamiseen. KumoaKumoa
-
-
-
Give it to a kid?
Kiitos. Käytämme tätä aikajanasi parantamiseen. KumoaKumoa
-
-
-
Brute force is energy expensive. If and when the energy used to solve a problem is of the order of magnitude of that spent by a human we could talk about intelligence.
Kiitos. Käytämme tätä aikajanasi parantamiseen. KumoaKumoa
-
-
-
My guess would be that evolution took a couple of million years to get to five fingers with an opposable thumb and the respective control circuitry. So given the advantage of constraint mechanics, 13k years is quite good, I’d assume.
-
Yes, if you forget how general the evolution approach is and how narrow the RL one.
Keskustelun loppu
Uusi keskustelu -
Lataaminen näyttää kestävän hetken.
Twitter saattaa olla ruuhkautunut tai ongelma on muuten hetkellinen. Yritä uudelleen tai käy Twitterin tilasivulla saadaksesi lisätietoja.