efyu_lemonardo
May I have a cookie?
Remember AlphaGo? The A.I that defeated Lee Sedol in 4 out of 5 games of Go last year?
A feat that was considered far out of reach for a.i. only a couple years prior?
AlphaGo Zero, the new bot on the block at Deep Mind, creators of AlphaGo, was able to defeat its predecessor only 3 days after going online. And that's not even the most amazing part!
Unlike AlphaGo, which learned from analyzing thousands upon thousands of human matches, Zero is entirely self taught.
As in, it was turned on, given nothing but the rules of the game, and within 3 days had progressed to such a level that it destroyed AlphaGo Lee (last year's version that beat the human player) in 100 matches out of 100. That's right, it didn't lose even once! After 3 days of training. With no previous knowledge of the game whatsoever.
Be afraid. Be very afraid.
Image from here:
http://www.sciencealert.com/new-sel...en-civilisation-inventing-its-own-mathematics
Further links:
Google Supercomputer learns 3000 years of human knowledge in 40 days.
A feat that was considered far out of reach for a.i. only a couple years prior?
AlphaGo Zero, the new bot on the block at Deep Mind, creators of AlphaGo, was able to defeat its predecessor only 3 days after going online. And that's not even the most amazing part!
Unlike AlphaGo, which learned from analyzing thousands upon thousands of human matches, Zero is entirely self taught.
As in, it was turned on, given nothing but the rules of the game, and within 3 days had progressed to such a level that it destroyed AlphaGo Lee (last year's version that beat the human player) in 100 matches out of 100. That's right, it didn't lose even once! After 3 days of training. With no previous knowledge of the game whatsoever.
Be afraid. Be very afraid.
Image from here:
http://www.sciencealert.com/new-sel...en-civilisation-inventing-its-own-mathematics
Further links:
Google Supercomputer learns 3000 years of human knowledge in 40 days.