In 2018, OpenAI discovered that the amount of compute used for AI training doubles every 3.5 months. By comparison, the number of transistors per square inch on integrated circuits only doubles every 18 months, a.k.a Moore’s Law.
The best human has a GO rating of 3695 while Alpha GO Zero has an ELO rating of 5185. Human players were beaten 100 to 0 and AI GO systems can only crush the prior best AI to show improvement.
The hardware cost for a single AlphaGo Zero system, including custom components, is estimated to be $25 million.
There has been an open-source implementation of Alpha Go Zero. ELF OpenGo is the first open-source Go AI to convincingly demonstrate superhuman performance with a perfect (20:0) record against global top professionals.
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
Is AlphaGoZero the AI that purely learns by playing against itself? If so then this is fair scary.
No, ratings only take into account the result of games (win, loss, draw) and the strength of the two players. Sounds circular, but it isn’t. Elo is relative. You can only say how much stronger one person is from another. You could give 5000 more Elo to everyone and it would make no difference to the system. The level is not an absolute measure of strength.
When you look at that table it gives an expected % result. If you are 200 Elo stronger than an opponent, and you play many games with them, you would be expected to get roughly 3/4 of the game points (1 point for win, 1/2 point for draw, 0 points for loss). Don’t confuse these points for the others though. You generally get a lot more than 1 Elo point for a win. If you draw someone stronger you get Elo points also.
Your rating is adjusted after every tournament/match based on how many game points you got and the level of the opposition you faced. Here is a calculator: https://www.3dkingdoms.com/chess/elo.htm
There is also a K value that allows more movement when you have not played many games yet (provisionally rated) or are a kid and thus are expected in increase in strength more rapidly than the system would normally be able keep up with.
Not sure what this table represents. Can it be said that a human “Go rating” is simply the score a human can achieve in a Go game?
https://www.remi-coulom.fr/WHR/WHR.pdf
But easier to comprehend is this table: http://pradu.us/old/Nov27_2008/Buzz/elotable.html
I can’t figure this “Go rating” for humans.