In which a commenter "ponta" offers following (citing a Nature paper):
"AlphaGo played Fan Hui using 1202 CPUs and 176 GPUs.
It's strenght was assessed to be 3140 ELO on the scale used by www.goratings.org/ (BTW, this is different from the European ELO system)
That would put it at #279 in the world. Fan Hui is number 633 at 2916. AlphaGo's next opponent will be Lee Sedol, #5 at 3515."
Plugging 3140 and 3515 into the ELO win probability calculator with weighting factor "50" (not positive that is correct weighting factor so perhaps this estimate could be improved upon with a better understanding of this tool)
http://bzstats.strayer.de/bzinfo/elo/?lang=en
The calculator gives AI a win probability of 10% for this matchup.
Noting that the AI is likely still under development and it is more likely that the AI will improve materially prior to the match than the human doing so, I will generously add another 5% to win probability (15%).
Finally noting that the GJ crowd is currently at 55% and the crowd is likely smarter than myself so I'll bump it up to 25% chance of AI win.
In Gary Markus discussion of the hybrid neural net / decision tree approach being used:
https://backchannel.com/has-deepmind-really-passed-go-adc85e256bec#.bz6s0q2c7
He points to an interesting discussion taking place in the Hacker News boards:
https://news.ycombinator.com/item?id=10981679
In which a commenter "ponta" offers following (citing a Nature paper):
"AlphaGo played Fan Hui using 1202 CPUs and 176 GPUs.
It's strenght was assessed to be 3140 ELO on the scale used by www.goratings.org/ (BTW, this is different from the European ELO system)
That would put it at #279 in the world. Fan Hui is number 633 at 2916. AlphaGo's next opponent will be Lee Sedol, #5 at 3515."
Plugging 3140 and 3515 into the ELO win probability calculator with weighting factor "50" (not positive that is correct weighting factor so perhaps this estimate could be improved upon with a better understanding of this tool)
http://bzstats.strayer.de/bzinfo/elo/?lang=en
The calculator gives AI a win probability of 10% for this matchup.
Noting that the AI is likely still under development and it is more likely that the AI will improve materially prior to the match than the human doing so, I will generously add another 5% to win probability (15%).
Finally noting that the GJ crowd is currently at 55% and the crowd is likely smarter than myself so I'll bump it up to 25% chance of AI win.
Very interesting use of ELO!