Mastering the game of Go without human knowledge¶
Silver, D., Schrittwieser, Simonyan, Antonoglou, Huang, Guez, Hubert, et al. (2017). Mastering the game of Go without human knowledge. Nature, 550(7676), 354-359.
Cited by¶
1 citation across 1 artifact.
Each citation links to the sentence it supports in the citing article.
Primes¶
- Markov Decision Processes (MDPs)
- Scale and computational cost (as Silver et al. (2017a) report for AlphaGo Zero): AlphaGo's training used approximately 50M self-play games; AlphaZero's chess training used approximately 44M games; each system used substantial TPU/GPU resources (AlphaGo Lee Sedol: approximately 48 TPU v1; AlphaZero training: approximately 5,000 TPU v1 for approximately 12 hours for chess, 34 hours for shogi).
This sourceAlphaGo Zero paper; trained from random play with no human game data, reaching superhuman performance and providing the scale and Elo benchmarks cited for AlphaGo/AlphaZero.
Supported in partVerified against the publisher's abstract
Abstract reports AlphaGo Zero trained in five million self-play games, not the 50M/44M or TPU counts the claim attributes to this paper.
“Starting from random moves, it can reach superhuman level in just a couple of days of training and five million games of self-play”
- Scale and computational cost (as Silver et al. (2017a) report for AlphaGo Zero): AlphaGo's training used approximately 50M self-play games; AlphaZero's chess training used approximately 44M games; each system used substantial TPU/GPU resources (AlphaGo Lee Sedol: approximately 48 TPU v1; AlphaZero training: approximately 5,000 TPU v1 for approximately 12 hours for chess, 34 hours for shogi).
Verification¶
Does it exist? Not checked yet. This entry carries no identifier to resolve. It was extracted from the citation as written in the article, normalized, and deduplicated against the rest of the registry.
Does it back the claim? Read against the text for 1 of 1 citation: 1 supported in part. Each verdict is shown under its citation below, with what in the work backs the sentence.
Was it audited? Yes. A second, independent pass read the citation against the article text and recorded a verdict.
Support is checked per citation rather than per work — the same source can be cited soundly in one article and wrongly in another. Per-citation recording began recently, so a citation with no recorded check is a gap in the record rather than evidence it went unchecked.
See how references were verified.
Registry ID ref:45196b543e09 · see in the full table