The DOI matches its Crossref record; 30 of the 30 references it cites were found in a registry or an index; 5 of the 16 sources its claims cite were checked for support.
A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play
Open accessAbstract
One program to rule them all: Computers can beat humans at increasingly complex games, including chess and Go. However, these programs are typically constructed for a particular game, exploiting its properties, such as the symmetries of the board on which it is played. Silver et al. developed a program called AlphaZero, which taught itself to play Go, chess, and shogi (a Japanese version of chess) (see the Editorial, and the Perspective by Campbell). AlphaZero managed to beat state-of-the-art programs specializing in these three games. The ability of AlphaZero to adapt to various game rules is a notable step toward achieving a general game-playing system. Science , this issue p. 1140 ; see also pp. 1087 and 1118
Overview
CoChat Summary
TL;DR
AlphaZero is a single reinforcement-learning and MCTS algorithm that, given only game rules, learned superhuman play in chess, shogi, and Go through self-play. It defeated world-champion programs Stockfish, Elmo, and AlphaGo Zero using the same algorithm and network architecture across all three games.
Key findings
- In chess, AlphaZero defeated Stockfish 8, winning 155 games and losing 6 out of 1,000; it also won matches from common human openings and 2016 TCEC opening positions.
- In shogi, AlphaZero defeated Elmo, winning 98.2% of games as black and 91.2% overall; in Go, it defeated AlphaGo Zero, winning 61% of games.
- AlphaZero first outperformed Stockfish after 4 hours (300,000 steps), Elmo after 2 hours (110,000 steps), and AlphaGo Lee after 30 hours (74,000 steps); training ran for 700,000 steps, taking about 9 hours in chess, 12 hours in shogi, and 13 days in Go.
- AlphaZero searched only about 60,000 positions per second in chess and shogi, versus 60 million for Stockfish and 25 million for Elmo, yet still defeated Stockfish with 1/10 thinking time and won 46% against Elmo with 1/100 thinking time.
Methods
- AlphaZero uses a deep neural network \(f_\theta(s)\) that outputs move probabilities \(p\) and a scalar value \(v\), trained by reinforcement learning from self-play starting from random parameters; MCTS uses these outputs to guide search, with 800 simulations per move during training.
- The same algorithm, network architecture, and hyperparameters were used for chess, shogi, and Go, with no domain knowledge except game rules; no opening book, endgame tablebases, or domain-specific heuristics were used.
- Training used 5,000 first-generation TPUs to generate self-play games and 16 second-generation TPUs to train networks; evaluation matches used 3 hours per game plus 15 seconds per move, with opponents on 44 CPU cores and AlphaZero on four first-generation TPUs and 44 CPU cores.
Limitations
- Stated: AlphaZero did not use opening books, endgame tablebases, or domain-specific heuristics; the authors note that some such techniques could further improve performance but were left for future research.
- Stated: hyperparameters were reused across games without game-specific tuning, except for exploration noise and the learning rate schedule; matches against an opening book were low in diversity, and forced randomization increased losses.
- Evident from the text: results cover only three games with separate trained instances, no cross-game transfer is demonstrated, and evaluation compares AlphaZero on TPU hardware against CPU-based opponents, so hardware and time-control differences qualify direct strength comparisons.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
3 Valid · 2 Not valid · 8 Not checked · 43 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“Apart from the representation of positions and actions described above, AlphaZero uses the same network architecture as AlphaGo Zero ( 9 ), briefly recapitulated here.”
What the sources say
- Supportedcites [9] source
The Neural Network Architecture Comparison (p. 29) identifies AlphaGo Zero’s architecture as a residual tower with policy and value heads. The citation supplies the AlphaGo Zero architecture that the citing authors say AlphaZero uses; the comparison to AlphaZero is their own statement. (LLM, conf=0.88)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“and in Go, AlphaZero first outperformed AlphaGo Lee ( 9 ) after 30 hours (74,000 steps).”
What the sources say
- Partly supportedcites [9] sourceExperimental — accuracy not yet measured
Section 2, p. 7 reports that AlphaGo Zero outperformed AlphaGo Lee, but says this happened after 36 hours, not 30. The passage does not give 74,000 steps, and it names the system AlphaGo Zero rather than AlphaZero. (LLM, conf=0.90)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The training algorithm achieved similar performance in all independent runs (see fig. S3), suggesting that the high performance of AlphaZero's training algorithm is repeatable.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“AlphaZero also defeated Stockfish when given 1 / 10 as much thinking time as its opponent (i.e. searching ~ 1 / 10 , 000 as many positions),”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“and won 46% of games against Elmo when given 1 / 100 as much time (i.e. searching ~ 1 / 40 , 000 as many positions), see Fig. 2.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
23 more claims
Create a free account to see all 56 claims and to verify the 3 references still open.
- Probably not a substantial claim. 10
- It may state two separate claims as one. 23
- The quote only partly states this claim. 1
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Games have historically served as highly effective metrics for evaluating progress in deep reinforcement learning research”
From BALROG: Benchmarking Agentic LLM and VLM Reasoning On Games (2024) · cites this paper as Silver et al., 2018
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“In 2017, AlphaGo, using the same learning algorithms in TD-Gammon and a billion times more computing power, soundly beat Ke Jie, the world Go champion, in a game that is to chess what chess is to backgammon in complexity (Silver et al., 2018).”
From Large Language Models and the Reverse Turing Test (2023) · cites this paper as Silver et al., 2018
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Most of the attention was on who won, but more important was how it won: AlphaGo was creative and had discovered new moves and board positions that were outside human experience.”
From Large Language Models and the Reverse Turing Test (2023) · cites this paper as Silver et al., 2018
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“The primary method for applying more computation at inference, common in the AlphaZero family of agents2,72, is scaling the tree search and running multiple independent search attempts.”
From Olympiad-level formal mathematical reasoning with reinforcement learning (2026) · cites this paper as [2]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“Modern AI can write computer programs (Li et al., 2022) or play chess at super-human level (Silver et al., 2018), but the ability of AI to perceive and act in the world remains far below human level.”
From Scaling Instructable Agents Across Many Simulated Worlds (2024) · cites this paper as Silver et al., 2018
1 more claim about this paper
Create a free account to see every claim other papers make about this one.
Citations
Cites
Mastering the game of Go with deep neural networks and tree search
Silver, D., Huang, A., Maddison, C. J., Guez, A., Sifre, L., van den Driessche, G., Schrittwieser, J., Antonoglou, I., Panneershelvam, V., Lanctot, M., et al. (2016) · Nature · doi:10.1038/nature16961 · Cited by 15,612
Self-citationAdvances in neural information processing systems 7
(1996) · Computers & Mathematics with Applications · doi:10.1016/s0898-1221(96)90463-0 · Cited by 14,281
Identity mappings in deep residual networks
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46493-0_38 · Cited by 10,741
Identity Mappings in Deep Residual Networks
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · arXiv · doi:10.48550/arxiv.1603.05027 · Cited by 10,665
Mastering the game of Go without human knowledge
Silver, D., Schrittwieser, J., Simonyan, K., Antonoglou, I., Huang, A., Guez, A., Hubert, T., Baker, L., Lai, M., Bolton, A., et al. (2017) · Nature · doi:10.1038/nature24270 · Cited by 9,084
Self-citationIn-Datacenter Performance Analysis of a Tensor Processing Unit
Jouppi, N. P., Young, C., Patil, N., Patterson, D., Agrawal, G., Bajwa, R., Bates, S., Bhatia, S., Boden, N., Borchers, A., et al. (2017) · Proceedings of the 44th Annual International Symposium on Computer Architecture · doi:10.1145/3079856.3080246 · Cited by 4,537
Encyclopedia of Artificial Intelligence
(2009) · IGI Global eBooks · doi:10.4018/978-1-59904-849-9 · Cited by 1,396
Campbell, M., Hoane, A. J., & Hsu, F.-H. (2002) · Artificial Intelligence · doi:10.1016/s0004-3702(01)00129-1 · Cited by 1,226
Lim, T.-S., Loh, W.-Y., & Shih, Y.-S. (2000) · Machine Learning · doi:10.1023/a:1007608224229 · Cited by 1,133
An analysis of alpha-beta pruning
Knuth, D. E., & Moore, R. W. (1975) · Artificial Intelligence · doi:10.1016/0004-3702(75)90019-3 · Cited by 1,004
Research works printed but not linked: 7
- C. J. Maddison A. Huang I. Sutskever D. Silver paper presented at the International Conference on Learning Representations 2015 San Diego CA 7 to 9 May 2015.
- D. N. L. Levy M. Newborn How Computers Play Chess (Ishi Press 2009).
- W. Steinitz The Modern Chess Instructor (Edition Olms 1990).
- O. Arenz “Monte Carlo chess ” master’s thesis Technische Universität Darmstadt (2012).
- 10.1023/a:1007634325138 · DOI
- T. Kaneko K. Hoki in Advances in Computer Games – 13th International Conference ACG 2011 Revised Selected Papers Tilburg Netherlands 20 to 22 November 2011 (Springer 2012) pp. 158–169.
- R. Ramanujan A. Sabharwal B. Selman in Proceedings of the 26th Conference on Uncertainty in Artificial Intelligence (UAI 2010) Catalina Island CA 8 to 11 July (AUAI Press 2010).
Cited by
Items 3,001–3,100 of 3,622
Doya, K. (2023) · Cambridge University Press eBooks · doi:10.1017/9781108755610.013 · Cited by 1
Investigating Action-Space Generalization in Reinforcement Learning for Recommendation Systems
Naik, A., Chang, B., Karatzoglou, A., Mladenov, M., Ed, H., & Chen, M. (2023) · International World Wide Web Conference (WWW) · doi:10.1145/3543873.3587661
“RL techniques have been applied with remarkable success on various kinds of applications—from games like TD-gammon [23], Atari [15], Go [20], to robotics manipulation tasks like solving a Rubik’s cube [1].”
MethodThe Survey of Self-play Method in Computer Games
Li, X., Wang, Z., Liu, B., & Wu, L. (2023) · Communications in computer and information science · doi:10.1007/978-981-99-2789-0_11 · Cited by 1
“Gongzhu” Strategy Based on Convolutional Neural Network
Wu, L., Wu, Q., Zhong, H., et al. (2023) · Communications in computer and information science · doi:10.1007/978-981-99-2789-0_10
One Class of Cubic Regularized Newton Methods for Min-Max Optimization Problems
高, 瑞. (2023) · Pure Mathematics · doi:10.12677/pm.2023.135129
Playing Fight the Landlord with Tree Search and Hidden Information Evaluation
He, Z., Zhang, X., Wang, X., Liu, L., & Xing, X. (2022) · Asian Conference on Artificial Intelligence Technology (ACAIT) · doi:10.1109/acait56212.2022.10137871
Multi-step Prediction using Tree Generation for Reinforcement Learning
Prakash, K. (2022) · San José State University ScholarWorks (San Jose State University) · doi:10.31979/etd.63be-cbuw
“3 Tree Search Tree search can be done with any algorithm suitable to the problem environment, such as the one proposed in [18, 3] The search algorithm needs to identify which leaf nodes are desirable to reach.”
BackgroundMethodInfluentialAggarwal, V., & Agarwal, M. (2023) · Springer handbooks · doi:10.1007/978-3-030-96729-1_8 · Cited by 1
The Role of Multiple Intelligences Theory in Learning
Winandar, A. K., Oktiawati, U. Y., Winantu, W., et al. (2023) · al-fikrah Jurnal Manajemen Pendidikan · doi:10.31958/jaf.v11i1.8588
“Multiple intelligences are developed into nine theories of children's intelligence, in each child this intelligence is developed, moderately developed and highly developed in the child himself (Chen & Zhu, 2022; Silver et al., 2018; Zhuang et al., 2021) .”
BackgroundA New Deep Reinforcement Learning Based Robot Path Planning Algorithm without Target Network
Cao, Y., Zhao, D. Y., & Cao, X. S. (2023) · International Symposium on Autonomous Systems (ISAS) · doi:10.1109/isas59543.2023.10164323
“In 2016, Google developed AplhaGo [17], using deep learning and Monte Carlo search tree, defeated the then human Go master Lee Sedol, and released an upgraded version AphlaZero [18] in the following year which set off a wave of deep learning.”
MethodPekiştirmeli Öğrenme ile Tenis Oyunu Simülasyonu Gerçekleştirimi
OSPANOV, B., & Demirci, M. F. (2023) · Journal of Turkish Operations Management · doi:10.56554/jtom.1273051
Comparing Reinforcement Learning and Human Learning using the Game of Hidden RulesPreprint
Pulick, E., Meñkov, V., Mintz, Y. D., et al. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2306.17766
“(*,*,*,[19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36],Nearby) \ (*,*,*,[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18],[0,1,2,3])”
BackgroundMuddling-Through and Deep Learning for Bureaucratic Decision-Making
Cox, L. A. (2023) · International series in management science/operations research/International series in operations research & management science · doi:10.1007/978-3-031-32013-2_8
The cost of passing - using deep learning AIs to expand our understanding of the ancient game of Go
Egri-Nagy, A., & Törmänen, A. (2023) · International Journal of Networking and Computing · doi:10.15803/ijnc.13.2_258
Chunxu, M. U. (2023) · Applied and Computational Engineering · doi:10.54254/2755-2721/5/20230526 · Cited by 1
Machine learning algorithm and training in Go—Take three influential program as example
Zhang, Q. (2023) · Applied and Computational Engineering · doi:10.54254/2755-2721/6/20230845
Of Flesh and Steel: Computational Creativity in Music and the Body Issue
Merlini, M., & Nicoletti, S. M. (2020) · INSAM Journal of Contemporary Music Art and Technology · doi:10.51191/issn.2637-1898.2020.3.4.24 · Cited by 1
“…for example, with the game of chess: while mastering chess was arguably considered a definitive mark of intelligence before the advent of computers, supporting this consideration is not so easy today, where computational devices can challenge each other to establish supremacy (Silver et al. 2018).”
BackgroundContinual Learning for Classification Tasks
Xue-jun, H. (2023) · Carleton University's Institutional Repository (MacOdrum Library, Carleton University) · doi:10.22215/etd/2023-15555
Monte Carlo Tree Search and Machine Learning Techniques on Block Go Programs
Yen, S.-J., & Lee, K.-W. (2023) · International Conference on Applied System Innovation · doi:10.1109/icasi57738.2023.10179581
“[1][2][3] Black Go, a game like Go, was developed after years of research, experimentation, and collaboration between early childhood education expert Zhang Yuanxi in Taiwan and professional Go player Zhang Xu in Japan.”
BackgroundSimilar failures of consideration arise in human and machine planningPreprint
Zhang, A., Langenkamp, M., Kleiman‐Weiner, M., et al. (2023) · doi:10.31234/osf.io/jqhac
Recent Applications and Future Research
Jardim, R. R.-A. J. (2023) · Springer eBooks · doi:10.1007/978-3-031-37345-9_6
Reinforcement Learning for Data Science
Barkan, J., Moran, M., & Gordon, G. (2023) · Springer eBooks · doi:10.1007/978-3-031-24628-9_24
Floridi, L. (2023) · Oxford University Press eBooks · doi:10.1093/oso/9780198883098.002.0009
Floridi, L. (2023) · Oxford University Press eBooks · doi:10.1093/oso/9780198883098.003.0001
Floridi, L. (2023) · Oxford University Press eBooks · doi:10.1093/oso/9780198883098.002.0004
Mapping the Ethics of Algorithms
Floridi, L. (2023) · Oxford University Press eBooks · doi:10.1093/oso/9780198883098.003.0007
How to Deliver a Good AI Society: Some Recommendations
Floridi, L. (2023) · Oxford University Press eBooks · doi:10.1093/oso/9780198883098.003.0010
Floridi, L. (2023) · Oxford University Press eBooks · doi:10.1093/oso/9780198883098.002.0007
Bad Practices: The Misuse of AI for Social Evil
Floridi, L. (2023) · Oxford University Press eBooks · doi:10.1093/oso/9780198883098.003.0008
List of Most Common Abbreviations and Acronyms
Floridi, L. (2023) · Oxford University Press eBooks · doi:10.1093/oso/9780198883098.002.0011
The Gambit: AI Impact on Climate Change
Floridi, L. (2023) · Oxford University Press eBooks · doi:10.1093/oso/9780198883098.003.0011
Floridi, L. (2023) · Oxford University Press eBooks · doi:10.1093/oso/9780198883098.002.0010
Good Practices: The Proper Use of AI for Social Good
Floridi, L. (2023) · Oxford University Press eBooks · doi:10.1093/oso/9780198883098.003.0009
Floridi, L. (2023) · Oxford University Press eBooks · doi:10.1093/oso/9780198883098.002.0005
Conclusion: The Green and the Blue
Floridi, L. (2023) · Oxford University Press eBooks · doi:10.1093/oso/9780198883098.003.0013
Glazer, B. J., Lifferth, J., & Lopez, C. F. (2023) · Frontiers in Cell and Developmental Biology · doi:10.3389/fcell.2023.1198359
“This is similar to the approach used by the AlphaZero algorithm (Silver et al., 2018), that proved to be exceptionally effective at searching the combinatorially large space of moves in games like chess and Go.”
MethodReinforcement Learning with Neural Network-based Deterministic Game Tree Approximation
Prakash, K., & Ishigaki, G. (2023) · IEEE International Conference on Big Data Computing Service and Applications (BigDataService) · doi:10.1109/bigdataservice58306.2023.00039
“Also, when there is additional information about an environment, such as its transition function, it is known that the use of a search algorithm for planning enhances the scalability and learning efficiency of RL [2].”
BackgroundHidayatullah, R., Mahendra, I., Wibowo, A. S., et al. (2021) · Jurnal Ilmu Komputer dan Informatika · doi:10.54082/jiki.4
Birkoben, T., & Kohlstedt, H. (2023) · Springer series on bio- and neurosystems · doi:10.1007/978-3-031-36705-2_1 · Cited by 8
Reinforcement Learning with Goal Relabeling and Dynamic Model for Robotic Tasks
Dong, K., Luo, Y., Wang, Y., Fang, S. C., Liu, Y., Cheng, E., Sun, Z., Zhang, Q., & Song, B. (2023) · IEEE International Conference on Real-time Computing and Robotics (RCAR) · doi:10.1109/rcar58764.2023.10250060
Discrete Denoising Diffusion Approach to Integer Factorization
Freivalds, K., Ozoliņš, E., & Bārzdiņš, G. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-44207-0_11
Adversarial Multi-Agent Reinforcement Learning for Fault-Tolerant Design of Complex Systems
Grimstad, J., & Morozov, A. S. (2023) · European Safety and Reliability Conference (ESREL) · doi:10.3850/978-981-18-8071-1_p375-cd
Diagnosing and exploiting the computational demands of videos games for deep reinforcement learningPreprint
Govindarajan, L. N., Liu, R. G., Linsley, D. A., Ashok, A. K., Reuter, M., Frank, M. J., & Serre, T. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2309.13181 · Cited by 1
“…steady progress in the field of dRL, and newer reinforcement learning algorithms have yielded agents that achieve human or super-human performance in a variety of challenges – from Chess to Go and from Atari games to Starcraft (Mnih et al., 2015; Silver et al., 2017; 2018; Vinyals et al., 2019).”
A Convolutional Neural Network Approach to General Game Playing
Wang, Y., Zhang, H., & Jiang, G. (2023) · Frontiers in artificial intelligence and applications · doi:10.3233/faia230563 · Cited by 2
Expediting Self-Play Learning in AlphaZero-Style Game-Playing Agents
Björnsson, Y., Jönsson, R., & Jónsson, S. I. (2023) · Frontiers in artificial intelligence and applications · doi:10.3233/faia230279
“For example, AlphaZero played 40 million chess games with the help of 5,000 special-purpose computing units (TPUs) to reach state-of-the-art performance in the game of chess [1].”
BackgroundAn Introduction to Machine Learning in Molecular Sciences
Liu, H., & Qu, C. (2023) · Challenges and advances in computational chemistry and physics · doi:10.1007/978-3-031-37196-7_1
Photonic Reservoir Computing for Spectrum Awareness Applications
Chembo, Y. K. (2023) · IEEE Research and Applications of Photonics in Defense Conference (RAPID) · doi:10.1109/rapid54473.2023.10264717
“It is well known that biological brains can effortlessly outperform Turing–von Neumann machines (TvN) for certain classes of computational tasks [1-3].”
BackgroundA Reinforcement Learning Based Slope Limiter for Two-Dimensional Finite Volume Schemes
Keim, J., Schwarz, A., Chiocchetti, S., et al. (2023) · Springer proceedings in mathematics & statistics · doi:10.1007/978-3-031-40860-1_22
Investigating quantum many-body systems with tensor networks, machine learning and quantum computers
Kottmann, K. (2022) · UPCommons institutional repository (Universitat Politècnica de Catalunya) · doi:10.5821/dissertation-2117-379461 · Cited by 3
Learning to Behave: Reinforcement Learning in Human Contexts
den Hengst, F. (2023) · doi:10.5463/thesis.390
Buzer, L., & Cazenave, T. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-44505-7_13 · Cited by 2
“It was used to build superhuman game playing programs such as AlphaGo [32], AlphaZero [33] and Katago [35].”
MethodA framework for safe decision making: A convex duality approach
Bernasconi, M., Cacciamani, F., & Castiglioni, M. (2023) · Intelligenza Artificiale · doi:10.3233/ia-230008
“These range from recreative applications, e.g., Chess [19], Texas hold’em poker [7], Go [18], to security applications such as anti-poaching patrolling [20].”
Cutajar, C., & Bajada, J. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-47546-7_16
Research on Strategies for Tripeaks Variant with Various Layouts
Gao, Y., Xu, S., & Du, S. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-46314-3_7
Impact of Heterogeneity on Multi-Agent Reinforcement Learning
Marques, R. F., & Patrocínio, Z. K. G. (2023) · doi:10.5753/eniac.2023.234582
“Examples of problems in which RL has been successfully applied are playing video games at a super-human level in Atari [Mnih et al. 2015], in Chess [Silver et al. 2018], in Go [Silver et al. 2016], but also in robotics [Kober et al. 2013], in routing [Nazari et al. 2018] or in chronic illness…”
BackgroundElucidating Common Fallacies and Misconceptions Around LLMsPreprint
Khan, M. A.-Z., Innan, N., & Hammujuddy, J. (2023) · Preprints.org · doi:10.20944/preprints202311.1559.v1
Kahn, S. J. M. (2023) · Asian Journal of Philosophy · doi:10.1007/s44204-023-00118-7
Offensive and Defensive Attention Reward in Reinforcement Learning for Football Game
Liu, Y., Zhu, Y., Liu, S., & Zhao, D. (2023) · Chinese Control and Decision Conference (CCDC) · doi:10.1109/ccdc58219.2023.10327432
“AlphaGo Zero [16] adopts a self-play paradigm and is able to train agents to master the skills of playing chess, shogi, and Go simultaneously without human data, which proves the huge potential of DRL combined with self-play training for the first time.”
BackgroundLedinauskas, E., & Anisimovas, E. (2023) · SciPost Physics · doi:10.21468/scipost.report.8117
Ledinauskas, E., & Anisimovas, E. (2023) · SciPost Physics · doi:10.21468/scipost.report.7880
Gradient-Norm Constrained Algorithm on Offline and Online LearningPreprint
Zhang, H. (2023) · SSRN Electronic Journal · doi:10.2139/ssrn.4663051
Marroquín-Cano, S. F., Escobar-Gómez, E. N., Morales, E. F., Ramírez-Álvarez, E., Chandomí-Castellanos, E., Gasga, P., & Rodríguez-Sanchez, F. J. (2023) · SSRN Electronic Journal · doi:10.2139/ssrn.4666187
Park, S., & Yoo, J. (2023) · Transactions of Korean Society of Automotive Engineers · doi:10.7467/ksae.2023.31.12.949
Exploring the constraints on artificial general intelligence: a game-theoretic perspectivePreprint
Ismail, M. S. (2023) · SSRN Electronic Journal · doi:10.2139/ssrn.4666645
Métodos eficientes para geração de desafios de xadrez do tipo mate em n movimentos
de Albuquerque, K. M. M., de Figueiredo, C. F. S., de Sousa, V. H. F., et al. (2019) · Anais do Simpósio Brasileiro de Pesquisa Operacional · doi:10.59254/sbpo-2019-106986
Mertens, P., Buxmann, P., Heß, T., Hinz, O. F., Muntermann, J., & Schümann, M. (2023) · Springer eBooks · doi:10.1007/978-3-662-67573-1_3
Exploring the Constraints on Artificial General Intelligence: A Game-Theoretic PerspectivePreprint
Ismail, M. S. (2023) · SSRN Electronic Journal · doi:10.2139/ssrn.4680247
Optimizing Strategy Games: Ant Colony Optimization vs. Minimax Algorithm
Angeles-García, Y., Legaria-Santiago, V. K., Anzueto-Ríos, Á., et al. (2023) · IEEE Symposium Series on Computational Intelligence (SSCI) · doi:10.1109/ssci52147.2023.10371917
“However, these techniques require knowledge from previous games from expert systems (such as humans or other AI systems) or compete against themselves until specializing in a specific game [5]–[7].”
BackgroundArtificial Intelligence in Neuroscience
Xiao, W., Zhang, M., & Kreiman, G. (2024) · Cambridge University Press eBooks · doi:10.1017/9781108917339.010
Attracting Investment and Reducing Poverty in Africa Through AI Technologies
(2023) · Engineering Open Access · doi:10.33140/eoa.01.04.07
Una introducciön amable pero rigurosa al aprendizaje por refuerzo
Computación, C. D. I. E., Meza, M. A. M., Menchaca-Méndez, R., Centro de Investigación en Computación del, I. P. N., Méndez, R. M., & Centro de Investigación en Computación del, I. P. N. (2023) · RECIBE REVISTA ELECTRÓNICA DE COMPUTACIÓN INFORMÁTICA BIOMÉDICA Y ELECTRÓNICA · doi:10.32870/recibe.v12i1.268 · Cited by 2
Robust and adaptive decision-making: a reinforcement learning perspective
Xue, W. (2023) · DR-NTU (Nanyang Technological University) · doi:10.32657/10356/173125
Implementing the AlphaZero algorithm for Connect Four: A deep reinforcement learning approach
Guo, Y. (2024) · Applied and Computational Engineering · doi:10.54254/2755-2721/33/20230228
A research of artificial intelligence game agent application
Lan, Y., & Li, Z. (2024) · Applied and Computational Engineering · doi:10.54254/2755-2721/37/20230486 · Cited by 1
NISHINA, K., & Fujita, S. (2024) · Journal of Japan Society for Fuzzy Theory and Intelligent Informatics · doi:10.3156/jsoft.36.1_571
Advanced Machine Learning Techniques in Gomoku: Strategy,Implementation, and Analysis
Han, J.-J. (2024) · Science and Technology of Engineering Chemistry and Environmental Protection · doi:10.61173/4sda3g67
Ramón, C. A. (2023) · Boletín das ciencias · doi:10.54954/202395251
Learning to Adapt: Communication Load Balancing via Adaptive Deep Reinforcement Learning
Wu, D., Xu, Y. T., Li, J., Jenkin, M. R. M., Hossain, E., Jang, S., Xin, Y., Zhang, C., Liu, X., & Dudek, G. (2023) · Globecom · doi:10.1109/globecom54140.2023.10437528 · Cited by 1
“Examples here include AlphaZero [21], which has defeated human world champions in the games of chess, Shogi and Go.”
BackgroundQuantifying Feature Importance of Games and Strategies via Shapley Values
Fujii, S. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-54968-7_8
“Notably, the amalgamation of conventional algorithms with deep neural networks — exemplified by AlphaZero [14] in perfect information games, Deep CFR [2] and NFSP [6] in imperfect information games, has demonstrated outstanding performance, occasionally surpassing top human players.”
BackgroundMerging Neural Networks with Traditional Evaluations in Crazyhouse
Makovec, A., Pirker, J., & Guid, M. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-54968-7_2
Pierre, M., Cohen-Solal, Q., & Cazenave, T. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-54968-7_13 · Cited by 1
Efficient deep reinforcement learning in robotic motion planning
Luo, S. (2024) · University of Groningen research database (University of Groningen / Centre for Information Technology) · doi:10.33612/diss.922202894
Go Players Should Not Trust AI Win Rate
Rendu, Q. (2023) · Journal of Go Studies · doi:10.62578/476299
Exploring the Impact of AI on Go Education: A Teacher Survey
Trinks, D., & Oh, C.-M. (2023) · Journal of Go Studies · doi:10.62578/689093 · Cited by 2
Is Artificial Intelligence for Retinopathy of Prematurity Ready to Go?
Binenbaum, G. (2024) · JAMA Ophthalmology · doi:10.1001/jamaophthalmol.2024.0218
Data analysis of tactical wargaming based on data mining
Wu, L. (2024) · Journal of Computational Methods in Sciences and Engineering · doi:10.3233/jcm-237083
“In recent years, with the continuous development of game agents such as chess, Go, StarCraft and Dota [11,12,13,14], the application of artificial intelligence related technology in game AI has been vigorously promoted.”
Enhanced Deep Reinforcement Learning for Parcel Singulation in Non-Stationary Environments
Shen, J., Lu, H., Zhang, H., Lyu, S., & Lu, Y. (2024) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp48485.2024.10446437 · Cited by 2
“However, the performance of the priority-learned method increases only a little as it ignored the uncertainty of input parcels when formulating the problem [8–10].”
BackgroundKhodadadi, Z., Trpevski, D., Lindroos, R., et al. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.03.12.584462
“Based on this simplification, powerful artificial neural systems have been created, outperforming humans on multiple tasks (Silver et al. , 2018).”
Singh, V. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.03.19.585672
“Its applications range from robotics [54], [55], [56] to games such as Go, chess [57], [58], and even complex video games such as StarCraft II [59].”
BackgroundInampudi, S. S. (2024) · Communications in computer and information science · doi:10.1007/978-3-031-56700-1_36
Artificial Intelligence und Quantum Computing
Wirtz, B. W. (2024) · Springer eBooks · doi:10.1007/978-3-658-41467-2_8
An Empirical Analysis of Gumbel MuZero on Stochastic and Deterministic Einstein Würfelt Nicht!
Kuo, C.-L., Chen, P., Guei, H., Sung, D.-R., Hsueh, C.-H., Wu, T.-R., & Wu, I. (2024) · Communications in computer and information science · doi:10.1007/978-981-97-1711-8_25
Ethical disagreements in technology and management research
Telkamp, J. B. (2023) · doi:10.31274/td-20240329-291
A Puppet Without Strings? Two Considerations on AI and Its Ethics
Marino, D., Cananzi, D. M., & Aragona, F. (2024) · Studies in systems, decision and control · doi:10.1007/978-3-031-50902-5_2
Evaluating Deep Learning-based Chess-Engine Endgame Strategies
Neto, E. P. D. B., Filho, T. M. S., do Rego, T. G., & Malheiros, Y. (2023) · Simpósio Brasileiro de Jogos e Entretenimento Digital (SBGames) · doi:10.5753/sbgames_estendido.2023.233983
“Deep Blue used a game tree search approach, but most recent chess engines have been using machine learning and neural networks [Silver et al. 2018].”
MethodStrang, A., Sewell, D., Kim, A. A., Alcedo, K., & Rosenbluth, D. (2024) · Information Visualization · doi:10.1177/14738716241239018 · Cited by 1
“4 Despite impressive game play, enhanced understanding of the game is typically only achieved by additional analysis of the algorithm’s play post facto.”
BackgroundA Generalist Reinforcement Learning Agent for Compressing Convolutional Neural Networks
Gonzalez-Sahagun, G., Conant-Pablos, S. E., Ortíz-Bayliss, J. C., et al. (2024) · IEEE Access · doi:10.1109/access.2024.3385857
“Another generalist agent learned to play chess, shogi, and Go [31].”
Background
Metadata
- Authors
- Silver, D., Hubert, T., Schrittwieser, J., Antonoglou, I., Lai, M., Guez, A., Lanctot, M., Sifre, L., Kumaran, D., Graepel, T., Lillicrap, T., Simonyan, K., & Hassabis, D.
- Affiliations
- Google DeepMind (United Kingdom) (GB)
- Corresponding author
- Silver, David
- Hubert, Thomas
- Schrittwieser, Julian
- Hassabis, Demis
- ORCID
- Venue
- Science
- Published
- 2018-12-07
- Volume
- 362
- Issue
- 6419
- Pages
- 1140-1144
- ISSN
- 0036-8075 (ISSN-L)
- 1095-9203
- OpenAlex
- W2902907165
- Cited by
- 3,562
- References
- 30
- Citations by year
- 2026: 257 · 2025: 427 · 2024: 512 · 2023: 514 · 2022: 569 · 2021: 546 · 2020: 462 · 2019: 256 · 2018: 13
- FWCI
- 179.89
- Type
- journal article
- Language
- en
- Open access
- bronze
- doi.org
- Locations
- Science (publishedVersion)
- PubMed (publishedVersion)
- UCL Discovery (University College London) (submittedVersion)
- Funders
- DeepMind
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Artificial Intelligence in Games
- Physical Sciences › Computer Science › Artificial Intelligence › Reinforcement Learning in Robotics
- Physical Sciences › Computer Science › Artificial Intelligence › Artificial Intelligence Applications
- Keywords
- AlphaZero, chess, shogi, Go, evaluation function learning, game tree search
- MeSH terms
- Algorithms, Artificial Intelligence, Humans, Reinforcement, Psychology, Software, Video Games
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.