A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play
Open accessThe 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.
Abstract
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 501–600 of 3,632
Will AI Replace Ophthalmologists?
Korot, E., Wagner, S. K., Faes, L., Liu, X., Huemer, J. C., Ferraz, D., Keane, P. A., & Balaskas, K. (2020) · Translational Vision Science & Technology · doi:10.1167/tvst.9.2.2 · Cited by 42
Deep Statistical Model Checking
Gros, T. P., Hermanns, H., Hoffmann, J., Klauck, M., & Steinmetz, M. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-50086-3_6 · Cited by 38
An Adaptive Model of Gaze-based Selection
Chen, X., Acharya, A., Oulasvirta, A., & Howes, A. (2021) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3411764.3445177 · Cited by 61
“(3) It illustrates how machine learning can be used to model human cognition through trial-and-error learning in an artifcial world, rather than through generalisation from large amounts of human data (similar to [58]’s learning through selfplay).”
BackgroundReinforcement learning of rare diffusive dynamics
Das, A., Rose, D. C., Garrahan, J. P., & Limmer, D. T. (2021) · The Journal of Chemical Physics · doi:10.1063/5.0057323 · Cited by 57
Existential risk from AI and orthogonality: Can we have it both ways?
Müller, V. C., & Cannon, M. (2021) · Ratio · doi:10.1111/rati.12320 · Cited by 56
Artificial virtuous agents: from theory to machine implementation
Stenseke, J. (2021) · AI & Society · doi:10.1007/s00146-021-01325-7 · Cited by 54
Han, T., Liao, X., Zeng, C., et al. (2022) · ACM International Conference on emerging Networking EXperiments and Technologies (CoNEXT) · doi:10.1145/3555050.3569125 · Cited by 37
“Armed with deep neural networks, deep reinforcement learning plays the core role in achieving super-human performance in many games and real-world decision-making tasks [33, 39].”
BackgroundForward–forward training of an optical neural network
Oğuz, İ., Ke, J., Weng, Q., Yang, F., Yildirim, M., Dinç, N. U., Hsieh, J.-L., Moser, C., & Psaltis, D. (2023) · Optics Letters · doi:10.1364/ol.496884 · Cited by 43
Audio-Visual Deception Detection: DOLOS Dataset and Parameter-Efficient Crossmodal Learning
Guo, X., Selvaraj, N. M., Yu, Z., Kong, A. W., Shen, B., & Kot, A. C. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.02023 · Cited by 62
“Deep learning algorithms have achieved comparable or even better performance than human in many complex tasks [22, 41, 40, 7, 38].”
BackgroundChitkeshwar, A. K. (2024) · Archives of Computational Methods in Engineering · doi:10.1007/s11831-024-10117-3 · Cited by 56
“AlphaZero trounced world champion programs in these games without any prior domain knowledge other than the game rules, proving the algorithm’s efficacy in learning intricate strategic games [85].”
Wang, J., Guo, Q., & Sun, H. (2024) · Applied Energy · doi:10.1016/j.apenergy.2024.123792 · Cited by 40
Ekundayo, O., & Ezugwu, A. E.-S. (2025) · Applied Soft Computing · doi:10.1016/j.asoc.2025.113378 · Cited by 62
Wang, Z., Li, H., Wu, H., & Wu, Z. (2020) · Mathematical Problems in Engineering · doi:10.1155/2020/7180639 · Cited by 41
Deep reinforcement learning for efficient measurement of quantum devices
Nguyen, V., Orbell, S. B., Lennon, D. T., Moon, H., Vigneau, F., Camenzind, L. C., Yu, L., Zumbühl, D. M., Briggs, G. A. D., Osborne, M. A., et al. (2021) · npj Quantum Information · doi:10.1038/s41534-021-00434-x · Cited by 66
Robot Navigation with Map-Based Deep Reinforcement Learning
Chen, G., Pan, L., Chen, Y., Xu, P., Wang, Z., Wu, P., Ji, J., & Chen, X. (2020) · IEEE International Conference on Networking, Sensing and Control (ICNSC) · doi:10.1109/icnsc48988.2020.9238090 · Cited by 52
“On the other hands, deep reinforcement learning (DRL) methods have achieved remarkable success in many challenging tasks [9]–[11].”
BackgroundOptimizing quantum annealing schedules with Monte Carlo tree search enhanced with neural networks
Chen, Y., Chen, Y., Lee, C., Zhang, S., & Hsieh, C. (2022) · Nature Machine Intelligence · doi:10.1038/s42256-022-00446-y · Cited by 57
“The following three points highlight the main differences between DeepMind’s AlphaZero and our modified algorithm QZero.”
BackgroundMethodInfluentialAccurate prediction of ice nucleation from room temperature water
Davies, M. B., Fitzner, M., & Michaelides, A. (2022) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2205347119 · Cited by 39
Revealing the impact of expertise on human planning with a two-player board gamePreprint
van Opheusden, B., Galbiati, G., Kuperwajs, I., Bnaya, Z., li, Y., & Ma, W. J. (2021) · PsyArXiv (OSF Preprints) · doi:10.31234/osf.io/rhq5j · Cited by 38
“A promising candidate for modeling this learning process is deep reinforcement learning, specifically algorithms such as AlphaZero[53] and SAVE[54], which combine prior experience with forward planning at decision time.”
MethodIs Power-Seeking AI an Existential Risk?
Carlsmith, J. (2022) · arXiv · doi:10.48550/arxiv.2206.13353 · Cited by 37
Phaf, R. H. (2020) · Theory & Psychology · doi:10.1177/0959354319898250 · Cited by 42
“The neural-network program AlphaZero, for instance, has been trained solely via playing against itself, not only playing the board game Go, but also chess and shogi, with only the basic rules being preprogrammed (Silver et al., 2018).”
MethodOptimization of DEM parameters using multi-objective reinforcement learning
Westbrink, F., Elbel, A., Schwung, A., & Ding, S. X. (2020) · Powder Technology · doi:10.1016/j.powtec.2020.10.067 · Cited by 38
Scaffolding Human Champions: AI as a More Competent Other
Sætra, H. S. (2022) · Human Arenas · doi:10.1007/s42087-022-00304-8 · Cited by 63
“…sacrifices previously thought to be madness against an opponent such as Stockfish (Nielsen, 2019) Following this breakthrough, DeepMind members also released an article on the algorithm behind it and how AlphaZero is one step away from becoming a general game-playing AI (Silver et al., 2018).”
AI in Human-computer Gaming: Techniques, Challenges and Opportunities
Yin, Q., Yang, J. J., Huang, K., Zhao, M., Ni, W., Liang, B., Huang, Y., Wu, S., & Wang, L. (2023) · Machine Intelligence Research · doi:10.1007/s11633-022-1384-6 · Cited by 55
“In this paper, the AIs cover: AlphaGo, AlphaGo Zero, AlphaZero for board game Go; Libratus, DeepStack, DouZero and Suphx for card games HUNL, DouDiZhu and Mahjong, respectively; FTW for FPS game Quake III Arena in Capture the Flag model; AlphaStar, Commander, OpenAI Five and JueWu for StarCraft, Dota2 and Honor of Kin…”
BackgroundInfluentialYang, D., Karimi, H. R., & Pawełczyk, M. (2023) · Control Engineering Practice · doi:10.1016/j.conengprac.2023.105475 · Cited by 36
Ahmed, S. N., & Prakasam, P. (2023) · Progress in Biophysics and Molecular Biology · doi:10.1016/j.pbiomolbio.2023.07.001 · Cited by 51
Amin, M. A., Suleman, A., Waseem, M., Iqbal, T., Aziz, S., Faiz, M. T., Zulfiqar, L., & Saleh, A. M. (2023) · IEEE Access · doi:10.1109/access.2023.3302895 · Cited by 37
“We calculate the whole storage cluster and store the cost every cycle until we get the capital cost and replace the battery and the cycle number given in [25].”
MethodAdopting artificial intelligence in cardiovascular medicine: a scoping review
Makimoto, H., & Kohro, T. (2023) · Hypertension Research · doi:10.1038/s41440-023-01469-7 · Cited by 47
“Recently, AIs have demonstrated memory, learning, and recognition abilities that surpass those of humans [2, 3].”
Reinforcement learning based temperature control of a fermentation bioreactor for ethanol production
Rajasekhar, N., Radhakrishnan, T. K., & Mohamed, S. N. (2024) · Biotechnology and Bioengineering · doi:10.1002/bit.28784 · Cited by 42
“…is updated until Q ‐ value converges into optimal Q ‐ function and it is shown in Equation (3). where, α → learning rate. chess, shogi and robots by learning optimal policies through trial ‐ and ‐ error methods and estimating them from environmental image inputs (Silver et al., 2016, 2017, 2018).”
Reinforcement Twinning: From digital twins to model-based reinforcement learning
Schena, L., Marques, P. A. D. M., Poletti, R., Ahizi, S., den Berghe, J. V., & Mendez, M. A. (2024) · Journal of Computational Science · doi:10.1016/j.jocs.2024.102421 · Cited by 54
Dogru, O., Velswamy, K., & Huang, B. (2021) · Engineering · doi:10.1016/j.eng.2021.04.027 · Cited by 37
A New Trajectory Tracking Algorithm for Autonomous Vehicles Based on Model Predictive Control
Huang, Z., Li, H., Li, W., Liu, J., Huang, C., Yang, Z., & Fang, W. (2021) · Sensors · doi:10.3390/s21217165 · Cited by 44
Hierarchical reinforcement learning with dynamic recurrent mechanism for course recommendation
Lin, Y., Lin, F., Zeng, W., Xiahou, J., Li, L., Wu, P., Liu, Y., & Miao, C. (2022) · Knowledge-Based Systems · doi:10.1016/j.knosys.2022.108546 · Cited by 36
Data-Driven Wind Farm Control via Multiplayer Deep Reinforcement Learning
Dong, H., & Zhao, X. (2022) · IEEE Transactions on Control Systems Technology · doi:10.1109/tcst.2022.3223185 · Cited by 36
“This design can significantly improve the learning stability, which has been proven by many relevant studies [4], [14], [17], [18], [25].”
BackgroundReinforced Genetic Algorithm Learning for Optimizing Computation GraphsPreprint
Paliwal, A., Gimeno, F., Nair, V. B., Li, Y., Lubin, M., Kohli, P., & Vinyals, O. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1905.02494 · Cited by 73
“Combining learning with a search algorithm has been shown to be successful (e.g., (Silver et al., 2017; 2018)), and our work can be seen as an instance of the same high-level idea.”
MethodYang, F., Yang, Y., Ni, S., Liu, S., Xu, C., Chen, D., & Zhang, Q. (2023) · Transportation Research Part C Emerging Technologies · doi:10.1016/j.trc.2023.104237 · Cited by 37
Scale-aware deep reinforcement learning for high resolution remote sensing imagery classification
Liu, Y., Zhong, Y., Shi, S., & Zhang, L. (2024) · ISPRS Journal of Photogrammetry and Remote Sensing · doi:10.1016/j.isprsjprs.2024.01.013 · Cited by 40
Wang, Y., Liu, H., Zhang, M., Luo, X., & Qu, H. (2024) · Neural Networks · doi:10.1016/j.neunet.2024.106244 · Cited by 55
The Paradigm Shifts in Artificial Intelligence
Dhar, V. (2024) · Communications of the ACM · doi:10.1145/3664804 · Cited by 62
“42 I start with the era of knowledge-based systems, which began to focus on the systematic representation and use of human knowledge at scale.”
The Evolution of Reinforcement Learning in Quantitative Finance: A Survey
Pippas, N., Ludvig, E. A., & Turkay, Ç. (2025) · ACM Computing Surveys · doi:10.1145/3733714 · Cited by 71
Liu, Q., Song, Z., Liang, Y., Xie, Z., Zhang, S., Zhang, J., & Li, Y. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.130113 · Cited by 57
Setting Up Experimental Bell Tests with Reinforcement Learning
Melnikov, A. A., Sekatski, P., & Sangouard, N. (2020) · Physical Review Letters · doi:10.1103/physrevlett.125.160401 · Cited by 46
Deep Reinforcement Learning for DER Cyber-Attack Mitigation
Roberts, C., Ngo, S.-T., Milesi, A., Peisert, S., Arnold, D., Saha, S. S., Scaglione, A., Johnson, N. G., Андреевич, К. А., & Fradkin, D. (2020) · IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids (SmartGridComm) · doi:10.1109/smartgridcomm47815.2020.9302997 · Cited by 38
“Thanks to its flexibility, DRL has been successfully applied to robotic control [9], video games [10], [11] and board game playing [12], [13].”
MethodWardi, G., Owens, R., Josef, C., Malhotra, A., Longhurst, C., & Nemati, S. (2023) · Critical Care Medicine · doi:10.1097/ccm.0000000000005894 · Cited by 35
“Although hardware improvements were also needed to achieve super-human performance, the latest iteration of such systems (1) had to embrace more powerful ML algorithms and were programmed to learn continually.”
BackgroundThe function of consciousness is to generate experiencePreprint
Cleeremans, A., & Tallon‐Baudry, C. (2021) · doi:10.31234/osf.io/jfpw2 · Cited by 34
Look-ahead based reinforcement learning for robotic flow shop scheduling
Kim, H.-J., & Lee, J.-H. (2023) · Journal of Manufacturing Systems · doi:10.1016/j.jmsy.2023.02.002 · Cited by 34
Reinforcement Learning-based approach for dynamic vehicle routing problem with stochastic demand
Zhou, C., Ma, J., Douge, L., Chew, E. P., & Lee, L. H. (2023) · Computers & Industrial Engineering · doi:10.1016/j.cie.2023.109443 · Cited by 40
Functional neuronal circuits emerge in the absence of developmental activity
Barabási, D. L., Schuhknecht, G. F. P., & Engert, F. (2024) · Nature Communications · doi:10.1038/s41467-023-44681-2 · Cited by 36
“Indeed, research in arti fi cial intelligence and computational neuroscience have high-lighted how weakly structured models can learn to perform complex tasks, even outperforming humans at Chess and Go 4 .”
BackgroundLearning Multipursuit Evasion for Safe Targeted Navigation of Drones
Xiao, J., & Feroskhan, M. (2024) · IEEE Transactions on Artificial Intelligence · doi:10.1109/tai.2024.3366871 · Cited by 48
“With self-play [43]–[45], RL has achieved remarkable success in adversarial environments (or competitive games), such as DeepMind’s AlphaZero [45], AlphaStar [46] and OpenAI Five [47].”
BackgroundArtificial Intelligence for Retrosynthetic Planning Needs Both Data and Expert Knowledge
Strieth‐Kalthoff, F., Szymkuć, S., Molga, K., Aspuru‐Guzik, A., Glorius, F., & Grzybowski, B. A. (2024) · Journal of the American Chemical Society · doi:10.1021/jacs.4c00338 · Cited by 51
Towards the future of physics- and data-guided AI frameworks in computational mechanics
Bai, J., Wang, Y., Jeong, H., Chu, S., Wang, Q. J., Alzubaidi, L., Zhuang, X., Rabczuk, T., Xie, Y. M., Feng, X., et al. (2025) · Acta Mechanica Sinica · doi:10.1007/s10409-025-25340-x · Cited by 44
Sample-efficient Reinforcement Learning in Robotic Table Tennis
Tebbe, J., Krauch, L., Gao, Y., & Zell, A. (2021) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra48506.2021.9560764 · Cited by 59
Enzymeless DNA Base Identification by Chemical Stepping in a Nanopore
Qing, Y., & Bayley, H. (2021) · Journal of the American Chemical Society · doi:10.1021/jacs.1c07497 · Cited by 35
A Coupled Spintronics Neuromorphic Approach for High‐Performance Reservoir Computing
Akashi, N., Kuniyoshi, Y., Tsunegi, S., Taniguchi, T., Nishida, M., Sakurai, R., Wakao, Y., Kawashima, K., & Nakajima, K. (2022) · Advanced Intelligent Systems · doi:10.1002/aisy.202200123 · Cited by 37
Recent progress in optoelectronic neuromorphic devices*
Guo, Y., & Zhu, L. Q. (2020) · Chinese Physics B · doi:10.1088/1674-1056/ab99b6 · Cited by 35
On the role of planning in model-based deep reinforcement learningPreprint
Hamrick, J. B., Friesen, A. L., Behbahani, F., Guez, A., Viola, F., Witherspoon, S., Anthony, T., Buesing, L., Veličković, P., & Weber, T. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2011.04021 · Cited by 64
“[57] David Silver, Thomas Hubert, Julian Schrittwieser, Ioannis Antonoglou, Matthew Lai, Arthur Guez, Marc Lanctot, Laurent Sifre, Dharshan Kumaran, Thore Graepel, et al.”
BackgroundInfluentialSelf-citationMapping Citizen Science through the Lens of Human-Centered AI
Rafner, J. F., Gajdacz, M., Kragh, G., Hjorth, A., Gander, A. J., Palfi, B., Berditchevskiaia, A., Grey, F., Gal, K., Segal, A., et al. (2022) · Human Computation · doi:10.15346/hc.v9i1.133 · Cited by 41
Applications of physics-informed neural networks for property characterization of complex materials
Lee, S., & Popovics, J. S. (2023) · RILEM Technical Letters · doi:10.21809/rilemtechlett.2022.174 · Cited by 42
“Go and Chess), the performance of ANNs is already superior to human ability [15].”
BackgroundDeep-reinforcement-learning-based water diversion strategy
Jiang, Q., Li, J., Sun, Y., Huang, J., Zou, R., Ma, W., Guo, H., Wang, Z., & Liu, Y. (2023) · Environmental Science and Ecotechnology · doi:10.1016/j.ese.2023.100298 · Cited by 35
Liu, Q., Guo, Y., Deng, L., Liu, H., Li, D., Sun, H., & Huang, W. (2024) · IEEE Transactions on Sustainable Energy · doi:10.1109/tste.2024.3376369 · Cited by 40
“Recently, deep reinforcement learning (DRL) methods are intensively studied because of its remarkable performance in games [4], robotics [5], as well as power system operation [6], [7], [8], [9], [10], [11], [12].”
BackgroundCEO Analysis of Demis Hassabis of DeepMind Technologies Limited
Aithal, S., & Kumar, K. R. N. (2025) · Poornaprajna International Journal of Teaching & Research Case Studies (PIJTRCS) · doi:10.64818/pijtrcs.3107.8494.0027 · Cited by 41
A Survey of Planning and Learning in Games
Duarte, F. F., Lau, N., Pereira, A., & Reis, L. P. (2020) · Applied Sciences · doi:10.3390/app10134529 · Cited by 44
“DRL, MCTS PGRD [254], TDTS [270], AlphaGo [17], AlphaGo Zero [287], ExIt [288], TD-MCTS [291], AlphaZero [293], MuZero [294], MC2 [295], [256,261,299] DRL, A* Cerberus [303] RL, DFS [298] IRL, PF, WFP [300] RL, α-β Meep [14] ML, α-β Logistello [11,52] ML, LAS CueCard [302] RL, UCT [262] RL, Minimax [286]”
MethodInfluentialBeating the Stock Market with a Deep Reinforcement Learning Day Trading System
Martinez, L. C., & Pereira, A. C. M. (2020) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn48605.2020.9206938 · Cited by 38
Imai, K., Jiang, Z., Greiner, D. J., Halen, R., & Shin, S. (2023) · Journal of the Royal Statistical Society Series A (Statistics in Society) · doi:10.1093/jrsssa/qnad010 · Cited by 83
“Although some of this evidence is decades old (e.g., Dawes et al., 1989), it has recently gained significant public attention for the spectacular defeats of humanity’s best in cerebral games (e.g., Silver et al., 2018).”
BackgroundZhang, C., Yao, M., Chen, W., Zhang, S., Chen, D., & Wu, Y. (2021) · Security and Communication Networks · doi:10.1155/2021/9956773 · Cited by 39
“)ese network models show superperformance in many professional fields like language translation [1], visual recognition [2], and decision-making [3].”
BackgroundRetractedHow to compete with robots by assessing job automation risks and resilient alternatives
Paolillo, A., Colella, F., Nosengo, N., Schiano, F., Stewart, W., Zambrano, D., Chappuis, I., Lalive, R., & Floreano, D. (2022) · Science Robotics · doi:10.1126/scirobotics.abg5561 · Cited by 49
Zhang, G., Zhang, C., Wang, W., Cao, H., Chen, Z., & Niu, Y. (2022) · Energy · doi:10.1016/j.energy.2022.126485 · Cited by 34
Performance comparison of reinforcement learning and metaheuristics for factory layout planning
Klar, M., Glatt, M. F., & Aurich, J. C. (2023) · CIRP journal of manufacturing science and technology · doi:10.1016/j.cirpj.2023.05.008 · Cited by 39
Evolutionary implementation of Bayesian computationsPreprint
Czégel, D., Giaffar, H., Zachar, I., Tenenbaum, J. B., & Szathmáry, E. (2019) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/685842 · Cited by 33
“…the theoretical development of this machinery: i) the machine learning revolution building on the boom of available data and computational power [4, 5] and ii) the possibility to perform large-scale (human and non-human) cognitive experiments that necessitates a modeling framework accounting for…”
BackgroundFinding Friend and Foe in Multi-Agent Games
Serrino, J., Kleiman-Weiner, M., Parkes, D. C., & Tenenbaum, J. B. (2019) · arXiv · doi:10.48550/arxiv.1906.02330 · Cited by 33
“Related work DeepRole builds on the recent success of heuristic search techniques that combine efficient depth-limited lookahead planning with a value function learned through self-play in twoplayer zero-sum games [27, 33, 34].”
BackgroundSocial Diversity and Social Preferences in Mixed-Motive Reinforcement Learning
McKee, K. R., Gemp, I., McWilliams, B., Duèñez-Guzmán, E. A., Hughes, E., & Leibo, J. Z. (2020) · International Joint Conference on Autonomous Agents and Multiagent Systems · doi:10.65109/dnuf7773 · Cited by 33
“In pure-conflict reinforcement learning, self-play solutions for Markov games [5, 22, 49] have gradually given way to populationbased approaches [26, 55] (Figure 1a).”
BackgroundChen, J. H., Dhaliwal, G., & Yang, D. (2022) · JAMA · doi:10.1001/jama.2022.13735 · Cited by 32
Agent‐based models of inflammation in translational systems biology: A decade later
Vodovotz, Y., & An, G. C. (2019) · WIREs Systems Biology and Medicine · doi:10.1002/wsbm.1460 · Cited by 47
“…specific type termed “ model-based DRL. ” This is the approach that has led to the recent advances in game-playing artificial intelligence agents (AIs) from Google's Deep Mind project, AlphaGo and AlphaZero, function by learning on simulated game played between two AIs (Silver et al., 2016, 2018).”
BackgroundPartitioned integrators for thermodynamic parameterization of neural networks
Leimkuhler, B., Matthews, C., & Vlaar, T. (2019) · Foundations of Data Science · doi:10.3934/fods.2019019 · Cited by 54
“al structure [43], medical decision making strategies for palliative care [1] and Alpha Zero which is able to master a complex challenge, e.g., learning to play Go or chess, in the span of a few days [44]. Yet there remain a number of mysteries regarding the performance of neural networks, their generality, and their…”
BackgroundFu, Y., Xu, S., Zhu, Q., O’Neill, Z. D., & Adetola, V. (2023) · Energy · doi:10.1016/j.energy.2023.127073 · Cited by 43
Li, J., Gu, X., Qiu, S., Zhou, X., Cangelosi, A., Loo, C. K., & Liu, X. (2024) · IEEE Transactions on Systems Man and Cybernetics Systems · doi:10.1109/tsmc.2024.3369071 · Cited by 41
“DRL is famous for the AlphaGo program developed by Google DeepMind [159], and it has been widely used in the human–computer cooperation planning of lower extremity exoskeleton robots at present.”
MethodMachine Learning Algorithm Guides Catalyst Choices for Magnesium‐Catalyzed Asymmetric Reactions
Baczewska, P., Kulczykowski, M., Zambroń, B. K., Jaszczewska‐Adamczak, J. A., Pakulski, Z., Roszak, R., Grzybowski, B. A., & Młynarski, J. (2024) · Angewandte Chemie International Edition · doi:10.1002/anie.202318487 · Cited by 37
Quantum circuit optimization with AlphaTensor
Ruiz, F. J. R., Laakkonen, T., Bausch, J., Balog, M., Barekatain, M., Heras, F. J. H., Novikov, A., Fitzpatrick, N., Romera‐Paredes, B., van de Wetering, J., et al. (2025) · Nature Machine Intelligence · doi:10.1038/s42256-025-01001-1 · Cited by 128
“In fault-tolerant quantum computation we cannot use a continuous family of gates like 𝑍 [ 𝛼 ] , and instead A particularly often used one is the T gate, 𝑇 ≜ 𝑍 [ ] .”
Redirection Controller Using Reinforcement Learning
Chang, Y., Matsumoto, K., Narumi, T., Tanikawa, T., & Hirose, M. (2021) · IEEE Access · doi:10.1109/access.2021.3118056 · Cited by 33
“Meanwhile, recent studies have encouraged the possibility of applying reinforcement learning (RL) to many fields, such as board games [7], robotics [8], and autonomous car [9],”
BackgroundDeep Q-learning decoder for depolarizing noise on the toric code
Fitzek, D., Eliasson, M., Kockum, A. F., & Granath, M. (2020) · Physical Review Research · doi:10.1103/physrevresearch.2.023230 · Cited by 59
Structure prediction of surface reconstructions by deep reinforcement learning
Meldgaard, S. A., Mortensen, H. L., Jo̷rgensen, M., & Hammer, B. (2020) · Journal of Physics Condensed Matter · doi:10.1088/1361-648x/ab94f2 · Cited by 34
Learning to Plan with Uncertain Topological Maps
Beeching, E., Dibangoye, J., Simonin, O., & Wolf, C. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-58580-8_28 · Cited by 58
Advances in neural networks and potential for their application to steel metallurgy
Smith, J. L. (2020) · Materials Science and Technology · doi:10.1080/02670836.2020.1839206 · Cited by 32
“This has been demonstrated many times in the ability of RL models to master games (i.e. perform better than the top human players) such as Atari [95], Chess [96], Go [97], and Starcraft II [98].”
BackgroundComparing driving behavior of humans and autonomous driving in a professional racing simulator
Remonda, A., Veas, E. E., & Luzhnica, G. (2021) · PLoS ONE · doi:10.1371/journal.pone.0245320 · Cited by 34
Deep reinforcement learning with credit assignment for combinatorial optimization
Yan, D., Weng, J., Huang, S., Li, C., Zhou, Y., Su, H., & Zhu, J. (2021) · Pattern Recognition · doi:10.1016/j.patcog.2021.108466 · Cited by 37
A deep learning guided memetic framework for graph coloring problems
Goudet, O., Grelier, C., & Hao, J. (2022) · Knowledge-Based Systems · doi:10.1016/j.knosys.2022.109986 · Cited by 42
“These attempts have been given a new lease of life, with the emergence of deep learning techniques for combinatorial optimization problems [7, 46], inspired from the great success of the AlphaZero al-gorithm for combinatorial games [40].”
Behaviour Suite for Reinforcement Learning
Osband, I., Doron, Y., Hessel, M., Aslanides, J., Sezener, E., Saraiva, A., McKinney, K., Lattimore, T., Szepesvári, C., Singh, S. P., et al. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1908.03568 · Cited by 31
“We have the existence proof that a single self-learning RL agent can master the game of Go purely from self-play (Silver et al., 2018).”
BackgroundInfluentialSelf-citationExploring Large Language Models for Communication Games: An Empirical Study on Werewolf
Xu, Y., Wang, S., Li, P., Luo, F., Wang, X., Liu, W., & Liu, Y. (2023) · arXiv · doi:10.48550/arxiv.2309.04658 · Cited by 32
Chen, Q., Kuang, Z., Liu, X., et al. (2024) · Applied Energy · doi:10.1016/j.apenergy.2024.123163 · Cited by 31
Chew, A. W. Z., He, R., & Zhang, L. (2024) · Archives of Computational Methods in Engineering · doi:10.1007/s11831-024-10145-z · Cited by 45
“Even for the state-of-the-art RL algorithms (e.g., OpenAI Five, AlphaZero), they are required to be trained for many days before being competent and effective in their respective game plays [212, 213].”
Enhancing the interoperability between deep learning frameworks by model conversion
Liu, Y., Chen, C., Zhang, R., Qin, T., Ji, X. L., Lin, H., & Yang, M. (2020) · ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering (ESEC/FSE) · doi:10.1145/3368089.3417051 · Cited by 52
The automaton as a surgeon: the future of artificial intelligence in emergency and general surgery
Rimmer, L. J., Howard, C., Picca, L., & Bashir, M. (2020) · European Journal of Trauma and Emergency Surgery · doi:10.1007/s00068-020-01444-8 · Cited by 36
Curriculum learning with Hindsight Experience Replay for sequential object manipulation tasks
Manela, B., & Biess, A. (2021) · Neural Networks · doi:10.1016/j.neunet.2021.10.011 · Cited by 37
“…to many breakthroughs in recent years for generating goal-directed behavior in artificial agents, ranging from playing Atari games with superhuman capabilities [3] to making animated figures walk [4, 5, 6], solving complex manipulation tasks [7, 8], and defeating the best GO player in the world [9].”
BackgroundForecasting Hamiltonian dynamics without canonical coordinates
Choudhary, A., Lindner, J. F., Holliday, E. G., Miller, S. T., Sinha, S., & Ditto, W. L. (2021) · Nonlinear Dynamics · doi:10.1007/s11071-020-06185-2 · Cited by 31
WD3: Taming the Estimation Bias in Deep Reinforcement Learning
He, Q., & Hou, X. (2020) · IEEE International Conference on Tools with Artificial Intelligence (ICTAI) · doi:10.1109/ictai50040.2020.00068 · Cited by 38
“Combined with deep learning, a lot of achievements have been produced in a wide range of field such as playing Atari games [1], playing chess, Go and shoji [4], beating human players in StarCraft [2], controlling robotic manipulation [16], playing Rubik’s cube [3] and so on.”
BackgroundConstrained Multiagent Markov Decision Processes: a Taxonomy of Problems and Algorithms
de Nijs, F., Walraven, E., Weerdt, M. M. D., & Spaan, M. T. J. (2021) · Journal of Artificial Intelligence Research · doi:10.1613/jair.1.12233 · Cited by 55
“The recent arrival of Artificial Intelligence (AI) as a household term comes from the success of well-known examples like digital personal assistants (Hoy, 2018), media recommender systems (Möller et al., 2018), and game-playing AI capable of surpassing human experts (Silver et al., 2018).”
BackgroundLearning equilibria in symmetric auction games using artificial neural networks
Bichler, M., Fichtl, M., Heidekrüger, S., Kohring, N., & Sutterer, P. (2021) · Nature Machine Intelligence · doi:10.1038/s42256-021-00365-4 · Cited by 77
Chen, Y., Bai, J., & Kang, Y. (2021) · IEEE Journal of Emerging and Selected Topics in Power Electronics · doi:10.1109/jestpe.2021.3128270 · Cited by 32
“The proposed idea is inspired by the AlphaGo [48] and AlphaZero [49], whose procedure is shown in Fig.”
MethodHierarchical deep reinforcement learning reveals a modular mechanism of cell movement
Wang, Z., Xu, Y., Wang, D., Yang, J., & Bao, Z. (2022) · Nature Machine Intelligence · doi:10.1038/s42256-021-00431-x · Cited by 34
“…reinforcement learning (DRL), which formulates the dynamic decision-making problem with a Markov decision process (MDP), has been highly successful in solving dynamic, global optimization problems such as game-playing 7–10 and robotic manipulation 11–13 , achieving or surpassing human performance.”
BackgroundOn Creativity, Music’s AI Completeness, and Four Challenges for Artificial Musical Creativity
Rohrmeier, M. (2022) · Transactions of the International Society for Music Information Retrieval · doi:10.5334/tismir.104 · Cited by 47
What the Near Future of Artificial Intelligence Could Be
Floridi, L. (2020) · Digital ethics lab yearbook · doi:10.1007/978-3-030-29145-7_9 · Cited by 37
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.