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 2,601–2,700 of 3,633
An accelerated asynchronous advantage actor-critic algorithm applied in papermaking
Wang, X., Zhuang, Z., Zou, L., et al. (2019) · Chinese Control Conference (CCC) · doi:10.23919/chicc.2019.8866243 · Cited by 1
Influences of Neural Network Structures on an Efficient Reinforcement Learning Policy Search
Zhu, W., & Rosendo, A. (2019) · IEEE International Conference on Robotics and Biomimetics (ROBIO) · doi:10.1109/robio49542.2019.8961475 · Cited by 1
“Many games such as Go [13], Chess [16], and Atari games [17] can be solved by different reinforcement learning methods.”
BackgroundTuniz, C., & Vipraio, P. T. (2020) · Springer eBooks · doi:10.1007/978-3-030-36522-6_12 · Cited by 1
K-spin Hamiltonian for quantum-resolvable Markov decision processes
Jones, E. B., Graf, P. A., Kapit, E., & Jones, W. B. (2020) · Quantum Machine Intelligence · doi:10.1007/s42484-020-00026-6 · Cited by 3
El factor estético en la automatización de tareas lógicas: el caso del ajedrez
Rementeria, S. (2020) · Artnodes · doi:10.7238/a.v0i26.3338 · Cited by 1
Research on Topology Planning for Wireless Mesh Networks Based on Deep Reinforcement Learning
Yin, C., Yang, R., Zou, X., & Zhu, W. (2020) · International Conference on Computer Communication and the Internet (ICCCI) · doi:10.1109/iccci49374.2020.9145985 · Cited by 1
“The method of MCTS and deep reinforcement learning is used to design the residual network module based on convolution.”
MethodZhao, J., Zhou, N., Zong, H., Cheng, J., & Long, H. (2019) · International Conference on Intelligent Systems and Knowledge Engineering (ISKE) · doi:10.1109/iske47853.2019.9170373 · Cited by 1
Campo, D., Martín, D., & Marcenaro, L. (2020) · IEEE Signal Processing Magazine · doi:10.1109/msp.2020.3002482 · Cited by 1
“In recent studies [2], it has been demonstrated that intelligent systems must learn to solve different tasks in diverse environments, focusing attention on the most relevant task and environmental features for solving a problem.”
BackgroundThe Future of Society: Choices, Decisions, and Consequences
Eaton, M. (2020) · Springer eBooks · doi:10.1007/978-3-030-55300-5_14 · Cited by 1
Bishop, R. R., Outkin, A. V., Eames, B., et al. (2020) · IEEE Transactions on Computational Social Systems · doi:10.1109/tcss.2020.3035635 · Cited by 1
“Moreover, the concept of training adversarial agents using successive competition is similar to and inspired by recent work in [14] using neural networks and competitive self-play to develop policies for playing Go, shogi, and chess.”
BackgroundDesigning Policy Network with Deep Learning in Turn-Based Strategy Games
Kimura, T., & Ikeda, K. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-65883-0_12 · Cited by 1
Zhang, H., Tian, Q., Wang, D., & Wei, K. (2020) · IEEE International Conference on Tools with Artificial Intelligence (ICTAI) · doi:10.1109/ictai50040.2020.00086 · Cited by 1
“Deep reinforcement learning (DRL) has led to a big success in various domains such as AlphaGo [1], AlphaGo-Zero [2], realistic simulated environments [3], robotics [4] and games [1], [5].”
BackgroundLearning Summarised Messaging Through Mediated Differentiable Inter-Agent Learning
Gopal, S., Mathur, R., Deshwal, S., et al. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-66412-1_35 · Cited by 1
Collins, J. B., & Dasgupta, P. (2020) · Springer eBooks · doi:10.1007/978-3-030-55692-1_1 · Cited by 1
Was kann Künstliche Intelligenz?
Paaß, G., & Hecker, D. (2020) · Künstliche Intelligenz · doi:10.1007/978-3-658-30211-5_2 · Cited by 1
Paradox of AlphaZero: Strategic vs. Optimal Plays
Dou, Z.-L., Ma, L., Nguyen, K., & Nguyen, K. X. (2020) · IEEE International Performance, Computing, and Communications Conference · doi:10.1109/ipccc50635.2020.9391562 · Cited by 1
“AlphaGo [1] and its subsequent variations culminating in AlphaZero [2], notching victories against professional Go players for the very first time in history and yet achieving them with seeming ease, and eventually becoming self-taught from a near tabula rasa in the span of a mere couple of years,…”
BackgroundVirtual Reinforcement Learning for Balancing an Inverted Pendulum in Real Time.
Bates, D. (2021) · NCSU Libraries Repository (North Carolina State University Libraries) · Cited by 1
The artificial intelligence revolution: Achievements and prospects
Smolin, V. S. (2021) · Proektirovanie buduŝego. Problemy cifrovoj realʹnosti · doi:10.20948/future-2021-13 · Cited by 1
What May Lie Ahead in Reinforcement Learning
Cansever, D. H. (2021) · Studies in systems, decision and control · doi:10.1007/978-3-030-60990-0_1 · Cited by 4
Curiosity Based RL on Robot Manufacturing Cell
Hameed, M. S. A., Khan, M. M., & Schwung, A. (2021) · IEEE International Conference on Industrial Technology (ICIT) · doi:10.1109/icit46573.2021.9453577 · Cited by 2
“Reinforcement Learning (RL) is becoming a popular al-gorithm to solve complex games and tasks like Atari [1], physics simulation [2], Go [3] and Chess [4], Robotics [5], and optimization problems [6] etc.”
BackgroundA Modified Q-Learning Algorithm for Control of Two-Qubit Systems
Shindi, O., Yu, Q., Dong, D., et al. (2020) · International Conference on Machine Learning and Cybernetics · doi:10.1109/icmlc51923.2020.9469044 · Cited by 1
Carle's Game: An Open-Ended Challenge in Exploratory Machine Creativity
Davis, Q. T. (2021) · IEEE Conference on Games (CoG) · doi:10.1109/cog52621.2021.9619011 · Cited by 1
“…world champion level chess play of Deep Blue [50] and the lesser known Chinook’s prowess at checkers [35] in the 1990s, to the more recent demonstrations of the AlphaGo lineage [37]–[40] and video game players like AlphaStar for StarCraft II [52] and OpenAI Five for Dota 2 [2] to name just a few.”
BackgroundPiette, É., Stephenson, M., Soemers, D. J. N. J., & Browne, C. (2021) · IEEE Conference on Games (CoG) · doi:10.1109/cog52621.2021.9618990 · Cited by 25
“Nowadays, computer players for classical board games like Chess [7], Go [8], Shogi [9] or Checkers [10], can de-feat expert human players.”
BackgroundTowards sample-efficient policy learning with DAC-ML
Freire, I. T., Amil, A. F., Vouloutsi, V., & Verschure, P. F. M. J. (2021) · Procedia Computer Science · doi:10.1016/j.procs.2021.06.035 · Cited by 2
“Reaching human-level performance in historically considered uniquely-human intellectual domains (such as chess [1], Go [2], and real-time strategy video games [3]) has led to renewed claims about its resemblance to human cognition.”
BackgroundNo Free Lunch: Overcoming Reward Gaming in AI Safety Gridworlds
Tsvarkaleva, M., & Dennis, L. A. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-83906-2_18 · Cited by 1
Tuning the Hyperparameters of Anytime Planning: A Deep Reinforcement Learning Approach
Bhatia, A., Svegliato, J., & Zilberstein, S. (2021) · Cited by 1
Revisiting “Recurrent World Models Facilitate Policy Evolution”
Esteves, B., & Melo, F. S. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-86230-5_26 · Cited by 1
“In recent years, reinforcement learning (RL) has been the focus of increasing interest, largely due to the impressive successes in video games (Mnih et al., 2015), classical games (Silver et al., 2018) and control (Lillicrap et al., 2016), among others.”
BackgroundEquiprobable and attacked-square chess entropies by phases, levels of play and game outcomes
Islas-García, J. D. A., del Castillo-Escribano, M., & del Castillo‐Mussot, M. (2021) · International Journal of Modern Physics C · doi:10.1142/s0129183122500279 · Cited by 1
Targeted Upskilling Framework based on Player Mistake Context in Online Skill Gaming Platforms
Agrawal, S., Sachdeva, M., & Mukherjee, T. (2021) · International Conference on AI-ML Systems (AIMLSystems) · doi:10.1145/3486001.3486234 · Cited by 1
Application of Machine Learning Algorithms for Solving Problems in the Oil and Gas Sector
Kalashnikov, I. S. T. U., & Maiorov, K. N. (2021) · Intellekt Sist Proizv · doi:10.22213/2410-9304-2021-3-55-64 · Cited by 1
Learn Quasi-Stationary Distributions of Finite State Markov Chain
Cai, Z., Lin, L., & Zhou, X. (2022) · Entropy · doi:10.3390/e24010133 · Cited by 1
“We first regard the learning process of the quasi-stationary distribution as the interaction with the environment which is constructed by the property of QSD. Reinforcement learning has recently shown tremendous advancements and remarkable successes in applications (e.g. [26, 34, 30]).”
BackgroundMimicking the Human Approach in the Game of Hive
Kampert, D., Varbanescu, A.-L., Müller-Brockhausen, M., et al. (2021) · IEEE Symposium Series on Computational Intelligence (SSCI) · doi:10.1109/ssci50451.2021.9659999 · Cited by 1
Piekutowski, P. F. (2021) · Przegląd Kulturoznawczy · doi:10.4467/20843860pk.21.037.14357 · Cited by 2
Multiple Suboptimal Policies Integrated Reinforcement Learning Algorithm for Path Planning
Hu, X., Chen, A., Sijie, Z., & Yang, Z. (2022) · International Conference on Consumer Electronics and Computer Engineering (ICCECE) · doi:10.1109/iccece54139.2022.9712751 · Cited by 1
Deep Learning: Why Neural Networks Are State of the Art
Ganesh, A., & Urs, N. (2022) · Educational communications and technology: issues and innovations · doi:10.1007/978-3-030-84729-6_2 · Cited by 2
The Cost of Reinforcement Learning for Game Engines
de Goede, D., Kampert, D., & Vărbănescu, A. L. (2022) · ACM/SPEC International Conference on Performance Engineering (ICPE) · doi:10.1145/3489525.3511685 · Cited by 1
“AlphaZero, Leela Chess Zero and MuZero have shown that it is possible to achieve superhuman performance in the games of Atari, chess, and Go, by tabula rasa reinforcement learning from games of self-play [13, 18].”
BackgroundAdvancing Human Understanding with Deep Learning Go AI Engines
Egri-Nagy, A., & Törmänen, A. (2022) · International Society for the Study of Information Summit (IS4SI Summit) · doi:10.3390/proceedings2022081022 · Cited by 3
“To answer the above question we need to investigate how existing epistemological approaches can operate in the presence of non-human knowledge (e.g., the output of the AlphaGo Zero [4] and AlphaZero [5] algorithms that use no human-played games as inputs for training) and to do an empirical study of what people instin…”
MethodA Homotopic Approach to Policy Gradients for Linear Quadratic Regulators with Nonlinear Controls
Chen, C. X., & Agazzi, A. (2022) · IEEE Conference on Decision and Control (CDC) · doi:10.1109/cdc51059.2022.9992367 · Cited by 1
“This type of trade-off has been considered before, most notably in [2] and [3] where”
BackgroundKnowledge Transfer for Deep Reinforcement Agents in General Game Playing
McEwan, C., & Thielscher, M. (2022) · Lecture notes in computer science · doi:10.1007/978-3-030-97546-3_5 · Cited by 2
“UCT has been successfully combined with deep reinforcement learning in AlphaGo [16], to beat eighteen-time world champion Go player Lee Sedol, and later in AlphaZero [17], which learned to play three different board games from scratch to a su-perhuman level.”
MethodThreat Matrix: A Fast Algorithm for Human–Machine Chinese Ludo Gaming
Han, F., & Zhou, M. (2022) · Electronics · doi:10.3390/electronics11111699 · Cited by 1
Value Refinement Network (VRN)
Wöhlke, J., Schmitt, F., & van Hoof, H. (2022) · International Joint Conference on Artificial Intelligence · doi:10.24963/ijcai.2022/494 · Cited by 1
Lynn, T. F., Ottino, J. M., Lueptow, R. M., & Umbanhowar, P. B. (2022) · Chemical Engineering Science · doi:10.1016/j.ces.2022.117840 · Cited by 1
Martínez, F., & Rendón, A. (2022) · International Journal of Computer Applications Technology and Research · doi:10.7753/ijcatr1107.1005 · Cited by 1
Competing for Resources: Estimating Adversary Strategy for Effective Plan Generation
Chrpa, L., Rytíř, P., Horčı́k, R., et al. (2022) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v36i9.21205 · Cited by 1
“Deep Reinforcement Learning (DRL) has shown impressive re-sults in Starcraft (Vinyals et al. 2019) and other (adversarial) domains such as the games of Chess or Go (Silver et al. 2018).”
BackgroundDeep Reinforcement Learning for Morpion Solitaire
Doux, B., Négrevergne, B., & Cazenave, T. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-11488-5_2 · Cited by 2
“Exit has been used in the notorious Alpha Zero [13] and Wang et al. [15] applied it for Morpion Solitaire.”
Routing in Reinforcement Learning Markov Chains
Moll, M., & Weller, D. (2022) · Lecture notes in operations research · doi:10.1007/978-3-031-08623-6_60 · Cited by 1
Improved QMIXs for Multi-entity Asynchronous Cooperative Learning in Tactical Wargame
Jiang, C., Zhang, Y., Luo, J., Li, X., Li, T., & Liu, G. (2022) · Lecture notes in electrical engineering · doi:10.1007/978-981-19-6052-9_50 · Cited by 1
Towards Run-time Efficient Hierarchical Reinforcement Learning
Abramowitz, S., & Nitschke, G. (2022) · IEEE Congress on Evolutionary Computation (CEC) · doi:10.1109/cec55065.2022.9870368 · Cited by 2
Bayesian Opponent Exploitation by Inferring the Opponent’s Policy Selection Pattern
Lee, K.-T., & Wang, S. (2022) · IEEE Conference on Games (CoG) · doi:10.1109/cog51982.2022.9893611 · Cited by 2
“Reinforcement learning (RL) has been employed to solve game playing [1]–[3] in stationary environments.”
BackgroundA Study on Efficient Reinforcement Learning Through Knowledge Transfer
Glatt, R., da Silva, F. L., Bianchi, R. A. C., & Costa, A. H. R. (2022) · Adaptation, learning, and optimization · doi:10.1007/978-3-031-11748-0_14 · Cited by 7
Artificial intelligence finds faster algorithms for multiplying matrices
(2022) · Nature · doi:10.1038/d41586-022-03023-w · Cited by 2
Report on the 2nd Digital Ludeme Project Workshop
Browne, C., Piette, É., Crist, W., et al. (2022) · ICGA Journal · doi:10.3233/icg-220211 · Cited by 1
“In contrast to the well-known AlphaZero (Silver et al. (2018)) line of work, this program uses a best-first variant of minimax search, rather than MCTS (Kocsis and Szepesvári (2006); Coulom (2007); Browne et al. (2012)), as the search engine.”
MethodKokel, H., Natarajan, S., Ravindran, B., & Tadepalli, P. V. (2022) · Neural Computing and Applications · doi:10.1007/s00521-022-08119-y · Cited by 8
“Here explicit dynamic models of actions are learned via exploration and used either offline to compute approximately optimal policies [4, 16] or online in look-ahead search [43].”
MethodNavaneeth, A. V., & Dileep, M. R. (2022) · Lecture notes in electrical engineering · doi:10.1007/978-981-19-5482-5_12 · Cited by 1
Neural Network Heuristic Functions: Taking Confidence into Account
Heller, D., Ferber, P., Bitterwolf, J., Hein, M., & Hoffmann, J. (2022) · Proceedings of the International Symposium on Combinatorial Search · doi:10.1609/socs.v15i1.21771 · Cited by 5
“Successes include the AlphaGo series (Silver et al. 2016, 2017, 2018), as well as heuristic search for single-agent games such as Rubik’s Cube (Agostinelli et al. 2019).”
BackgroundBeyond Value: CheckList for Testing Inferences in Planning-Based RL
Lam, K.-H., Tabatabai, D., Irvine, J., Bertucci, D., Ruangrotsakun, A., Kahng, M., & Fern, A. (2022) · Proceedings of the International Conference on Automated Planning and Scheduling · doi:10.1609/icaps.v32i1.19848 · Cited by 1
Glyph-Based Visual Analysis of Q-Learning Based Action Policy Ensembles on Racetrack
Groß, D., Klauck, M., Gros, T. P., Steinmetz, M., Hoffmann, J., & Gumhold, S. (2022) · International Conference on Information Visualisation (IV) · doi:10.1109/iv56949.2022.00011 · Cited by 2
“Deep Q-learning has been used successfully on several decision-making problems [3], [4], [8].”
BackgroundMethodA Relative Value Function Based Learning Beam Search for the Longest Common Subsequence Problem
Huber, M., & Raidl, G. R. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-25312-6_10 · Cited by 1
“The overall approach is inspired by the way Silver et al. [7] mastered chess, shogi, and Go with AlphaZero.”
MethodOptimizing Heterogeneous Platform Allocation Using Reinforcement Learning
Brumwell, X., Kitchen, S., & Zulch, P. A. (2023) · IEEE Aerospace Conference (AERO) · doi:10.1109/aero55745.2023.10115631 · Cited by 1
“The field has made significant strides in the last decade, particularly in the subfield of deep reinforcement learning (DRL) [1, 2].”
BackgroundOpponent Exploitation Based on Bayesian Strategy Inference and Policy Tracking
Lee, K.-T., Huang, Y.-Y., Yang, J.-R., et al. (2023) · IEEE Transactions on Games · doi:10.1109/tg.2023.3285031 · Cited by 1
“R EINFORCEMENT learning has been successfully employed to solve various kinds of decision-making problems [1], [2], [3], such as game playing [6], [7], [8], robotics [4], and recommendation [5].”
MethodWalsh, T. (2019) · Journal and proceedings of the Royal Society of New South Wales · doi:10.5962/p.361855 · Cited by 1
Augmentative Topology Agents For Open-Ended Learning
Nasir, M. U., Beukman, M., James, S. D., & Cleghorn, C. W. (2023) · Genetic and Evolutionary Computation Conference Companion (GECCO Companion) · doi:10.1145/3583133.3590576 · Cited by 4
“Machine learning has successfully been used to solve numerous problems, such as classifying images [1], writing news articles [2, 3] or solving games like Atari [4] or chess [5].”
MethodKyoung, D., & Sung, Y. (2023) · Sensors · doi:10.3390/s23177411 · Cited by 2
SAGE: Generating Symbolic Goals for Myopic Models in Deep Reinforcement Learning
Chester, A., Dann, M., Zambetta, F., & Thangarajah, J. (2023) · Lecture notes in computer science · doi:10.1007/978-981-99-8391-9_22 · Cited by 2
Perspective Chapter: Deep Learning Misconduct and How Conscious Learning Avoids It
Weng, J. (2023) · Artificial intelligence · doi:10.5772/intechopen.113359 · Cited by 1
Artificial Intelligence and Quantum Computing
Wirtz, B. W. (2024) · Springer texts in business and economics · doi:10.1007/978-3-031-50289-7_8 · Cited by 1
Wang, M., Hershey, P. C., & Ioakimidis, T. E. (2019) · doi:10.1109/datacom.2019.00019 · Cited by 1
“Their results demonstrate that a general-purpose RL algorithm can learn, tabula rasa – without domain-specific human knowledge or data, as evidenced by the same algorithm succeeding in multiple challenging games [5 – 7].”
BackgroundTheory of Cognitive Relativity: A Promising Paradigm for True AIPreprint
Li, Y. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1812.00136 · Cited by 2
“In contrast, weak AI claims that a computer would not necessarily have a mind and consciousness, even it is a super-intelligent machine like AlphaGo Zero (Silver et al. 2017; 2018).”
BackgroundDecision-Making in Reinforcement LearningPreprint
Rehman, A. J., & Tomar, P. (2019) · arXiv (Cornell University) · doi:10.13140/rg.2.2.12367.33443 · Cited by 3
Synthetic Genetic Codes Designed to Hinder EvolutionPreprint
Calles, J., Justice, I., Brinkley, D., Garcia, A., & Endy, D. (2019) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/695569 · Cited by 1
Knowledge reuse for deep reinforcement learning.
Glatt, R. (2019) · Digital Library of Theses and Dissertations (Universidade de São Paulo) · doi:10.11606/t.3.2019.tde-18092019-074805 · Cited by 2
Foundations of human problem solvingPreprint
Zarr, N., & Brown, J. W. (2019) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/779322 · Cited by 1
“Many variants of reinforcement learning solve a specific case of this problem, in which the rewarded states are relatively fixed, such as winning a game of Go (25).”
BackgroundFast Task-Adaptation for Tasks Labeled Using Natural Language in Reinforcement LearningPreprint
Hutsebaut-Buysse, M., Mets, K., & Latré, S. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1910.04040 · Cited by 5
“RL has achieved tremendous successes, especially in beating humans in games (Silver et al., 2018; Jaderberg et al., 2019) and robotics (Levine et al.”
BackgroundColosseumRL: A Framework for Multiagent Reinforcement Learning in $N$-Player GamesPreprint
Shmakov, A., Lanier, J. B., McAleer, S., Achar, R., Lopes, C. V., & Baldi, P. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1912.04451 · Cited by 1
“ing humans in 2-player zero-sum games such as Go (Silver et al. 2018), Starcraft (Vinyals et al.”
BackgroundBiologically Inspired Visual System Architecture for Object Recognition in Autonomous SystemsPreprint
Malowany, D., & Guterman, H. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2002.03472 · Cited by 1
“In addition, recent successes of deep reinforcement learning (RL) techniques have raised widespread interest in their potential for solving practical problems, with most famous cases being in the field of autonomous driving [29] and game playing [30,31].”
BackgroundEgoMap: Projective mapping and structured egocentric memory for Deep RLPreprint
Beeching, E., Wolf, C., Dibangoye, J., & Simonin, O. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2002.02286 · Cited by 1
Latent Variable Machine Learning Algorithms: Applications in a Nuclear Physics Experiment
Solli, R. (2019) · NORA - Norwegian Open Research Archives · Cited by 1
Mininni, C. J., & Zanutto, B. S. (2020) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2020.04.20.051565 · Cited by 1
FORECASTER: A Continual Lifelong Learning Approach to Improve Hardware EfficiencyPreprint
Nguyen, P. T., Kansal, A., Taur, A., Zahran, M., & Muzahid, A. (2020) · arXiv (Cornell University) · Cited by 1
Off-the-shelf deep learning is not enough: parsimony, Bayes and causalityPreprint
Vasudevan, R. K., Ziatdinov, M. A., Vlček, L., & Kalinin, S. V. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2005.01557 · Cited by 1
“8,11,12 Generally, ML methods provide a universal and extremely powerful framework for analysis of physical problems when the causal chain is clearly known.”
MethodSingle-Agent Optimization Through Policy Iteration Using Monte-Carlo Tree SearchPreprint
Seify, A., & Buro, M. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2005.11335 · Cited by 2
“One of the major contributions is the work on AlphaZero (Silver et al. 2018), which reached superhuman performance in the games of Go, Chess, and Shogi.”
BackgroundMethodInfluentialManipulating the Distributions of Experience used for Self-Play Learning in Expert IterationPreprint
Soemers, D. J. N. J., Piette, É., Stephenson, M., et al. (2020) · arXiv (Cornell University) · Cited by 1
Indexed Metrics for Link Prediction in Graph Analytics
Lim, M., Abdullah, A. B., Jhanjhi, N. Z., et al. (2020) · International Journal of Advanced Computer Science and Applications · doi:10.14569/ijacsa.2020.0110525 · Cited by 1
“With only the basic rules of the games provided, AlphaGo Zero succeeded in mastering different two-player games with complete information such as Go, Shogi and Chess, using its self-learning algorithm to compete with different versions of itself [26].”
MethodOn the Expressivity of Neural Networks for Deep Reinforcement LearningPreprint
Dong, K., Luo, Y., & Ma, T. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1910.05927 · Cited by 3
Adaptation and Deception in Adversarial Cyber Operations
Cybenko, G. V. (2020) · Wiley eBooks · doi:10.1002/9781119593386.ch5 · Cited by 3
Evaluating the Apperception EnginePreprint
Evans, R., Hernández‐Orallo, J., Welbl, J., Kohli, P., & Sergot, M. J. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2007.05367 · Cited by 10
“David Silver et al. [20] have shown that a model-based Monte Carlo tree search planner with policy distillation can achieve superhuman level performance in a number of board games.”
BackgroundUnsupervised Learning of Solutions to Differential Equations with Generative Adversarial NetworksPreprint
Randle, D., Protopapas, P., & Sondak, D. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2007.11133 · Cited by 8
“While traditional methods for solving differential equations perform well, and the theory for their stability and convergence are well established, the recent success of deep learning techniques [19, 38, 2, 39, 30, 5, 8, 34] has inspired researchers to apply neural networks to solving differential equations [33, 10, 3…”
BackgroundImproving Multi-Agent Cooperation using Theory of MindPreprint
Lim, T. X., Tio, S., & Ong, D. C. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2007.15703 · Cited by 17
“rks (and better search methods) to learn complex non-linear approximations of the “value” of states and actions have enabled deep RL agents to tackle increasingly complex games in recent years (e.g., Silver et al., 2018; Carroll et al., 2019). However, while human players also learn by trial-and-error, they learn much…”
MethodTracking the Race Between Deep Reinforcement Learning and Imitation Learning -- Extended VersionPreprint
Gros, T. P., Höller, D., Hoffmann, J., & Wolf, V. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2008.00766 · Cited by 3
Efficient Competitive Self-Play Policy OptimizationPreprint
Zhong, Y., Zhou, Y., & Peng, J. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2009.06086 · Cited by 2
“For example, AlphaGo [1] always competes with the latest agent, while the later generation AlphaGo Zero [2] and AlphaZero [3] generate self-play data with the maintained best historical agent.”
BackgroundMethodQuantum Search with Prior KnowledgePreprint
He, X., Zhang, J., & Sun, X. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2009.08721 · Cited by 1
Dream and Search to Control: Latent Space Planning for Continuous ControlPreprint
Koul, A., Kumar, V. V., Fern, A., & Majumdar, S. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2010.09832 · Cited by 6
“This is done by computing upper confidence bound (UCB) score (Rosin, 2011; Silver et al., 2018) for each child action(edge) using the stored internal statistics.”
MethodDeepRepair: Style-Guided Repairing for DNNs in the Real-world Operational EnvironmentPreprint
Yu, B., Qi, H., Guo, Q., Juefei-Xu, F., Xie, X., Ma, L., & Zhao, J. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2011.09884 · Cited by 1
“One of the most milestone examples is what we have witnessed in the game of Go where AlphaGo [51] and subsequent AlphaGo Zero [52] as well as AlphaZero [53] dominate the complicated game.”
BackgroundJob Transitions in a Time of Automation and Labor Market CrisesPreprint
Dawson, N., Rizoiu, M.-A., & Williams, M. (2020) · arXiv (Cornell University) · Cited by 1
On limitations of learning algorithms in competitive environmentsPreprint
Klimenko, A. Y., & Klimenko, D. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2011.12728 · Cited by 1
Assessing and Accelerating Coverage in Deep Reinforcement LearningPreprint
Kusari, A. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2012.00724 · Cited by 1
Learning to Stop: Dynamic Simulation Monte-Carlo Tree SearchPreprint
Lan, L.-C., Tsai, M.-Y., Wu, T.-R., Wu, I., & Hsieh, C. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2012.07910 · Cited by 1
“After fine-tuning the thresholds on NoGo, DS-MCTS can speed up more than 2.5x and still maintain a 46.5% win rate under a much larger Nmax.”
BackgroundInfluentialKégl, B., Hurtado, G., & Thomas, A. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2107.11587 · Cited by 7
“This is arguably one of the main reasons why recent beautiful advances in deep reinforcement learning (RL) (Silver et al., 2018; Vinyals et al., 2019; Badia et al., 2020) stay mostly in the realm of simulated worlds and do not immediately translate to practical success in the real world.”
BackgroundThe Intersection of Planning and Learning
Moerland, T. M. (2021) · Research Repository (Delft University of Technology) · doi:10.4233/uuid:5437884e-0078-4b36-b2c7-c6edfea3b418 · Cited by 3
Anomalous diffusion : from life to machines
Gil, G. M. (2020) · UPCommons institutional repository (Universitat Politècnica de Catalunya) · doi:10.5821/dissertation-2117-335421 · Cited by 1
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.