reinforcement learning
In plain wordsA way of teaching a computer program by trial and error: it tries actions, earns points for good results and penalties for bad ones, and gradually learns which choices pay off over time.
A family of machine learningmachine learning, n.: A way of building software that learns from examples instead of following rules written by a programmer. Shown thousands of labeled photos, for instance, a program can learn to recognize cats. problems and methods for learning behavior from its consequences: an agentagent, n.: Something that acts. In AI, a program that observes its situation and takes actions toward a goal: a game-playing learner, a software helper, or a chatbot that uses tools to complete tasks. acts in an environmentenvironment, n.: Everything an AI agent interacts with but does not control — a game, a simulator, a website, the physical world. It answers each of the agent's actions with a new situation and, in reinforcement learning, a score., receives a numerical rewardreward, n.: The score an AI system receives after an action, telling it how good the outcome was. Reinforcement learning trains the system to collect as much reward as possible over time. after its actions, and gradually adjusts its policypolicy, n.: An AI agent's strategy: the rule it follows to pick an action in each situation. Training in reinforcement learning means improving this rule until it earns high rewards., its rule for choosing what to do in each situation, so as to collect as much reward as possible over the long run (the returnreturn, n.: The total reward an AI agent collects from now until the end of a task, usually counting near-term rewards more than distant ones. Agents are trained to make this total as large as possible.). Because the feedback grades what the agent did instead of naming the correct action, as labels do in supervised learningsupervised learning, n.: Teaching a computer by example: you give it many inputs paired with the right answers, such as photos labeled 'cat' or 'dog', and it learns to answer correctly for new inputs., the agent must keep testing actions whose value is still uncertain rather than only repeating those that have paid off (the exploration–exploitation trade-offexploration–exploitation trade-off, n.: The choice between trying something new that might turn out better and sticking with what already works — like picking a familiar restaurant versus testing a new one.), and must work out which earlier decisions deserve credit for outcomes that arrive late (the credit assignment problemcredit assignment problem, n.: The puzzle of figuring out which of many earlier decisions, or which parts of a system, deserve the credit or blame for a final outcome — like working out which moves won a chess game.). Problems are usually formalized as Markov decision processesMarkov decision process, n.: A mathematical description of a step-by-step decision problem: each step you pick an action, receive a reward and then land in a new situation, drawn at random from odds that depend only on the current situation and your choice.. Methods are broadly value-based (temporal difference learningtemporal difference learning, n.: A way of learning predictions by adjusting each guess toward the next, slightly better-informed guess instead of waiting for the final outcome — like revising a weather forecast every day as the date approaches., Q-learningQ-learning, n.: A trial-and-error method in which an agent keeps a score for every move in every situation and nudges each score toward the reward it got plus the best score available next.), policy-based (policy gradientpolicy gradient, n.: A family of methods that improve an AI's strategy directly: make actions that led to good results a bit more likely, and actions that led to bad results a bit less likely., actor–criticactor–critic, n.: A way of training an AI agent in which one part, the 'actor', makes the moves while another, the 'critic', learns to predict how well they will work out, and the actor improves by heeding the critic.) or model-basedmodel-based reinforcement learning, n.: Learning in which an AI agent builds, or is given, an inner simulation of how its world responds, and uses it to plan or practice in imagination before acting.; combined with deep neural networks they form deep reinforcement learningdeep reinforcement learning, n.: Reinforcement learning in which a large neural network does the learning, letting an AI learn tasks such as video games or robot control directly from raw inputs like screen pixels or camera images.. Landmark results include TD-GammonTD-Gammon, prop. n.: A backgammon program built at IBM in the early 1990s that taught itself to play at near world-champion level by playing more than a million games against itself., the Atari-playing deep Q-networkdeep Q-network, n.: A method, made famous by DeepMind in 2013–2015, in which a neural network learned to play dozens of Atari video games from screen pixels and the score alone, many at human level., AlphaGoAlphaGo, prop. n.: A DeepMind program that in 2016 beat Lee Sedol, one of the world's best players of the board game Go, a decade or more before many experts expected a computer to manage it. and robot control. Since 2022 RL has also been a central stage of post-trainingpost-training, n.: Everything done to an AI model after its initial bulk training on huge amounts of text: the extra teaching that turns a raw text predictor into a helpful, safe assistant or a careful reasoner. for language modelslarge language model, n.: A computer program trained on enormous amounts of text to predict the words that come next. At vast scale, that simple skill lets it chat, answer questions, translate, summarize and write code., first as reinforcement learning from human feedbackreinforcement learning from human feedback, n.: A way of training AI using people's judgments: humans compare pairs of answers, a scoring model learns their preferences, and the AI is then trained to produce answers that score well. and then as reinforcement learning with verifiable rewardsreinforcement learning with verifiable rewards, n.: Training an AI by letting it attempt problems whose answers can be checked automatically — a math result, a passing program — and rewarding it only when the check succeeds. for reasoning modelsreasoning model, n.: An AI model trained to work through a problem step by step in hidden 'thinking' text before it answers. Letting it think longer generally makes it better at math, coding and planning..
We moved from imitating demonstrations to reinforcement learning, rewarding the agent only when the tests pass.
Origin
The idea descends from animal-learning psychology — Edward Thorndike's law of effect and the 'reinforcement' of conditioning research — and from early computing: Alan Turing's 1948 proposal of 'pleasure–pain' training, Marvin Minsky's 1954 Princeton thesis on neural-analog reinforcement systems, and Arthur Samuel's checkers program (1959). The phrase entered engineering literature in the mid-1960s (Waltz and Fu, 1965; Mendel, 1966). Andrew Barto and Richard Sutton recast the field from the 1980s, wrote its standard textbook (1998) and received the 2024 ACM Turing Award.
Usage note
The field borrowed its vocabulary from behavioral psychology, but 'reward' here is just a number chosen by a designer, and an agent optimizes whatever that number measures — the root of reward hackingreward hacking, n.: When an AI finds a shortcut that earns the reward or score it is trained on without doing the task its designers wanted — like a student gaming the grading rubric.. Reinforcement learning from human feedbackreinforcement learning from human feedback, n.: A way of training AI using people's judgments: humans compare pairs of answers, a scoring model learns their preferences, and the AI is then trained to produce answers that score well. is one application of RL, not a synonym for it.
Related
- Markov decision processMarkov decision process, n.: A mathematical description of a step-by-step decision problem: each step you pick an action, receive a reward and then land in a new situation, drawn at random from odds that depend only on the current situation and your choice.
- policypolicy, n.: An AI agent's strategy: the rule it follows to pick an action in each situation. Training in reinforcement learning means improving this rule until it earns high rewards.
- rewardreward, n.: The score an AI system receives after an action, telling it how good the outcome was. Reinforcement learning trains the system to collect as much reward as possible over time.
- value functionvalue function, n.: An estimate of how promising a situation is for an AI agent, judged by the rewards it is likely to collect from that point on, the way a chess player sizes up a board.
- deep reinforcement learningdeep reinforcement learning, n.: Reinforcement learning in which a large neural network does the learning, letting an AI learn tasks such as video games or robot control directly from raw inputs like screen pixels or camera images.
- temporal difference learningtemporal difference learning, n.: A way of learning predictions by adjusting each guess toward the next, slightly better-informed guess instead of waiting for the final outcome — like revising a weather forecast every day as the date approaches.
- policy gradientpolicy gradient, n.: A family of methods that improve an AI's strategy directly: make actions that led to good results a bit more likely, and actions that led to bad results a bit less likely.
- reinforcement learning from human feedbackreinforcement learning from human feedback, n.: A way of training AI using people's judgments: humans compare pairs of answers, a scoring model learns their preferences, and the AI is then trained to produce answers that score well.
- supervised learningsupervised learning, n.: Teaching a computer by example: you give it many inputs paired with the right answers, such as photos labeled 'cat' or 'dog', and it learns to answer correctly for new inputs.
- unsupervised learningunsupervised learning, n.: Machine learning without answer keys: the computer looks at unlabeled data and finds structure on its own, such as groups of similar customers or the main ways in which data vary.
- imitation learningimitation learning, n.: Teaching an AI by showing it examples of an expert doing a task, so it learns to copy the behavior instead of discovering it through trial and error.
Sources
- Sutton, R. S. & Barto, A. G. (2018). Reinforcement Learning: An Introduction (2nd ed.). MIT Press.
- ACM. 2024 ACM A.M. Turing Award: Andrew G. Barto and Richard S. Sutton. awards.acm.org/about/2024-turing
- Waltz, M. D. & Fu, K. S. (1965). A heuristic approach to reinforcement learning control systems. IEEE Transactions on Automatic Control 10(4), 390–398.
Mentioned in
adversarial exampleadversarial example, n.: An input altered on purpose, often too slightly for people to notice, so that an AI model gets it badly wrong — like a panda photo that a classifier confidently calls a gibbon., AlphaGoAlphaGo, prop. n.: A DeepMind program that in 2016 beat Lee Sedol, one of the world's best players of the board game Go, a decade or more before many experts expected a computer to manage it., artificial intelligenceartificial intelligence, n.: Computer systems that do things we would normally credit to a thinking mind, such as holding a conversation, reading an X-ray or planning a route, and the field of research that builds them., deep researchdeep research, n.: An AI feature that spends minutes rather than seconds on a question — planning, running dozens of web searches, reading many sources — and then writes a long report with citations., DeepSeekDeepSeek, prop. n.: A Chinese artificial intelligence company whose free, cheaply trained models startled the world in January 2025 by nearly matching the best American systems., loss functionloss function, n.: A score that measures how wrong a model's predictions are. Training means adjusting the model, step by step, to push this number down., machine learningmachine learning, n.: A way of building software that learns from examples instead of following rules written by a programmer. Shown thousands of labeled photos, for instance, a program can learn to recognize cats., optimizationoptimization, n.: Finding the best possible choice according to some score, such as the lowest error or the highest reward. Training an AI is optimization: adjusting its settings until its mistakes are as small as possible., reasoning modelreasoning model, n.: An AI model trained to work through a problem step by step in hidden 'thinking' text before it answers. Letting it think longer generally makes it better at math, coding and planning., reinforcement learning from human feedbackreinforcement learning from human feedback, n.: A way of training AI using people's judgments: humans compare pairs of answers, a scoring model learns their preferences, and the AI is then trained to produce answers that score well., rewardreward, n.: The score an AI system receives after an action, telling it how good the outcome was. Reinforcement learning trains the system to collect as much reward as possible over time., reward hackingreward hacking, n.: When an AI finds a shortcut that earns the reward or score it is trained on without doing the task its designers wanted — like a student gaming the grading rubric., training datatraining data, n.: The examples an AI system learns from. Whatever patterns, skills, gaps and biases are in the training data tend to show up in the finished system., accidental cyberattackaccidental cyberattack, n.: A break-in on real computer systems carried out by AI agents that were only supposed to be in training or being tested, with no person meaning to attack anyone., actor–criticactor–critic, n.: A way of training an AI agent in which one part, the 'actor', makes the moves while another, the 'critic', learns to predict how well they will work out, and the actor improves by heeding the critic., agentagent, n.: Something that acts. In AI, a program that observes its situation and takes actions toward a goal: a game-playing learner, a software helper, or a chatbot that uses tools to complete tasks., agent loopagent loop, n.: The repeating cycle at the heart of an AI agent: the model decides what to do next, software carries it out, the result is shown to the model, and the cycle repeats until the task is finished., agentic RAGagentic RAG, n.: Retrieval in which the AI runs its own research: it decides what to search for, looks at the results, searches again if needed, and only then answers., AlphaGo ZeroAlphaGo Zero, prop. n.: A program from DeepMind that taught itself to play the board game Go better than any human in 2017, starting from nothing but the rules and learning only by playing millions of games against itself., AlphaZeroAlphaZero, prop. n.: A 2017 DeepMind program that taught itself chess, shogi and Go from nothing but the rules by playing against itself, and within hours of training beat the strongest existing programs., automated red teamingautomated red teaming, n.: Using AI programs, not just human testers, to invent huge numbers of tricky inputs that try to make another AI misbehave, so problems can be found and fixed at scale., catastrophic forgettingcatastrophic forgetting, n.: When a neural network trained on something new gets suddenly and badly worse at what it could already do, because the new training rewrites the internal settings the old skills relied on., causal inferencecausal inference, n.: Working out whether one thing actually causes another, rather than merely occurring alongside it — for example, whether a new feature made users stay longer or just happened at the same time., cognitive sciencecognitive science, n.: The scientific study of how minds work — how we reason, use language, remember and see — combining psychology, neuroscience, linguistics, philosophy, anthropology and artificial intelligence.