PowerPlay: Training an Increasingly General Problem Solver by Continually Searching for the Simplest Still Unsolvable Problem
Open accessThe DOI matches its Crossref record; 13 of the 13 references it cites were found in a registry or an index.
Abstract
Most of computer science focuses on automatically solving given computational problems. I focus on automatically inventing or discovering problems in a way inspired by the playful behavior of animals and humans, to train a more and more general problem solver from scratch in an unsupervised fashion. Consider the infinite set of all computable descriptions of tasks with possibly computable solutions. Given a general problem-solving architecture, at any given time, the novel algorithmic framework PowerPlay (Schmidhuber, 2011) searches the space of possible pairs of new tasks and modifications of the current problem solver, until it finds a more powerful problem solver that provably solves all previously learned tasks plus the new one, while the unmodified predecessor does not. Newly invented tasks may require to achieve a wow-effect by making previously learned skills more efficient such that they require less time and space. New skills may (partially) re-use previously learned skills. The greedy search of typical PowerPlay variants uses time-optimal program search to order candidate pairs of tasks and solver modifications by their conditional computational (time and space) complexity, given the stored experience so far. The new task and its corresponding task-solving skill are those first found and validated. This biases the search toward pairs that can be described compactly and validated quickly. The computational costs of validating new tasks need not grow with task repertoire size. Standard problem solver architectures of personal computers or neural networks tend to generalize by solving numerous tasks outside the self-invented training set; PowerPlay's ongoing search for novelty keeps breaking the generalization abilities of its present solver. This is related to Gödel's sequence of increasingly powerful formal theories based on adding formerly unprovable statements to the axioms without affecting previously provable theorems. The continually increasing repertoire of problem-solving procedures can be exploited by a parallel search for solutions to additional externally posed tasks. PowerPlay may be viewed as a greedy but practical implementation of basic principles of creativity (Schmidhuber, 2006a, 2010). A first experimental analysis can be found in separate papers (Srivastava et al., 2012a,b, 2013).
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Hochreiter, S., & Schmidhuber, J. (1997) · Neural Computation · doi:10.1162/neco.1997.9.8.1735 · Cited by 102,318
Self-citationReinforcement Learning: A Survey
Kaelbling, L. P., Littman, M. L., & Moore, A. W. (1996) · Journal of Artificial Intelligence Research · doi:10.1613/jair.301 · Cited by 8,972
On computable numbers, with an application to the entscheidungsproblem
Turing, A. M. (1937) · Proceedings of the London Mathematical Society · doi:10.1112/plms/s2-42.1.230 · Cited by 8,283
Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme I
Gödel, K. (1931) · Monatshefte für Mathematik und Physik · doi:10.1007/bf01700692 · Cited by 2,395
A formal theory of inductive inference. Part II
Solomonoff, R. J. (1964) · Information and Control · doi:10.1016/s0019-9958(64)90131-7 · Cited by 1,758 · cited 1×
“s effective size. The compression drive tends to improve generalization performance, according to the principles of Occam’s Razor and Minimum Description Length (MDL) and Minimum Message Length (MML) [50, 16, 57, 58, 51, 26, 18, 14]. On the other hand, the system also continually tries to invent new tasks that break t…”
MethodFormal theory of creativity, fun, and intrinsic motivation (1990–2010)
Schmidhuber, J. (2010) · IEEE Transactions on Autonomous Mental Development · doi:10.1109/tamd.2010.2056368 · Cited by 747
Self-citationUniversal Artificial Intellegence
Hutter, M. (2005) · Texts in Theoretical Computer Science An EATCS Series · doi:10.1007/b138233 · Cited by 523
Learning to Control Fast-Weight Memories: An Alternative to Dynamic Recurrent Networks
Schmidhuber, J. (1992) · Neural Computation · doi:10.1162/neco.1992.4.1.131 · Cited by 357 · cited 2×
“P may itself be the set of weight matrices of a separate RNN called RNN2, computing tasks for RNN1, and modifications of RNN1, using techniques for network-modifying networks as described in previous work (Schmidhuber, 1992b, 1993a,b).”
BackgroundMethodSelf-citationPlanning to be surprised: optimal bayesian exploration in dynamic environments
Sun, Y., Gomez, F., & Schmidhuber, J. (2011) · Lecture notes in computer science · doi:10.1007/978-3-642-22887-2_5 · Cited by 134 · cited 2×
“xp loration seem obvious. One should analyze theoretically and experimentally under which conditions the creation of self-generated tasks can accelerate the solution to externally generated tasks—see [30, 54, 37, 38, 19, 4, 28, 62] for previous simple experimental studies in this vein. 6.1 Self-Reference Through Novel…”
BackgroundResultInfluentialGödel Machines: Fully Self-referential Optimal Universal Self-improvers
Schmidhuber, J. (2007) · Cognitive Technologies · doi:10.1007/978-3-540-68677-4_7 · Cited by 110
Self-citation
Cited by
Deep learning in neural networks: an overview
Schmidhuber, J. (2015) · Neural Networks · doi:10.1016/j.neunet.2014.09.003 · Cited by 18,570
Self-citationAndrychowicz, M., Wolski, F., Ray, A., Schneider, J., Fong, R., Welinder, P., McGrew, B., Tobin, J., Abbeel, P., & Zaremba, W. (2017) · arXiv · doi:10.48550/arxiv.1707.01495 · Cited by 1,355
Solving Rubik's Cube with a Robot Hand
OpenAI, Akkaya, I., Andrychowicz, M., Chociej, M., Litwin, M., McGrew, B., Petron, A., Paino, A., Plappert, M., Powell, G., et al. (2019) · arXiv · doi:10.48550/arxiv.1910.07113 · Cited by 602
Active learning of inverse models with intrinsically motivated goal exploration in robots
Baranes, A., & Oudeyer, P.-Y. (2013) · Robotics and Autonomous Systems · doi:10.1016/j.robot.2012.05.008 · Cited by 455
Towards Continual Reinforcement Learning: A Review and Perspectives
Khetarpal, K., Riemer, M., Rish, I., & Precup, D. (2020) · arXiv · doi:10.48550/arxiv.2012.13490 · Cited by 227
Towards Continual Reinforcement Learning: A Review and Perspectives
Khetarpal, K., Riemer, M., Rish, I., & Precup, D. (2022) · Journal of Artificial Intelligence Research · doi:10.1613/jair.1.13673 · Cited by 224
Curriculum Learning for Reinforcement Learning Domains: A Framework and Survey
Narvekar, S., Peng, B., Leonetti, M., Sinapov, J., Taylor, M. E., & Stone, P. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2003.04960 · Cited by 224
“order of the goals. Other methods they propose include reducing the number of goals, creating tasks to solve dicult subgoals, and changing domain operators and objects available for binding. Lastly, Schmidhuber (2013) introduced Powerplay, a framework that focuses on inventing new problems to train a more and more gen…”
BackgroundInfluentialMulti-Task Deep Reinforcement Learning with PopArt
Hessel, M., Soyer, H., Espeholt, L., Czarnecki, W., Schmitt, S., & Van Hasselt, H. (2019) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v33i01.33013796 · Cited by 196
Play in predictive minds: A cognitive theory of play.
Andersen, M. M., Kiverstein, J., Miller, M., & Roepstorff, A. (2023) · Psychological Review · doi:10.1037/rev0000369 · Cited by 122
Play, Curiosity, and Cognition
Chu, J., & Schulz, L. E. (2020) · Annual Review of Developmental Psychology · doi:10.1146/annurev-devpsych-070120-014806 · Cited by 93
Clune, J. (2019) · arXiv · doi:10.48550/arxiv.1905.10985 · Cited by 82
“One project in this direction is PowerPlay [151, 163].”
BackgroundIntrinsic motivations and open-ended development in animals, humans, and robots: an overview
Baldassarre, G., Stafford, T., Mirolli, M., Redgrave, P., Ryan, R. M., & Barto, A. (2014) · Frontiers in Psychology · doi:10.3389/fpsyg.2014.00985 · Cited by 73
Open-Ended Learning Leads to Generally Capable Agents
Open Ended Learning Team, Stooke, A., Mahajan, A., Barros, C., Deck, C., Bauer, J., Sygnowski, J., Trebacz, M., Jaderberg, M., Mathieu, M., et al. (2021) · arXiv · doi:10.48550/arxiv.2107.12808 · Cited by 56
“In PowerPlay, Schmidhuber (2013) propose a framework to continuously seek the simplest unsolved challenge to train on.”
BackgroundWang, R., Lehman, J., Rawal, A., Zhi, J., Li, Y., Clune, J., & Stanley, K. O. (2020) · arXiv · doi:10.48550/arxiv.2003.08536 · Cited by 53
Empowerment contributes to exploration behaviour in a creative video game
Brändle, F., Stocks, L. J., Tenenbaum, J. B., Gershman, S. J., & Schulz, E. (2023) · Nature Human Behaviour · doi:10.1038/s41562-023-01661-2 · Cited by 34
Replay-Guided Adversarial Environment Design
Jiang, M., Dennis, M., Parker-Holder, J., Foerster, J., Grefenstette, E., & Rocktäschel, T. (2021) · arXiv · doi:10.48550/arxiv.2110.02439 · Cited by 21
“In inducing parallel curricula, DCD follows a rich lineage of curriculum learning methods [2, 32, 28, 23].”
BackgroundGeneral intelligence requires rethinking exploration
Jiang, M., Rocktäschel, T., & Grefenstette, E. (2022) · arXiv · doi:10.48550/arxiv.2211.07819 · Cited by 17
“Similar to the outlined open-ended exploration process, prior work on the POWERPLAY algorithm has considered finding increasingly general agents by jointly searching through the space of MDP programs and that of solution programs [33].”
BackgroundMethodInfluentialAbsolute Zero: Reinforced Self-play Reasoning with Zero Data
Zhao, A., Wu, Y., Yue, Y., Wu, T., Xu, Q., Yue, Y., Lin, M., Wang, S., Wu, Q., Zheng, Z., et al. (2025) · arXiv · Cited by 13
Wang, R., Lehman, J., Clune, J., & Stanley, K. O. (2019) · arXiv
Metadata
- Authors
- Schmidhuber, J.
- Venue
- Frontiers in Psychology
- Publisher
- Frontiers Media SA
- Published
- 2013
- Volume
- 4
- Pages
- 313-313
- ISSN
- 1664-1078 (ISSN-L)
- OpenAlex
- W2116386744
- Cited by
- 118
- References
- 13
- FWCI
- 6.02
- Type
- journal article
- Language
- en
- Open access
- gold
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.
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.