Your privacy, your choice

We use essential cookies to make sure the site can function. We also use optional cookies for advertising, personalisation of content, usage analysis, and social media, as well as to allow video information to be shared for both marketing, analytics and editorial purposes.

By accepting optional cookies, you consent to the processing of your personal data - including transfers to third parties. Some third parties are outside of the European Economic Area, with varying standards of data protection.

See our privacy policy for more information on the use of your personal data.

for further information and to change your choices.
Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • NEWS AND VIEWS

AI discovers learning algorithm that outperforms those designed by humans

An artificial-intelligence algorithm that discovers its own way to learn achieves state-of-the-art performance, including on some tasks it had never encountered before.
By
  1. Joel Lehman

    1. Joel Lehman is at Lila Sciences, Cambridge, Massachusetts 02141, USA, and at the Laboratory for Human-Centered AI, University of Oxford, Oxford OX1 3Q, UK.

Access through your institution
Buy or subscribe

Rapid progress in machine learning has been brought about by algorithms that are trained using increasingly large data sets and vast computational resources, while depending less and less on human expertise. The design of such algorithms has remained the province mainly of human programmers, but, writing in Nature, Oh et al.1 report an algorithm that can create a state-of-the-art machine-learning algorithm belonging to an area of artificial intelligence called reinforcement learning.

Access options

Access through your institution

Rent or buy this article

Prices vary by article type

from$1.95

to$39.95

Learn more

Prices may be subject to local taxes which are calculated during checkout

doi: https://doi.org/10.1038/d41586-025-03398-6

References

  1. Oh. J. et al. Nature https://doi.org/10.1038/s41586-025-09761-x (2025).

    Article  Google Scholar 

  2. Wang, J. X. Curr. Opin. Behav. Sci. 38, 90–95 (2021).

    Article  Google Scholar 

  3. Hendrycks, D., Carlini, N., Schulman, J. & Steinhardt, J. Preprint at arXiv https://doi.org/10.48550/arXiv.2109.13916 (2021).

  4. Sharma, M. et al. Preprint at arXiv https://doi.org/10.48550/arXiv.2310.13548 (2023).

  5. Taylor, T. et al. Artif. Life 3, 408–442 (2016).

    Article  Google Scholar 

  6. Hughes, E. et al. In Proc. 41st Int. Conf. Mach. Learn. 235, 20597–20616 (Proceedings of Machine Learning Research, 2024).

  7. Clune, J. Preprint at arXiv https://doi.org/10.48550/arXiv.1905.10985 (2019).

  8. Mouret, J.-B. Nature 625, 452–453 (2024).

    Article  PubMed  Google Scholar 

  9. Lu, C. et al. Preprint at arXiv https://doi.org/10.48550/arXiv.2408.06292 (2024).

  10. Packard, N. et al. Artif. Life 25, 93–103 (2019).

    Article  PubMed  Google Scholar 

Download references

Competing Interests

J.L. has some OpenAI shares from time spent working there, and currently works for Lila.ai as a research scientist.

Subjects

Latest on:

Nature Careers

Jobs

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing