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

Physics solves a training problem for artificial neural networks

Systems that emulate biological neural networks offer an efficient way of running AI algorithms, but they can’t be trained using the conventional approach. The symmetry of these ‘physical’ networks provides a neat solution.
  1. Damien Querlioz

    1. Damien Querlioz is at Université Paris-Saclay, CNRS, Centre de Nanosciences et de Nanotechnologies, 91120 Palaiseau, France.

    View author publications

    You can also search for this author in PubMed  Google Scholar

Huge changes are already under way in health care, industry and education as a result of advances in artificial intelligence (AI), but the costs could outweigh the benefits if AI’s enormous energy consumption is not reined in. The problem is that AI relies heavily on deep neural networks, which are layered algorithms that involve millions or even billions of computations, requiring massive, energy-hungry access to the memories in conventional computers. One possible solution is to replace the computers with systems that more closely reflect the physical structure of biological neural networks, but such systems are typically incapable of performing one of the main steps in training the network. In a paper in Nature, Xue et al.1 report an ingenious workaround that uses physics to overcome the problem.

Access options

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

Nature 632, 264-265 (2024)

doi: https://doi.org/10.1038/d41586-024-02392-8

References

  1. Xue, Z. et al. Nature 632, 280–286 (2024).

    Article  Google Scholar 

  2. McMahon, P. L. Nature Rev. Phys. 5, 717–734 (2023).

    Article  Google Scholar 

  3. Ross, A. et al. Nature Nanotechnol. 18, 1273–1280 (2023).

    Article  PubMed  Google Scholar 

  4. Ambrogio, S. et al. Nature 620, 768–775 (2023).

    Article  PubMed  Google Scholar 

  5. Lee, R. H., Mulder, E. A. B. & Hopkins, J. B. Sci. Robot. 7, eabq7278 (2022).

    Article  PubMed  Google Scholar 

  6. Marković, D., Mizrahi, A., Querlioz, D. & Grollier, J. Nature Rev. Phys. 2, 499–510 (2020).

    Article  Google Scholar 

  7. Wright, L. G. et al. Nature 601, 549–555 (2022).

    Article  PubMed  Google Scholar 

  8. Scellier, B., Ernoult, M., Kendall, J. & Kumar, S. Adv. Neural Inf. Proc. Syst. 36, 52705–52731 (2023); available at https://go.nature.com/4fmijnb.

Download references

Competing Interests

The author declares no competing interests.

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

Search

Advanced search

Quick links