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.

  • Correspondence
  • Published:

Don’t pause giant AI for the wrong reasons

Nature Machine Intelligence volume 5, pages 470–471 (2023)Cite this article

Subjects

An open letter1 to the Future of Life Institute has called on all AI labs to “immediately pause for at least 6 months the training of AI systems more powerful than GPT-4”. The letter has been signed by prominent AI experts, intellectuals, industry actors and other commentators. The authors go so far as to state that “if such a pause cannot be enacted quickly, governments should step in and institute a moratorium”.

This prescriptive thesis is based on two premises: (i) “contemporary AI systems are now becoming human-competitive at general tasks” and thereby “can pose profound risks to society and humanity”, and (ii) the current “level of planning and management” of AI is insufficient. Both premises are empirically testable. However, the available evidence does not to support either. Regarding the first premise, there is no evidence that systems like GPT-4 are “human-competitive at general tasks”, at least not to the point of being reasonably definable as ‘non-human minds’, as implied in the open letter. Large language models (LLMs) such as GPT-4 — which seems to be the focus of the letter’s signatories’ concerns — show potential for solving new and difficult tasks ranging from mathematics, coding and vision to medicine, psychology and more, without the need for special prompts2. Although extensive, this list of tasks is far from encompassing the full range of human cognitive abilities. Moreover, several preliminary observations show how such systems are prone to factual error and have linguistic blind spots3. To have a mind, LLMs should have features such as understanding and consciousness. It is unclear whether LLMs have comprehension capabilities: they appear to behave as stochastic emulators, with no demonstrable ability to comprehend the content of the text they generate4. Because they have neither a body nor sensory processing, LLMs lack, at a minimum, bodily and sensory consciousness. As such, they are far from what could be called non-human minds. These considerations do not rule out the logical possibility of artificial minds. However, glossing over them does not add clarity to the public debate about AI.

This is a preview of subscription content, access via your institution

Relevant articles

Open Access articles citing this article.

Access options

Buy this article

  • Purchase on Springer Link
  • Instant access to full article PDF
Buy now

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

References

  1. Bengio, Y. et al. Pause Giant AI Experiments: an Open Letter. Future of Life Institute; https://futureoflife.org/open-letter/pause-giant-ai-experiments/ (2023).

  2. Bubeck, S. et al. Preprint at https://doi.org/10.48550/arXiv.2303.12712 (2023).

  3. Floridi, L. Philos. Technol. 36, 15 (2023).

    Article  Google Scholar 

  4. Bender, E. M., Gebru, T., McMillan-Major, A. & Mitchell, M. In Proc. 2021 ACM Conference on Fairness, Accountability, and Transparency 610–623 (ACM, 2021).

  5. Jobin, A., Ienca, M. & Vayena, E. Nat. Mach. Intell. 1, 389–399 (2019).

    Article  Google Scholar 

  6. Gebru, T., Bender, M., McMillan-Major, A. & Mitchell, M. DAIR Institute Blog https://www.dair-institute.org/blog/letter-statement-March2023 (2023).

  7. Oye, K. A. Proactive and adaptive governance of emerging risks: the case of DNA synthesis and synthetic biology (International Risk Governance Council, Geneva, 2012).

  8. McLennan, S. et al. Nat. Mach. Intell. 2, 488–490 (2020).

    Article  Google Scholar 

Download references

Author information

Authors and Affiliations

  1. Laboratory of Ethics of AI and Neuroscience, School of Medicine, Technical University of Munich (TUM), Munich, Germany

    Marcello Ienca

  2. College of Humanities, Swiss Federal Institute of Technology in Lausanne (EPFL), Lausanne, Switzerland

    Marcello Ienca

Authors
  1. Marcello Ienca
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Marcello Ienca.

Ethics declarations

Competing interests

The author declares no competing interests.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Ienca, M. Don’t pause giant AI for the wrong reasons. Nat Mach Intell 5, 470–471 (2023). https://doi.org/10.1038/s42256-023-00649-x

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1038/s42256-023-00649-x

This article is cited by

Search

Advanced search

Quick links

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