Position: Open-Endedness is Essential for Artificial Superhuman Intelligence

Edward Hughes, Michael D Dennis, Jack Parker-Holder, Feryal Behbahani, Aditi Mavalankar, Yuge Shi, Tom Schaul, Tim Rocktäschel
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:20597-20616, 2024.

Abstract

In recent years there has been a tremendous surge in the general capabilities of AI systems, mainly fuelled by training foundation models on internet-scale data. Nevertheless, the creation of open-ended, ever self-improving AI remains elusive. In this position paper, we argue that the ingredients are now in place to achieve open-endedness in AI systems with respect to a human observer. Furthermore, we claim that such open-endedness is an essential property of any artificial superhuman intelligence (ASI). We begin by providing a concrete formal definition of open-endedness through the lens of novelty and learnability. We then illustrate a path towards ASI via open-ended systems built on top of foundation models, capable of making novel, human-relevant discoveries. We conclude by examining the safety implications of generally-capable open-ended AI. We expect that open-ended foundation models will prove to be an increasingly fertile and safety-critical area of research in the near future.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-hughes24a, title = {Position: Open-Endedness is Essential for Artificial Superhuman Intelligence}, author = {Hughes, Edward and Dennis, Michael D and Parker-Holder, Jack and Behbahani, Feryal and Mavalankar, Aditi and Shi, Yuge and Schaul, Tom and Rockt\"{a}schel, Tim}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {20597--20616}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/hughes24a/hughes24a.pdf}, url = {https://proceedings.mlr.press/v235/hughes24a.html}, abstract = {In recent years there has been a tremendous surge in the general capabilities of AI systems, mainly fuelled by training foundation models on internet-scale data. Nevertheless, the creation of open-ended, ever self-improving AI remains elusive. In this position paper, we argue that the ingredients are now in place to achieve open-endedness in AI systems with respect to a human observer. Furthermore, we claim that such open-endedness is an essential property of any artificial superhuman intelligence (ASI). We begin by providing a concrete formal definition of open-endedness through the lens of novelty and learnability. We then illustrate a path towards ASI via open-ended systems built on top of foundation models, capable of making novel, human-relevant discoveries. We conclude by examining the safety implications of generally-capable open-ended AI. We expect that open-ended foundation models will prove to be an increasingly fertile and safety-critical area of research in the near future.} }
Endnote
%0 Conference Paper %T Position: Open-Endedness is Essential for Artificial Superhuman Intelligence %A Edward Hughes %A Michael D Dennis %A Jack Parker-Holder %A Feryal Behbahani %A Aditi Mavalankar %A Yuge Shi %A Tom Schaul %A Tim Rocktäschel %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-hughes24a %I PMLR %P 20597--20616 %U https://proceedings.mlr.press/v235/hughes24a.html %V 235 %X In recent years there has been a tremendous surge in the general capabilities of AI systems, mainly fuelled by training foundation models on internet-scale data. Nevertheless, the creation of open-ended, ever self-improving AI remains elusive. In this position paper, we argue that the ingredients are now in place to achieve open-endedness in AI systems with respect to a human observer. Furthermore, we claim that such open-endedness is an essential property of any artificial superhuman intelligence (ASI). We begin by providing a concrete formal definition of open-endedness through the lens of novelty and learnability. We then illustrate a path towards ASI via open-ended systems built on top of foundation models, capable of making novel, human-relevant discoveries. We conclude by examining the safety implications of generally-capable open-ended AI. We expect that open-ended foundation models will prove to be an increasingly fertile and safety-critical area of research in the near future.
APA
Hughes, E., Dennis, M.D., Parker-Holder, J., Behbahani, F., Mavalankar, A., Shi, Y., Schaul, T. & Rocktäschel, T.. (2024). Position: Open-Endedness is Essential for Artificial Superhuman Intelligence. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:20597-20616 Available from https://proceedings.mlr.press/v235/hughes24a.html.

Related Material