Regulating advanced artificial agents

Michael K Cohen^{1

2}, Noam Kolt^{3

4}, Yoshua Bengio^{5

6}, Gillian K Hadfield^{2

3

4

7}, Stuart Russell^{1

2}

Affiliations

¹ University of California, Berkeley, CA, USA.
² Center for Human-Compatible Artificial Intelligence, Berkeley, CA, USA.
³ University of Toronto, Toronto, Ontario, Canada.
⁴ Schwartz Reisman Institute for Technology and Society, Toronto, Ontario, Canada.
⁵ Université de Montréal, Montréal, Québec, Canada.
⁶ Mila-Quebec AI Institute, Montréal, Québec, Canada.
⁷ Vector Institute for Artificial Intelligence, Toronto, Ontario, Canada.

Michael K Cohen et al. Science. 2024.

. 2024 Apr 5;384(6691):36-38.

doi: 10.1126/science.adl0625. Epub 2024 Apr 4.

Michael K Cohen^{1

2}, Noam Kolt^{3

4}, Yoshua Bengio^{5

6}, Gillian K Hadfield^{2

3

4

7}, Stuart Russell^{1

2}

¹ University of California, Berkeley, CA, USA.
² Center for Human-Compatible Artificial Intelligence, Berkeley, CA, USA.
³ University of Toronto, Toronto, Ontario, Canada.
⁴ Schwartz Reisman Institute for Technology and Society, Toronto, Ontario, Canada.
⁵ Université de Montréal, Montréal, Québec, Canada.
⁶ Mila-Quebec AI Institute, Montréal, Québec, Canada.
⁷ Vector Institute for Artificial Intelligence, Toronto, Ontario, Canada.

Abstract

Governance frameworks should address the prospect of AI systems that cannot be safely tested.