A response to the UK’s Future of Compute Review, co-signed by Jess Whittlestone (CLTR), Shahar Avin (CSER), Lennart Heim (GovAI), Markus Anderljung (GovAI), and Girish Sastry (OpenAI). It states that the Future of Compute review fails to provide adequate strategies for ensuring responsible and well-governed use of compute and gives four suggestions for how the UK Government could improve their approach to managing high-risk AI development:
- A tiered access approach to compute provision.
- Requiring AI companies to report, or possibly in the future apply for a license for, training runs above a certain (very high) threshold.
- Requiring compute providers to have “Know Your Customer (KYC)” processes.
- Facilitating academic access to large pre-trained models to address the compute divide.