General Notes, Ideas, and Thoughts

Ideas (add them at aisi.ai/list/ai-governance where you can also indicate interest)

• Cloud compute providers (inspired by https://docs.google.com/document/u/0/d/13LJhP3ksrcEBKxYFG5GkJaC2UoxHKUYAHCRdRlpePEc/)
  • sliding scale regulation? (based on amt of compute one has access to or what it is used for)
  • global coordination for cloud compute
  • standardized APIs?
  • Structured access: an emerging paradigm for safe AI deployment
  • Oversight for Frontier AI through a Know-Your-Customer Scheme for Compute Providers
  • “What can we learn from analogous industries and technologies such as DNA synthesis companies, social media platforms, controls of nuclear materials, and international finance (e.g. the proliferation of Know Your Customer guidelines)?”

Key considerations for the weekend's project

Notes from the breakout session

Overview and Resource Notes

One of the biggest problems of successful societal-scale AI legislation compared to legislation for other high-risk technologies is that it is difficult to properly control compute and AI development. One bad actor has a significant amount of power.

Take notes in the foldout

• How to Catch a Chinchilla – a systemic overview on how we might audit compute usage for foundation models (Shavit, 2023)
• Compute Governance Introduction – an overview from Lennart Heim of how we can legislate AI by controlling GPUs (listen also to the podcast with Lennart Heim)
• [Additional reading] Representation Engineering – a technical framework to understand how capabilities are presented in various models
• [Additional reading] Export Controls of IaaS Controlled AI Chips – how the use of compute rental services affects export controls