AI
#257 – Tom Milton on how we could verify a US–China deal on AI
Publication: 80,000 Hours Podcast
Author: The 80,000 Hours team
Source published: 23/10/2026, 15:33:52 UTC
Briefing created: 08/11/2026, 05:06:58 UTC
Based on the episode description, not a full transcript. Source evidence retrieved 03/11/2026, 22:03:24 UTC.
Briefings
Tom Milton, CEO of Amodo Design, argues that verifying AI compliance during a US-China deal requires shutting down data centers, which wastes resources. He estimates only 50 engineers worldwide work on compute verification, compared to billions spent on Cold War nuclear testing. Amodo's pilot tool reruns inference on a separate server to prove traffic matches agreed models. The episode was recorded September 29, 2026, and covers topics like nuclear arms control lessons and AI hardware design, but verification tools remain small with long lead times.
Key points
- Tom Milton, CEO of Amodo Design, argues that verifying AI compliance during a US-China deal requires shutting down data centers, which wastes resources and makes agreements costly.
- Amodo's pilot tool reruns inference on a separate server to prove traffic matches agreed models, while Tom Milton estimates only 50 engineers worldwide work on compute verification compared to billions spent on Cold War nuclear testing.
Context
Tom Milton, CEO of Amodo Design, a UK hardware lab building prototype tools to verify how AI chips are being used, argues that verification is essential for affordable AI slowdowns. The episode was recorded September 29, 2026, and covers topics like nuclear arms control lessons and AI hardware design, but verification tools remain small with long lead times.
Limitations
Evidence is limited to a single podcast episode from September 2026. Coverage is restricted to Tom Milton's arguments and Amodo's specific tools; no independent verification data, government reports, or long-term impact studies are provided. Uncertainty exists regarding the feasibility of scaling verification efforts beyond the pilot stage and the exact timeline for tool deployment.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 08/11/2026, 05:06:58 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeihkztkea7lkkpjy6orqbgnnc6o6rv2llwmpcj6dkisbvnq3mmxpke