Hinton, Li, and Ng Debate Open Weights and Concentration of AI Power at Ai4

Skynet Chance (+0.02 pp): Hinton's acknowledgment that open weights are now irreversible and cheaply fine-tunable for malicious uses like cyberattacks implies proliferation of powerful models beyond any central control point, modestly raising loss-of-control and misuse risk. The counterweight is that all three speakers endorsed regulation, keeping the net increase small.

Skynet Date (no change): Widespread availability of open weights plus a framed US-China race dynamic slightly accelerates the diffusion of frontier-level capability to uncontrolled actors. This is commentary on an existing trend rather than a new capability, so the timeline effect is marginal.

AGI Progress (+0.00 pp): The discussion concerns distribution, governance, and market structure rather than any new capability, algorithm, or scaling result, so direct progress toward AGI is essentially unchanged. Broad open access can modestly widen the pool of researchers building on frontier models.

AGI Date (no change): Arguments for openness and competitive pressure against Chinese open-weight models point toward more actors iterating on strong base models, a very slight accelerant. No funding, compute, or research breakthrough is announced that would meaningfully shift AGI timing.