Pacing the Frontier: researchers launch a new field on whether to slow AI, citing 298 papers and mapping open questions
gleech · x · 2026-09-17
The authors behind pacing.tech introduce an effort to turn the debate over whether AI progress should be slowed ("pacing") into a real research field. Their full paper cites 298 relevant works, including end-to-end analyses, and they stress they are not assuming the conclusion — they start from reasons not to pace, aiming for reasoned views to genuinely cohere. A companion quiz puts all positions on one page and collects reader attitudes toward pacing.
The appendices lay out the open questions:
- Why pace? How much would more calendar time improve AI risk management, and which safety measures are ready now vs. need years (citing MacAskill & Moorhouse 2025, Hobbhahn 2025)?
- Will pacing get harder or easier? What investments preserve optionality; does distributed training undermine compute governance (citing Sastry et al. 2024)?
- How do actual decision-makers pace? How can external research reach government or lab leadership (citing METR 2025 on frontier safety policies)?
- Which risks would trigger intervention? How dangerous capabilities sit on the offense-defense balance over time (citing Garfinkel & Dafoe, Shevlane & Dafoe).
More from AGI Musings
- Gary Marcus: 'Rogue agents' is AI's excuse for irresponsibly built software — GaryMarcus · 2026-09-18
- Roman Yampolskiy returns to DOAC to update his AI safety warning that reached 20M+ viewers — MaxUnfried · 2026-09-18
- Ex-OpenAI's Yacine: enterprises are the next thinking machines — learn AI or get eaten — yacineMTB · 2026-09-18
- yacineMTB Predicts AI Tech Is Too Powerful for a Few to Control — yacineMTB · 2026-09-18
- Cell Focus: Biology Needs World Models That Predict Interventions, Not Just Descriptions — rishabh16_ · 2026-09-18
- AI Agents Find Drug Targets With Single-Cell Evidence Are 48% More Likely to Reach Market — james_y_zou · 2026-09-18