DeepMind paper: let AI models bid for prompts — auctions hit 99.2% of optimal allocation
rohanpaul_ai · x · 2026-10-23
A new Google DeepMind paper, "Price of Anarchy of First-Price Auctions in Agentic Markets," proposes a decentralized serving-time market where AI agents bid to claim each user prompt, with bids based on expected reward minus inference costs; the highest bidder answers.
Key result: even without knowing which model is best for a given question, this first-price auction achieves results 99.2% as good as always picking the best model. The paper identifies two features that drive the price-of-anarchy toward zero:
- Competitive floor: multiple bidders typically value prompts above a baseline share of optimal welfare (e.g., the best models are often distilled into cheaper, similar-performing alternatives), so competition limits welfare loss.
- Near-independence of values conditioned on the prompt.
FPAs aren't truthful — bidders may shade bids — but they're simple, transparent, and require no central knowledge of private capabilities and costs.
Related event: DeepMind paper: first-price auctions route AI queries at 99.2% of optimal(3 posts)→
More from AGI Musings
- Humanities dissertations have collapsed, social sciences are next — with AI as one driver — gsiemens · 2026-10-23
- Greg Brockman: AI makes hacking 100x more profitable and easier, upending hacker economics — gabriel1 · 2026-10-23
- 'Nothing is real now on social media': viral demo shows AI-generated content everywhere — saibharadwaj · 2026-10-23
- Lilian Weng to Discuss Recursive Self-Improvement at SGLang Summit 2026 — BanghuaZ · 2026-10-23
- mmitchell_ai questions whether the AI safety vs security framing stems from EA — mmitchell_ai · 2026-10-23
- AI, Robotics and Abundant Energy Will Undo Millennia of Scarcity-Driven Life, Argues Kardashev-Scale Optimist — Dr_Singularity · 2026-10-23