DeepMind report: AI supercharges hypothesis generation while validation bottlenecks move downstream
JMateosGarcia · x · 2026-10-21
Google DeepMind's Alex Imas (Director of AGI Economics) and James Manyika publish "Bending the Curve: AI in Science Today and Tomorrow," mapping how scientists actually use AI.
Key points:
- Historically the wet lab was the primary site of ideation with computation as an aid; now in silico simulation is becoming the main source of ideation, pushing physical labs downstream into testing and verification
- Just under half of surveyed scientists say their primary constraints have migrated into wet-lab testing, bench analysis and clinical trials — creating a fast-growing backlog of untested hypotheses
- An AlphaFold impact study found less-studied proteins did become basic-research targets, with little evidence yet of downstream applied contributions
- The report frames this via Mokyr's "positive feedback loop" of science-driven growth and US-EU divergence since the 1990s
This matches the Citrini/Claude antivenom experiment: 64 hours of dry-lab design and 39 days to results, but 4-7 years to release — AI accelerates hypothesis generation while bottlenecks shift to verification.
More from AGI Musings
- Polymarket puts just 7% odds on OpenAI pausing AI training this year — Polymarket · 2026-10-21
- Daniel Lemire: AI-generated papers are exposing academia's 'Profzi scheme' — lemire · 2026-10-21
- Universities block thesis downloads while each dissertation averages barely a handful of lifetime reads — generativist · 2026-10-21
- Berkeley wins Schmidt Sciences HAVI grant to decode whale and parrot communication with AI — begusgasper · 2026-10-21
- 3Blue1Brown says he'll keep making videos long after LLMs can generate clear explainers — 3blue1brown · 2026-10-21
- Talos researcher: AI agent attacks today are loud pentests — quiet swarms are the real threat — matthew_d_green · 2026-10-21