Stanford Computational Biologists Clash Over AI Capability Limits and Data Bottlenecks

Stanford computational biologist Anshul Kundaje and colleague Batzoglou engaged in a multi-round public debate over the boundaries of AI capability. Kundaje criticized unbounded trend extrapolation of model abilities, saying terms like Scaling, ASI, and RSI offer no concrete roadmap and feel like fortune-telling rather than science; Batzoglou, who previously shared this skepticism, said his stance has recently changed — "I was like you, until recently. It looks like it just works."

Confirmed

• Kundaje argues that AI model capability has a clear ceiling rooted in the nature of trainable data, citing hard biology problems in data-scarce settings where frontier models have so far performed poorly.
• He questioned claims that AI will "cure all diseases in 5-10 years" as an example of how far trend extrapolation can go without a roadmap, easily leading to absurd conclusions.
• Responding to TuXinming's proposal to train on interaction trajectories rather than pure text, Kundaje said such proposals have been made repeatedly, but proponents underestimate the diversity and scale of data required, and physical bottlenecks stand in the way.
• Kundaje added that the data bottleneck will persist until AI-driven assay technology improves recursively and truly massive fully robotic labs with automated supply chains emerge.
• He was lukewarm on recent AI math proof breakthroughs (including Millennium Prize-type results): a major milestone for AI capability, but mostly without direct real-world impact as far as he knows.
• TuXinming's concrete path: pretraining compresses "world→observation→text record" into models, while tool use provides a closed loop of "act→get feedback→train on trajectories," already being practiced in labs through heavy tool calls and agent runs.
• Batzoglou said he moved from long-term skepticism to acknowledging that AI recently "just works."

Why it matters

This debate crystallizes a core divide in the AI community: one side demands verifiable implementation paths for capability leaps, the other abandons skepticism based on recent real-world performance. Kundaje's "physical bottleneck of data diversity and scale" offers a concrete criterion for assessing AI prospects in data-scarce fields like biology, and whether fully automated robotic labs can unlock this bottleneck is a key variable to track.

2026-09-09 ~ 2026-09-09 · 10 related posts

• Stanford's Anshul Kundaje: models are soberingly bad at data-poor biology, stop extrapolating from scaling — anshulkundaje · 2026-09-09
• "Scaling, ASI, RSI is fortune-telling, not science": Kundaje's critique of capability prognostication — anshulkundaje · 2026-09-09
• Stanford's Batzoglou drops his AI skepticism: "It looks like it just... works" — s_batzoglou · 2026-09-09
• Stanford professor pushes back on claims AI will cure all diseases in 5-10 years — anshulkundaje · 2026-09-09
• [source] A concrete path past the data wall: train agents on interaction trajectories, not just text — TuXinming · 2026-09-09
• Stanford professor and computational biologist clash over claims AI will cure all diseases in 5-10 years — anshulkundaje · 2026-09-09
• Stanford professor unmoved by AI math proofs: no real-world impact yet, and biology sees nothing surprising either — anshulkundaje · 2026-09-09
• Stanford's Kundaje: interaction-based training proposals underestimate data scale and physical barriers — anshulkundaje · 2026-09-09
• [source] Kundaje: the data barrier holds until AI-built assays and fully robotic labs arrive — anshulkundaje · 2026-09-09

1 near-duplicate retellings: anshulkundaje