Computational genomics expert rebuts 'AI slop' criticism of DeepMind's AlphaGenome
anshulkundaje · x · 2026-09-18
Computational genomics researcher Anshul Kundaje published a long thread rebutting Steven Salzberg's claim that DeepMind's AlphaGenome Atlas is 'AI slop taking over genomics.' The Atlas, covered by Nature, charts predicted effects of 9 billion single-letter mutations across the human genome and is freely available for non-commercial use.
- AlphaGenome is a supervised foundation model mapping DNA sequence to genome-wide biochemical profiles from thousands of experiments (binding, accessibility, transcription) across cell contexts.
- The model class has been extensively evaluated on regulatory sequence syntax discovery, variant effect prediction, and regulatory sequence design, spanning small single-task to large multi-task models.
- Evaluations use perturbation datasets, synthetic libraries, and silver-standard causal variant sets. Kundaje acknowledges known biases but argues the evals are not fundamentally flawed.
- The paper also includes molecular-level experiments on selected variant examples.
His core argument: AG and the AVI score rest on a well-established paradigm with concrete evidence, not evidence-free hype.
More from AGI Musings
- As training scales up, labs will only spend compute on math and coding data, researcher argues — moultano · 2026-09-18
- e/acc leader Beff Jezos calls top Effective Altruists "turbo sociopath cult leaders" — beffjezos · 2026-09-18
- e/acc figure slams EA for wanting 'a total global AI panopticon' — beffjezos · 2026-09-18
- Meme mocks Dreamforce 2026 keynote feat. EA: 'the power of dreams' — Dr_Singularity · 2026-09-18
- Every non-tech manager is a coder until the 2AM production outage hits, dev argues — ashishllm · 2026-09-18
- Stanford philosopher defends p(doom): subjective probabilities are perfectly legitimate — sethlazar · 2026-09-18