Alex Rives argues current virtual cell models fail because biology's measuring tools cannot observe the fine-grained causal structure models need. He says better architectures help, but Biohub is investing heavily in new measurement modalities and data generation because observation limits are the deeper bottleneck first.
Alex Rives notes that with infinite data, model architectures converge, but cellular biology is far from that limit. He says inductive bias therefore matters most when data is scarce, and today's transformer lacks the multi-scale structure needed to represent molecular and tissue levels in practice.
Alex Rives describes Biohub's Virtual Biology Initiative with DOE, NIH, Google DeepMind, Isomorphic, and Meta as a coordinated push to generate cellular data. Top priorities include spatial biology and cell atlases linking molecular and imaging measurements across hundreds of billions of healthy and diseased cells.
Alex Rives explains the laser phase plate is phase contrast for electron microscopy, using the world's most powerful standing-wave laser to shift an electron beam's phase. He says this decades-in-the-making tool helps cryo-electron tomography pick out individual proteins and macromolecules inside otherwise noisy cellular images.
Alex Rives says a complete cell model must include cells in tissue and living systems, not just isolated cultures. Biohub builds real imaging across atomic, cell-culture, and whole-organism scales, including zebrafish microscopes that track every cell and immune-cell trafficking as an embryo develops over time.
Alex Rives points out ESM and AlphaFold trained across all evolution, not just human proteins. He says core biological machinery is conserved while programs are rewired, so predictive biological models can transfer from many organisms instead of relying on a single imperfect laboratory mouse proxy.
Alex Rives expects biology to follow a pretraining-plus-elicitation recipe, but says reinforcement learning must be verified in laboratories. Because biology is empirical, he envisions AI systems proposing real experiments, running them, learning from actual results, and recursively improving their biological models over repeated improvement cycles.
Alex Rives says health-record models learn correlations, which can route existing drugs to the right patients better but won't uncover causal biology. Biohub instead focuses on fundamental measurement and models, arguing that without the right basis, patient-history AI cannot explain the truly deep underlying mechanisms of disease.
Alex Rives says language captures centuries of biological knowledge but not empirical facts that only direct experiments reveal. He notes protein models learn features for unknown biology that lacks names, so biology may need its own separate latent space rather than mere arm's-length tool calls.
Alex Rives says continuous molecular monitoring would be immensely valuable and likely converge with cell models eventually. Measuring more molecular parameters continuously could directly link bottom-up personal health data with top-down biological understanding of disease, though Biohub itself stays at the enabling technology layer.
Alex Rives frames cell-state transitions as a probabilistic transition function: a model takes a cell or tissue state and inputs, then predicts a distribution of possible next states. This makes virtual cells harder than deterministic predictors, but better matched to biology's noisy, branching real outcomes.
Alex Rives treats disease as a regular biological program encoded in human life, opening the door to programmable medicine. Reading and rewriting those programs could directly let interventions target underlying mechanisms rather than just surface symptoms, which is central to Biohub's ambitious human cure-all-disease mission.
