Vivodyne's Robotic Labs Aim to Fix AI Drug Discovery's Data Gap
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Vivodyne says the AI drug-discovery industry has a data problem and has built modular robotic labs called HIVE that grow 20 kinds of human tissue, autonomously dose them, and generate causal biological data missing from today's training sets
- CEO Andrei Georgescu argues AI models trained on static cellular snapshots lack causal context, pointing to a Nature Methods study finding no clear data scaling laws for generative AI on cellular data — models learn 'cell state A' and 'cell state B' but never that B is the effect of inflaming A
- Vivodyne claims specific predictive accuracy for its tissues: 94% for liver toxicity, 96% for airway tissue, and 100% concordance for bone marrow across 20 chemotherapy drugs, and says its new facility already achieves twice the throughput of all US animal trials
- Vivodyne has raised just under $80 million across two rounds led by Khosla Ventures and says it is working with multiple unnamed major pharma companies to predict clinical-trial outcomes before the trials themselves — which typically cost tens of millions of dollars and fail 90% of the time for drugs that passed animal testing
- Anthropic CEO Dario Amodei wrote over the weekend that AI-curing-cancer claims have become "more cliche than credible," joining a chorus that includes Sam Altman's repeated cancer-cure justifications for OpenAI's AGI push and Demis Hassabis's prediction that AI could cure all disease within a decade
- Isomorphic Labs, founded to build on AlphaFold, is now expecting its first drug trials by the end of this year (originally planned for 2025) and acknowledged in February that true drug discovery will require "highly accurate predictive models, across an expansive range of biochemical properties and interactions."
Why it matters: Vivodyne is targeting a specific gap that major AI labs have begun acknowledging in public: causal data from living human tissue, not compute scale. With 90% of animal-tested drugs failing FDA approval and clinical trials costing tens of millions of dollars, even modest improvements in pre-trial prediction could redirect significant pharma R&D spend — though Vivodyne has not named its pharma partners or published peer-reviewed validation of its throughput claims.
Ask SkimNews
