A biotech startup called Vivodyne is arguing that the AI drug-discovery field suffers from a fundamental data shortage—and claims to have built a machine to solve it. The company’s HIVE platform consists of modular robotic labs that can grow 20 different types of human tissue, then automatically dose and monitor them. This generates the kind of causal biological data that today’s AI models lack—data that currently comes mostly from animal testing or studies of individual cells and proteins, not living tissue. “Absent human testing, what are these [AI] models going to do?” asks Andrei Georgescu, Vivodyne’s CEO and co-founder. “They’re going to cure cancer in mice.”

Even Anthropic CEO Dario Amodei noted over the weekend that claims about AI curing cancer have become more cliché than credible—“the thing that will work is actually curing cancer,” as he put it. To be fair, Amodei himself has floated that idea in earlier essays. Sam Altman has repeatedly cited curing cancer as a justification for OpenAI’s push toward AGI and ever-larger compute buildouts, and Google DeepMind’s Demis Hassabis said last year that AI could potentially cure all disease within a decade. The actual results, however, remain modest. A few AI-designed drugs have entered human trials—one reaching Phase III, which involves widespread human testing—but the reality is that many roadblocks aren’t ones AI can solve today. Nobel-winning AlphaFold was a major breakthrough for understanding the building blocks of life, yet it hasn’t produced a new drug. Isomorphic Labs, founded to build on AlphaFold, is expecting its first trials, originally planned for 2025, by the end of this year. In February, the company wrote that true drug discovery will require “highly accurate predictive models, across an expansive range of biochemical properties and interactions.”

Georgescu says the field needs “a sanity check”—that existing models don’t have enough data to capture the complexity of human biology. This is already a challenge facing the pharmaceutical industry, where 90% of drugs that work in animal testing and enter clinical trials don’t receive regulatory approval for humans. Vivodyne’s approach is different. The company was spun out of the University of Pennsylvania in 2021, after Georgescu earned a PhD in bioengineering there. Vivodyne says its tissues closely mirror the behavior of real human organs: its liver cells have 94% predictive accuracy compared to human trials testing for toxicity, its airway tissue matches real human tissue behavior 96% of the time, and its bone marrow has achieved 100% concordance in tests of 20 different chemotherapy drugs. Last week, the company—which has raised just under $80 million across two rounds led by Khosla Ventures—opened what it calls the world’s largest “human data center” just outside San Francisco. Georgescu says his team is already achieving twice the throughput of all animal trials held in the US.

The goal is to accelerate the path of drug candidates by having a clearer idea of what will work before spending tens of millions of dollars on a clinical trial. Vivodyne won’t name its partners publicly, but says it’s working with multiple major pharma companies to tackle a problem Georgescu compares to automotive crash tests: An automaker is usually confident its car will pass NHTSA requirements before testing it, but drugmakers rarely have that same confidence heading into a clinical trial, where the vast majority of drugs fail to win FDA approval. But there’s a bigger vision at play. Georgescu sees his autonomous biology labs as essential to generating the causal data needed to train new models on human biology. He points to studies like this one, published in Nature Methods last month, which found no clear data scaling laws when training generative AI models on existing cellular data. “In other words, the model learns ‘this is cell state A,’ ‘this is cell state B,’ but never ‘cell state B is the effect of inflaming cell state A.’”

Vivodyne’s HIVE machines, however, are tracking hundreds of thousands of ongoing experiments where diseased tissue is exposed to some stimulus. Georgescu expects this to provide the kind of reinforcement learning that will produce AI models capable of understanding human biology well enough to make more meaningful progress in healthcare. He believes that will be key not just for today’s medical challenges, but also for a future where complex diseases require drugs that—unlike most available today—target multiple pathways. “You have to say, ‘I want this effect to happen, so what cause should I invoke.’ Establishing causality in human biology is the basis of all of this.”