We often hear that AI is on the verge of curing complex diseases. Yet, despite massive investment and high hopes, those breakthroughs have been slow to materialize. The problem is not necessarily the AI itself; it is the quality of the information we are using to teach it.
WHAT'S HAPPENING
A biotech company called Vivodyne is arguing that current AI drug discovery is built on a shaky foundation. Most AI models are trained on data gathered from animal testing or from looking at static, individual cells. The issue, according to Vivodyne, is that these snapshots do not accurately reflect the chaos and complexity of a living human body. To change this, they have developed robotic systems that can grow human tissue in a lab, then automatically test drugs on that tissue. By watching how these tissues react to different medicines in real-time, the company is generating new data that they hope will teach AI systems how human biology works, rather than just showing them pictures of it.
Teaching AI to understand human biology
HOW IT WORKS
To understand why this is a big deal, think of how most AI is trained today. Imagine trying to learn how a car engine works by looking at thousands of photographs of a car parked in a garage. You might learn what the engine looks like, but you would never understand how the parts move or what happens when you press the gas pedal. Current AI models are essentially looking at these static photos of biological cells. They can identify patterns, but they don't understand the cause and effect of a disease. If you want to know what happens to a liver cell when you introduce a specific drug, you need to see the action unfold. Vivodyne’s robotic labs are designed to run these experiments continuously, creating a feed of cause-and-effect data. This allows the AI to learn not just the starting state of a cell and its ending state, but the actual journey it took to get there. It is the difference between reading a summary of a movie and watching it play out in full.
WHY IT MATTERS
Drug development today is a game of trial and error where the odds are stacked against us. Nearly nine out of ten drugs that show promise in animal tests fail once they reach humans, leading to billions of dollars in wasted research and years of lost time. If we can create AI that truly understands human biology—by training it on real human tissue data—we might eventually stop guessing and start designing treatments that are likely to work before they ever reach a clinical trial. This is a shift from treating biology like a mystery to treating it like an engineering challenge. We are still a long way from the medical breakthroughs we hear about in headlines, but we are finally starting to build the tools that might make them possible.
Liked this one? The next lands at breakfast.
Every story in tomorrow's AI news, rebuilt in plain English — five minutes, sources linked, free forever.
By joining you agree to receive Article's daily newsletter — unsubscribe in one click. Privacy