← The Vault
Everyday AI

Why AI is changing how we discover new medicines

In medicine, AI is starting to act as a digital scientist, helping researchers design better drugs and predict how they will perform in the human body. Unlike the AI tools we use for writing or images, medical AI cannot simply learn from public internet data. To be useful, these systems must be built on private, protected sets of biological information, creating a tension between proprietary silos and the hope for a more collaborative future.

Edition № 491Room: Everyday AI29 August 20262 min readSources: 1
Article

Investment expert Vijay Pande, who helped pioneer the use of AI in medicine at a major venture capital firm before launching his own, is narrowing his focus. Instead of spreading his bets across dozens of companies, he is now prioritizing a tiny handful of high-potential ventures. His work centers on a major shift in how we approach disease: moving from accidental discovery to intentional engineering.

WHAT'S HAPPENING

Scientists are using computers to map out the complex mechanics of diseases. They aim to identify exactly what a drug should target and how to build it. A significant part of this work involves clinical trials, the expensive and lengthy process of testing new drugs on people. Currently, many drugs fail in these trials because they were tested on animals, which often react very differently than humans. AI is being trained to predict human responses more accurately, potentially reducing both the cost and the failure rate of developing new life-saving treatments.

The tension of private data

HOW IT WORKS

When you use tools like ChatGPT, the underlying model—the brain of the system—learned by reading massive amounts of text from the public internet. It essentially memorized patterns in human language. Biology does not work that way. There is no open, digital library of the human body that an AI can simply download and read. Because biological data is highly complex and often sensitive, every company developing these technologies must create its own private collection of information, known as a dataset. They have to run their own physical experiments or analyze proprietary medical records to feed their models. This creates a difficult bottleneck; because this data cannot be easily scraped, it stays locked inside private, walled-off silos. Many in the industry aspire to build shared, foundational models that could act as universal maps of biological information, but this remains an ongoing challenge as individual companies prefer to protect their own unique research findings.

WHY IT MATTERS

The current system of medicine is often fragmented, with specialists focusing on single organs or diseases without seeing the full picture of a patient’s health. AI offers a different possibility: a system that acts like a team of the world’s best doctors collaborating at once. By comparing a person’s unique biological data against advanced models rather than just general population averages, doctors may one day stop guessing which medicine will work. Instead, the goal is to get the right treatment to the right person on the first try, turning medicine into a precise, individual-focused science.

Sources
← PreviousRunning Your Own Private AI at HomeNext →Why AI security experts are worried about cyberattacks
Tomorrow's edition · free

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

← Back to the Vault