Entering the New Era of Medical Discovery
Between us, we have six children. The oldest is 12. None of them has a phone yet, and none has asked us whether artificial intelligence is going to ruin their lives. They will. They will learn that some of the people building the most powerful AI systems have warned that the technology could become dangerous in ways we cannot yet control. They will hear predictions that many of the jobs they might want in the future could disappear. They will face heated debates about truth, power, safety, and our institutions' ability to keep pace with machines advancing faster than our capacity to govern them.
We cannot tell our children all these fears are unrealistic, because we don't believe that ourselves. Still, another danger looms over this historic moment. As public conversation about AI grows darker, we risk being so absorbed by what might go wrong that we lose sight of what is beginning to work well. This is not an argument for complacency. The risks are real and demand immediate action. It is, however, an argument for what we believe becomes equally necessary: pragmatic optimism.
We operate in one of the most failure-prone domains existing, trying to understand disease and determine which medicines might work on human beings. From our position on the front lines of drug investment and creation, we're beginning to see how artificial intelligence is changing what's possible. Drug development has always been a brutal business. Historically, about nine out of ten therapies entering human testing fail. A successful drug can require a decade or more and billions of dollars to develop. Of course, these metrics hide the real cost. A failed or missing drug is a patient who continues to wait.
For most of modern drug development history, researchers started from a biological hypothesis, sought a molecule that might affect it, tested that molecule, and then, in crushing fashion, discovered they had deceived themselves. Artificial intelligence is beginning to change this story. We've had the privilege of watching, almost firsthand, this shift toward discovery and human understanding amplified by artificial intelligence. Recently, we followed a research team using artificial intelligence to design antibodies from scratch against difficult medical targets, unlocking approaches that have blocked some of the most sophisticated groups in the pharmaceutical industry. Nearby, another team of just five scientists used an AI-driven experimental system to go from an idea for a new difficult drug to demonstrating what we believe are unique capabilities for treating cancer in primates in less than six months. Programs with such ambitions would conventionally require teams of specialists working for years.
We operate in one of the most
What is changing isn't merely the tools available to help us predict human biology. It's the speed of the discovery process itself: I propose an idea, build it, test it in the physical world, learn what happened, and start again. The scientific method itself is beginning to accelerate, and the number of scientific investigations, opportunities to improve health per dollar spent, is beginning to expand. Last month, Merck and Moderna reported that a personalized cancer vaccine for melanoma significantly delayed the disease's spread. The result is remarkable in itself, and studies are already underway in other contexts. What seems particularly encouraging is how this growing clinical evidence combined with artificial intelligence tools and automated laboratories could open the field to more competitors. Evidence reduces uncertainty about approach, while these tools can reduce costs and time needed to follow it.
Together, they can bring more teams and more capital into our fight against cancer and expand the number of promising ideas that will be tested at an unprecedented rate. We are just getting started.
Anyone claiming artificial intelligence has solved drug discovery is selling something. Of the 117 AI-assisted drug programs in clinical trials last year, at 63 companies, none has been approved. Biology has humbled generations of brilliant scientists, and artificial intelligence will not transcend biology. Models will deceive. Drugs will fail. Patients will react in unanticipated ways.
Yet we see green shoots and a remarkable recalibration of direction, not destination. Direction is the whole game in fighting disease and improving human health. Artificial intelligence is beginning to give us better ways to find it. That is the source of our optimism. Not a belief that the surrounding dangers of artificial intelligence are exaggerated, or that medicine is becoming easy. We don't believe either. It comes from observing a technology beginning to alter a field where significant progress has historically been excruciatingly slow.
Our children will inherit all of this. They will live with the disruptions we debate now and with others that none of us anticipated. It is time for leaders who take those risks seriously and institutions willing to act before every consequence is known.