Demis Hassabis and the AI Quest to Master Our Molecular World
On a cool October evening, Demis Hassabis and I discussed the stars.
This setting was fitting. Firstly, our location: we were in a London observatory that first opened its doors in 1929. Secondly, the metaphor: discussing stars and constellations is an excellent method to explore concepts of immensity, and the marvels and constraints of human comprehension.
Yet there exists an immensity much nearer to us that surpasses even the stars. It might appear unbelievable, but the number of potential chemical compounds in our world actually exceeds the number of stars in the sky. The difference is staggering: conservative figures place the quantity of small, drug-like molecules at roughly 10^60, whereas stars in the observable universe number around 10^22 (or possibly 10^24 by some counts).
This fact has captivated me for months, as it illustrates the profound challenge of drug discovery. I was aware, for instance, that penicillin was discovered by chance. However, I hadn’t fully grasped that every successful drug is essentially a miracle, a remarkable victory against formidable scientific odds.
This week, a magazine feature I’ve contemplated since autumn was published, offering an insider’s view of Hassabis’s Isomorphic Labs. Hassabis, a pioneering Nobel Laureate, founded Isomorphic in 2021 to apply the capabilities of artificial intelligence to the enormous landscape of chemical space, with the ultimate goal to “solve all disease.”
It’s important to clarify what that ambitious slogan truly signifies (and I arrived in London fully intending to question Hassabis rigorously on this point). As I noted:
When I raise this with Hassabis, he explains his philosophy: The concept of “solving disease” is more expansive and pragmatic than eradicating illness permanently. He deliberately avoids the word “cure” for a reason. While one cannot guarantee that people will never fall ill again, he argues, it is possible to create a systematic, repeatable, and scalable process—driven by advanced AI and technological platforms—for the discovery, design, and optimization of drugs or treatments as they are needed.
Therefore, it is less a slogan and more of a guiding theory, a mission, and in a way, a pledge of technological potential. A drug represents more than just a compound; it could be an improved therapy for a stubborn condition, or it might be a cure. Undoubtedly, it equates to more time for us or for those we cherish. Isomorphic does not currently have a drug in clinical trials (and remains vague about the precise timeline for this milestone). However, the ultimate measure of success is unequivocal: a transformative drug that makes it to patients.
“To demonstrate the value of a venture like this, you must provide concrete proof,” Google Ventures managing partner Krishna Yeshwant, a physician-turned-investor involved in Isomorphic’s early discussions, told me during my reporting. “You must discover your own drugs. You must shepherd them to the public and validate their efficacy.”
Isomorphic, along with the broader field of AI-driven drug discovery, is stepping into a new phase—one that requires tangible outcomes. Should these outcomes be achieved, they will deliver more than just enhanced treatments for autoimmune diseases or cancer.
In the years ahead, if Hassabis’s vision proves correct, a completely new system will arise. One that holds the promise of, quite feasibly, harnessing the vast complexity within each of us.
See you Monday,
Allie Garfinkle
X:
Email:
Submit a deal for the Term Sheet newsletter .
Joey Abrams curated the deals section of today’s newsletter. .