A handful of the most influential people in artificial intelligence have just walked away from their long-term home at Google to start a new company called Discovery Loop. Jeff Dean, a legendary engineer who was one of Google’s first 30 employees, is leaving along with three other high-profile researchers who helped build the systems that define modern AI today.
The team is building a startup focused on automating scientific discovery. In a traditional lab, researchers come up with a hypothesis, run an experiment, analyze the data, and start over. Discovery Loop wants to replace that slow, human-led cycle with AI that can perform thousands of experiments simultaneously. They are starting as a small group, but they have already secured funding from major venture capital firms and even Google itself, which remains a partner and investor in their project.
Making the machine the researcher
To understand Discovery Loop, think about how an AI currently works. Most people know AI as a tool that answers questions or writes text by predicting what comes next based on patterns it learned from vast amounts of data. This is useful, but it is passive. It waits for a human to give it a prompt.
Discovery Loop wants to turn this model into an active researcher. They are building systems that act like a scientist in a loop. The AI proposes a theory, designs the physical or digital test needed to prove it, runs the test, observes the results, and then learns from what happened to refine its next attempt. This is a leap from simple data processing to true scientific inquiry. Instead of just organizing information for us, the AI is tasked with generating new ideas, testing them against reality, and iterating until it finds a solution that works.
The departure of these leaders is significant because they are not just experienced employees; they are the architects of the technology that powers Google’s current AI tools. By leaving to form a new company, they are betting that a small, focused team can move faster than a tech giant. Large companies have massive amounts of data and computing power, but they also have internal inertia that can slow down radical new ideas.
The bigger question is whether they can actually teach an AI to make creative, scientific discoveries. If they succeed, it could mean much faster breakthroughs in complex fields like drug discovery, material science, and clean energy. If they struggle, it highlights just how difficult it is to move AI from being a clever assistant to an independent engine of scientific progress. We are about to see if the best way to invent the future of AI is to have the machine do the inventing itself.
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