In a cramped, windowless room at UC Berkeley, two identical microscopes run around the clock, each one a machine that has fundamentally reshaped how biologists see living cells. These are MOSAIC microscopes—the name stands for Multimodal Optical Scope with Adaptive Imaging Correction—and they represent a quiet revolution in cellular science that's quietly spreading around the world.
The breakthrough lies in what MOSAIC does: it combines twelve separate high-powered imaging technologies into a single switchable system, from standard phase contrast to the latest lattice light-sheet technology. Described this week in Nature Methods, the microscope has already been replicated in more than a dozen labs worldwide, thanks to detailed assembly instructions researchers have shared over six years. But at Berkeley's Advanced Bioimaging Center, MOSAIC is just one piece of something far more ambitious—an effort to create what they're calling a Cell Observatory, complete with AI powerful enough to make sense of the flood of data these machines produce.
The data itself is staggering. MOSAIC can track live specimens over seconds, hours, or days—from individual molecules and cells to entire embryos—capturing how cells move through tissue, how internal cellular structures evolve, and how proteins shuttle within the cell. The result: petabytes of information. For context, that's equivalent to roughly 500 billion pages of text. No human researcher can parse that alone.
This is where the AI comes in. Eric Betzig, a UC Berkeley professor of molecular and cell biology and of physics who won the 2014 Nobel Prize in Chemistry for developing super-resolution fluorescence microscopy (a technique now built into MOSAIC), explains the fundamental problem his team is trying to solve: "We are the world's best at collecting data at 5D, and have been for a decade. But we don't know how to interpret the data at scale; we can't think in petabytes and we don't see in 5D. That's why we're developing a 5D AI—it's a sherpa to guide us."
That 5D dimension is crucial. The microscopes capture three spatial dimensions, plus time and color—the latter coming from fluorescent labels that let scientists track multiple cellular structures simultaneously as they migrate, change shape, divide, and interact. Betzig notes that understanding cells holistically requires seeing them in living tissue over both fast and slow timescales. "You can't study something as complex as a cell or organism just by looking at the parts individually," he says. A single human cell contains roughly 40 million protein molecules of 20,000 different types—a complexity no traditional analysis can untangle.
Srigokul Upadhyayula, the assistant professor in residence in molecular and cell biology who led MOSAIC's development alongside Betzig, frames the challenge plainly: "Biology is entering an era in which the data are too complex and too large to interpret by human inspection alone." His vision is to build what amounts to a specialized ChatGPT for biology—an AI that can reason natively with 3D videos of living systems and let researchers query those dynamics through language rather than traditional computational tools.
The two microscopes running in that Berkeley room aren't just collecting data for curiosity's sake. They're gathering the foundation for what could transform how biologists understand life itself, creating a new era where machines help us see what's always been there but was simply too intricate for human minds alone to comprehend.