How UC scientists are putting AI to the test

Nervous about artificial intelligence? You're not alone: half of Americans now say they're more concerned than excited about the ways AI is changing daily life. But most people surveyed also recognize that AI is changing some things for the better. Take science: with its ability to detect never-before-seen patterns in complex systems, AI is speeding the search for everything from new medicines to new galaxies.

We talked to researchers across the University of California who are using AI to advance health, energy, agriculture and meteorology. It turns out they're as skeptical of AI as anyone - and that's actually a good thing. Even as they're pushing the boundaries of research and discovery, UC scientists are asking thoughtful questions about the transformation this technology brings: Can we trust AI to make high-stakes decisions? How can we ensure its benefits outweigh its risks? And how can we make sure human wisdom and well-being stays at the center of scientific progress?

Why extreme weather is tricky for AI to predict

Each time you glance at a weather app, you're tapping into one of humanity's most impressive technological feats: a global network of instruments collecting hundreds of billions of observations each day and beaming the data to supercomputers, which take hours to solve trillions of equations approximating the physics that govern all atmospheric and oceanic dynamics on Earth.

The U.S. National Weather Service launches hundreds of weather balloons every day, which contribute vital data to the billions of observations that go into creating a typical weather forecast. Credit: Caroline Brehman-Pool/Getty Images

This all makes weather forecasting one of the most computationally demanding tasks humanity routinely undertakes, says Ashesh Chattopadhyay, an applied mathematician at UC Santa Cruz. Chattopadhyay's research aims to use AI to see further into the future more accurately, using a fraction of the time, energy and computing power of today's methods.

Using supercomputers at the UC-managed Lawrence Berkeley National Laboratory, Chattopadhyay worked with NVIDIA, CalTech and Rice University to develop FourCastNet, the first AI that can go toe-to-toe with traditional forecasting methods for accuracy and range. Like ChatGPT taking in text and images from the internet to suggest a response to a user's prompt, FourCastNet looks at the past 40 years of weather and predicts what will happen next.

Because it's not solving trillions of equations from scratch each time, AI can generate a forecast in seconds instead of hours, using thousands to hundreds of thousands of times less computing power. The venerable European Centre for Medium-Range Weather Forecasting is now using FourCastNet and similar tools from Google and Huawei in its daily operations.

But recent research from Chattopadhyay's group, including collaborators at the University of Chicago and NYU, suggests it's probably too soon to turn forecasting over to AI entirely.

"AI works great for day-to-day weather over, say, Houston," Chattopadhyay says. "But what about when Houston is facing something that's never been seen in recorded history, like Hurricane Harvey?" The 2017 storm dumped over five feet of rain on parts of south Texas, a once-every-two-millennia event.

The crew of the International Space Station snapped this photo of Hurricane Harvey as it barreled toward the Texas Gulf Coast in September, 2017. Credit: NASA / Getty Images

"The fact that a storm like that can happen is embedded in the physics of the system, so the traditional physics-based models predicted it, which is the great thing about them," Chattopadhyay says. If an AI model is only trained on data going back 40 years, would it have been able to predict Harvey?

To find out, Chattopadhyay's group trained a version of FourCastNet using decades of weather observations, but filtered out all the hurricanes stronger than a Category 2. Then they fed it an atmospheric condition that they knew would generate a Category 5 hurricane in a few days. The AI model clocked the storm, but seriously underestimated its intensity, predicting it would top out at a Category 2.

"We found that it couldn't really extrapolate beyond what it had seen in its training data," Chattopadhyay says. "Despite how good these models are with routine weather, getting the extremes right is still a problem. And those extremes are actually the thing scientists and forecasters care most about."

Paul Morris checks on neighbors homes in a flooded district of Orange as Texas slowly moves toward recovery from the devastation of Hurricane Harvey on September 7, 2017 in Orange, Texas. Credit: Spencer Platt / Getty Images

The study flagged an important consideration for meteorologists integrating AI into their predictions, and pointed Chattopadhyay and his colleagues towards their next task for improving AI forecasts. Now they're experimenting with integrating algorithms designed to model longer-term climate trends into the shorter-term pipeline for forecasting weather, an approach that seems to boost the capacity of AI to foresee dangerous, unprecedented storms.

More from UC Santa Cruz: AI is good at weather forecasting. Can it predict freak weather events?

AI powers progress in brain-computer interfaces

Can AI learn what farmers know by instinct?

As director of the Digital Agriculture Laboratory at UC Davis, Alireza Pourreza has spent a lot of time tagging along with Central Valley farmers as they walk around their fields during growing season. "Some of these guys can take one leaf off a tree and hold it up to the light and they just seem to know if the plant needs more or less water or fertilizer. It's like magic," Pourreza says. "But this is not something everyone can do."

For everyone else, Pourreza's team has developed Leaf Monitor, an AI system that scans a leaf and returns a real-time snapshot of plant health. "Our system is based on recent advances in optic knowledge of how different elements absorb and reflect light in characteristic ways," Pourreza says. It uses a handheld spectrometer to measure light outside the range visible to humans, then cross-references those measurements with the known spectral signatures of ten different nutrients that affect crop health, including nitrogen, magnesium and potassium.

Leaf Monitor offers big advantages over the standard approach to monitoring crop health during growing season, which involves sending samples into a lab for chemical analysis. Results can take weeks to come back, "and by the time farmers get the data, that critical moment to make a decision has probably already passed," Pourreza says. So farmers tend to do more guessing than they'd like, erring on the side of adding fertilizer or water where they might not actually be needed.

"If we're going to move towards sustainable, regenerative agriculture that has less impact on the environment, we need to be able to manage in a real-time, site-specific manner," Pourreza says. "Leaf Monitor is an example of a tool that can help us get there."

More from UC Davis: AI Tool to Help Farmers Measure Real-Time Crop Health from the Field

AI tools promise a more accurate cancer diagnosis, but how good are they?

Each year, about 40 million women in the U.S. get a mammogram to screen for breast cancer, and a growing share of these mammograms are read with help from artificial intelligence. "AI in clinical mammography is not quite the standard practice yet, but it's expanding very fast," says Hannah Milch, associate professor of radiology at UCLA. The FDA has cleared a half dozen commercial algorithms for clinical use in recent years, and patients are increasingly likely to seek out the clinicians who use these tools.

Yet the evidence that AI-assisted radiology actually improves patient care or outcomes is surprisingly scant. "These tools advertise that they'll help us catch more cancers and make us radiologists do our jobs better, but the evidence behind those claims is based largely on someone reading a couple of hundred mammograms in a lab setting. That's very different from the daily work of interpreting thousands of scans in the real world," Milch says. "What happens when you take those products and apply them to a general population? Do they work as advertised?"

Dr. Hannah Milch studies what happens when AI radiology tools developed in a lab are introduced for clinical use. Credit: UCLA Health.

In a study published in May 2025, Milch investigated whether an AI radiology tool called Transpara could detect breast cancers that human radiologist missed while reading routine screening mammograms. Studying nearly 185,000 mammograms conducted between 2010 and 2019, her team found that Transpara spotted about 30 percent of these so-called interval cancers.

"It's encouraging to see that if that radiologist were in that situation again, and they had the AI support tool flagging that spot, that cancer may have been caught five, six, eight months earlier, when it's easier to treat," Milch says.

Now Milch is co-leading a larger study in which over 400,000 mammograms will be randomly assigned to be read either by a radiologist on their own or with assistance from Transpara. The PRISM trial, also led by Milch's UCLA Health Jonsson Comprehensive Cancer Center colleague Dr. Joanna Elmore, Dr. Diana Miglioretti at UC Davis and Dr. Christoph Lee at the University of Wisconsin-Madison, is the nation's largest-ever randomized prospective trial of AI in breast cancer screening. "We'll be looking to see if patient outcomes are actually better when the radiologist had AI helping them," Milch says. "We want to make sure that these tools are safe and effective before they're widely bought up and put into clinical use."

More from UCLA: AI could help improve early detection of breast cancers between screenings, UCLA-led study shows

4 ways artificial intelligence is poised to transform medicine

Can AI turn fracking tech into a tool for clean energy?

People have been tapping into geothermal energy for millennia, finding hot spots in the Earth's crust to warm their homes or slip into soothing natural hot springs. But the use of geothermal energy has long been limited to these rare places where heat from the Earth's molten core naturally rises to within a few hundred feet of the surface.

"If you drill down deep enough - three, four, even six kilometers - you reach the point where the crust is uniformly hot," says Mohammad Javad Abdolhosseini Qomi, UC Irvine associate professor of civil and environmental engineering and materials science and engineering. "If you can get down there, you can pipe in cold water, and the crust will heat the water up so you can run turbines and generate electricity."

These so-called enhanced geothermal systems could generate tons of sustainable energy 24 hours a day, regardless of weather - enough to power 40 to 70 million homes across the U.S. Thanks to advances in drilling technology pioneered by the fracking industry, engineers now have reliable methods to reach kilometers into the crust, meaning a future powered by sustainable geothermal energy is within reach.

But Qomi says there remain "some very significant challenges" Engineers will need to safely operate complex machines miles underground, where pressure, heat, fluid and chemistry interact in ways that scientists are only beginning to understand. (We know just enough to know that injecting lots of fluid that in the Earth's crust can cause what's known in the business as induced seismicity: Between 2008 and 2019, Oklahoma saw a 900-fold increase in earthquakes as fracking expanded throughout the state.) These are the kinds of problems that artificial intelligence, which can model and make predictions about overlapping forces, is primed to help solve.

With a $6 million grant from the UC Office of the President, Qomi is leading Geophysicist.ai, an effort to blend large language models, physics-based models, and real-world data from geothermal drilling sites across the Western U.S. into an AI that will give engineers a much sharper understanding of the conditions and risks of drilling in a particular area. That will help with site selection for geothermal development and reassure people who live close by that the projects are safe.

More from UC Irvine: UC Irvine to lead use of AI in solving grand challenges below Earth's surface

People-centered progress

Artificial intelligence is still a new technology, but at UC it's guided by the same ethic of rigorous science for public benefit that has steered our world-class research enterprise for over 150 years. These are just a few of the thousands of ways that UC is leading the world in AI-powered discovery - not just by pulling off impressive technological feats, but by ensuring new tools work safely and benefit people in the real world.