California has problems. AI can help solve them.

UC Berkeley scientists are testing whether sprinkling just a few AI-equipped vehicles into the rush hour mix can ease these phantom jams, making your commute faster, safer and less energy-intensive. Over one work week in 2022, human drivers piloted a hundred AI-augmented vehicles onto a chronically congested stretch of Interstate 24 in Nashville during the morning commute. Each vehicle in the AI fleet gathered data about and responded to traffic in its vicinity and was also networked with all of the other cars in the fleet.

It worked. "Preliminary results suggest that, even with a small proportion of these vehicles on the road, we can effectively change the overall behavior of traffic," said Alexandre Bayen, director of the Center for Information Technology Research in the Interest of Society and the Banatao Institute (CITRIS). Bayen led the study, which was funded by the National Science Foundation, the U.S. Department of Transportation, the U.S. Department of Energy and the Federal Highway Administration, among others. "The game changer here was the coordination - the fact that the vehicles leverage each other's presence and can react preemptively to downstream traffic conditions and improve overall fuel efficiency of traffic on the freeway."

Discover more ways UC researchers are using AI to make it easier, safer and more fun to get around the Golden State.

• AI designed by programmers at UC Berkeley rewards carpoolers with more green lights, incentivizing more people to ditch their solo rides or opt for public transit.

• Electrical engineers at UC Merced are testing AI's ability to backstop air traffic controllers, improving safety in fast-moving, high-stakes situations.

• Students and faculty in the Autonomous Vehicle Lab at UC San Diego are getting ready to launch a fleet of scooters that will be able to drive themselves to areas with high demand.

Helping prevent homelessness

According to recent research from UC San Francisco's Benioff Homelessness and Housing Initiative, most unhoused Californians say a bit of financial help at the right time would have been all they needed to hang on to the housing they'd lost. Public agencies provide that kind of support, but they've struggled to get it out to people who are at the highest risk of becoming homeless - in part because it's hard to know who falls into that category.

"Here in Los Angeles, you might have several million people on public assistance, all of whom seem vulnerable. But in any given year, only one to two percent of them will enroll in homelessness services for the first time," said Janey Rountree, executive director of the California Policy Lab at UCLA. With the predictive capabilities of AI, Rountree's team is helping Los Angeles County identify the people for whom a little bit of assistance will make the biggest difference.

A UCLA medical student and a patient at a mobile health clinic in downtown Los Angeles. UCLA California Policy Lab Director Janey Rountree has developed a machine learning algorithm that can help public service providers identify the Angelenos who are at highest risk of losing their housing, so agency staff know who to contact to offer additional support.

Her research group uses anonymized records from a half-dozen county agencies, reflecting emergency room visits, mental health and substance use disorder care, and use of public benefits. They've created an algorithm that predicts which people are at highest risk of experiencing homelessness in the near future. Rountree says the algorithm is more than three times better at predicting a person's chances of becoming homeless than a random guess. These are the folks that agency staff track down to offer tailored support, such as short-term funding to cover expenses and help people stabilize their housing situation.

Since the pilot launched in 2021, Los Angeles County has offered support to over 700 Angelenos who are at high risk for homelessness. Eight-six percent of them remain housed.

In 2021, the University of California became the first in the nation to adopt recommendations for ethical development and implementation of AI. Today, UC's AI Council is translating those recommendations into practice with training and tools to help staff, faculty and students judge the technology's opportunities and risks. These resources will be available later this year.

Read "Responsible Artificial Intelligence: Recommendations to guide the University of California's artificial intelligence strategy"

Growing more and better food

Humans started tinkering with genes way back in Mesopotamia, when early farmers started selecting the seeds of the tastiest plants from the harvest to plant the next season. Despite 10,000 years of advances in our understanding of genetics, it's still time consuming to untangle the complex, nonlinear relationships between a plant's genes and the traits that emerge as it grows.

"We can get gains more quickly if we can accurately predict how the next generation is going to perform, or how a given crop variety might perform in a new environment where we haven't tested it before," said UC Davis Plant Sciences professor Christine Diepenbrock. She's a researcher with the AI Institute for Next Generation Food Systems, a six-university consortium based at UC Davis funded by the U.S. Department of Agriculture and the National Science Foundation.

Diepenbrock uses artificial intelligence to accelerate the age-old process of trying to grow more and better food with fewer resources. In a 2023 study, her lab used machine learning to pinpoint the genes that boost levels of provitamin A carotenoids in maize. This vital nutrient is scarce in most strains of the crop, which is part of the reason that vitamin A deficiency is most common in parts of the world where maize is a dietary staple. Vitamin A deficiency is linked with vision loss, eczema, infection, infertility and more.

In general, the oranger the corn, the higher its concentration of provitamin A carotenoids. Christine Diepenbrock's lab is using machine learning to identify the genes that control provitamin A carotenoids in maize, which could ultimately help reduce disease in parts of the world where maize is a menu staple.

Discover more ways UC scientists are using AI to benefit growers, distributers, and eaters across the state.

• A UC Riverside study applies machine learning to help identify which molecules contribute to our favorite smells, with possible applications in flavors and fragrances.

• A UC Merced engineer is designing a smart robot to ease the physical burden of strawberry harvesting.

Diagnosing Alzheimer's in time for early intervention

An Alzheimer's drug called donanemab recently cleared another hurdle on the path to F.D.A approval. An expert panel determined that the new treatment's benefit - a modest slowing of cognitive decline - outweighed its risks. These benefits were most pronounced for patients in earlier stages of the disease, a finding that underscores the life-changing importance of catching Alzheimer's as soon as possible.

Now, UC San Francisco scientists are using AI to predict Alzheimer's disease years before symptoms appear. Alice Tang, an M.D./Ph.D. student in the Sirota Lab, wrote an algorithm to analyze the clinical records of 5 million patients, comparing those who eventually developed Alzheimer's with those who didn't. The computer looked for patterns in other conditions, such as hypertension, high cholesterol and vitamin D deficiency. Tang's algorithm was 72% accurate at identifying patients who'd go on to develop Alzheimer's, up to seven years before symptoms appeared.

Alice Tang, left, created a machine learning algorithm that can identify patients at risk of Alzheimer's disease by analyzing patterns in medical records.

"This is a first step towards using AI on routine clinical data, not only to identify risk as early as possible, but also to understand the biology behind it," said Tang of the National Institutes of Health-funded study. "The power of this AI approach comes from identifying risk based on combinations of diseases."

Tang is one of many UC researchers who are using AI to get diagnoses quicker, cheaper, more accurately, and even more equitably. Studies show promise for conditions ranging from heart disease to cancer to traumatic brain injuries and many more.

Explore more ways UC researchers are using algorithms to superpower health care and medical research in California and around the world:

• Doctors and computer scientists from UC Riverside, UC Santa Barbara and UC San Francisco found that ChatGPT could provide an early warning of outbreaks of contagious disease by scanning tweets.

• UC Riverside scientists used machine learning to discover drugs to treat COVID-19.

Alerting first responders to wildfire

As climate change fuels more and bigger fires statewide, rapid response from firefighters has never been more important. Now, a UC San Diego program called ALERTCalifornia is bringing the power of artificial intelligence to assist the age-old job of fire spotting.

UC San Diego scientists utilized 20 years of archival footage of wildfires and combined it with CAL FIRE's expertise to train computers to recognize smoke and flames within the network's camera views. Now, they're gathering data from cameras aimed at fire-prone areas throughout the state. When smoke appears in a frame, the system alerts firefighters, so they can confirm ignition and start making decisions about how to respond sooner. ALERTCalifornia's cameras never sleep and can spot signs of fire 60 miles off during the day and 120 miles away on a clear night. With more than a thousand cameras distributed throughout the state, the system has also helped first responders and scientists monitor mudslides, atmospheric rivers and even the elusive, endangered California condor.

UC San Diego developed a computer program that can spot wildfire smoke and alert first responders. Courtsey: ALERTCalifornia.

CAL FIRE piloted the AI tool in six regions starting last summer. It worked so well, in some cases tipping firefighters off to a new ignition before the first 911 call came in, that the agency expanded it to all 21 of its units within a few months. And this summer, CAL FIRE added feeds from ALERTCalifornia's cameras to every fire incident page, giving everyday Californians up-to-date information on fires that could affect their communities.

Check out more ways UC is using AI to help California deal with climate change:

• UC Santa Barbara scientists developed a computer that can detect methane, an invisible but potent greenhouse gas that often goes underreported in climate science

• Meteorologist at UC Irvine are using AI to forecast clouds - a notoriously difficult art, but one with big implications for climate research.

Preserving California's biodiversity

California is home to more kinds of birds than any other state, from the teeny Calliope's hummingbird to the colossal condor and over 700 species in between. As climate change and development rearrange wildlife habitat and threaten bird life here and around the world, the importance of keeping tabs on how species are faring only grows. But traditional bird surveys, carried out by people hiking into bird territory and keeping their eyes and ears out for calls or sightings, are a big job.

UC Merced Computer Science and Engineering Professor Shawn Newsam and Electrical Engineering and Computer Science graduate student Shrishail Baligar used a type of artificial intelligence called a convolutional neural network to make bird surveys more efficient and accurate. With funding from NASA and others, bird experts trained an algorithm to recognize a bird's call. Then Newsam and Baligar fed the algorithm over 8,000 hours of audio from 600 recorders distributed around Sonoma County. Much like a skilled human birder can, the algorithm learned to parse and identify many of the overlapping calls the recorders picked up, eventually detecting 54 species.

Scientists at UC Merced deployed 600 cheap, compact and rugged audio recorders around Sonoma County. A computer later scanned the audio from these recorders and parsed out bird species by their calls. Credit: UC Merced

Unlike human birders, each recorder stayed in the field for 2 weeks straight, improving any one microphone's chance of picking up whatever birds were nearby. And the tiny microphones didn't alter the birds' behavior like the presence of humans can, another factor than can affect the accuracy of a survey's data.

UC researchers are using AI to help Californians experience and understand the natural world.

• A linguist and a computer scientist at UC Berkeley are part of the international team using machine learning to decipher whale song, thought to be the largest effort at interspecies communication in human history.

• At UC Santa Cruz, computer scientists developed a program to help lifeguards monitor rip currents and prevent drownings at the beach.