UC Merced Magazine | Volume XX, Issue VI

MAGAZINE

THE artificial intelligence ISSUE

What can AI teach us about ourselves? Page 19

Fall/Winter 2024 Volume XIX, Issue 6

MESSAGE FROM THE CHANCELLOR

Chancellor Muñoz takes a selfie with new Bobcats at this year’s Scholars Bridge Crossing.

T he transformative potential of arti cial intelligence has become increasingly evident across numerous elds, including research at UC Merced. AI's capability to analyze vast datasets, identify intricate patterns and make predictions has greatly accelerated the pace of discovery in a growing number of disciplines. In health care, AI analyzes medical images, forecasts patient outcomes and even discovers new pharmaceuticals. In astronomy, AI algorithms are si ing through enormous volumes of data to pinpoint celestial objects and phenomena. In agriculture, AI is again helping analyze big data and is a critical part of remote sensors and automated systems. Moreover, AI is streamlining research methodologies by automating repetitive tasks, allowing researchers at UC Merced to concentrate on more innovative and analytical aspects of their work. As you will see in this issue, AI has crossed all disciplines at this university. But as with any new technology, there comes a litany of issues and challenges that require thoughtful and consistent attention. As teachers, faculty members must deal with whether and how to incorporate AI into lessons and assessing student work. Students must wrestle with ethical issues and creative challenges and think about their futures and how AI will t into their careers. Faculty members and students creatively analyze the best ways to incorporate AI into the important research they do, and how this technology might enable them to push the envelope of their research. As AI and related generative platforms advance, their in uence on research at UC Merced is poised for further growth, potentially leading to breakthroughs and innovations that will positively impact life here in the Valley, across the state and around the world.

Fiat Lux, Juan Sánchez Muñoz, Ph.D. Chancellor

UC MERCED MAGAZINE // ucmerced.edu

ON THE COVER: AI tools in ChatGPT and Adobe Photoshop generated this interpretation of Michelangelo's famous painting on the ceiling of the Sistine Chapel, "The Creation of Adam."

Fall/Winter 2024 Volume XIX, Issue 6 MAGAZINE

In is Issue:

Chancellor Juan Sánchez Muñoz Vice Chancellor/ Chief External Relations Officer E. Edw. Klotzbier Assistant Vice Chancellor and Senior Advisor, External Relations Danielle Armedilla

4

Donor Impact: The Boldly Forward campaign is benefitting students and the Valley Opinion: How advances in technology can benefit agriculture and the Earth

Editor Senior Writer and Public Information Officer Lorena Anderson Writers Francesca Dinglasan Patty Guerra

6

Alyssa Johansen Andrew Mitchell

Jody Murray Brenda Ortiz Sam Yniguez

AI or human-made? Page 28

Photography & AI-generated Images Veronica Adrover Melvin Diaz Alfaro ChatGPT Dall-E Ahmed Gaber Erik Jepson

14

Faculty Profile: Professor Shijia Pan is making the world a smarter place Cover Story: What AI can teach us about ourselves AI and the Arts: Is it art if a machine creates it? Sports: UC Merced has joined the NCAA

Jody Murray Zachary Silva Danay Suárez Sam Yniguez

19 26 29

Graphics / Design / Printing Colemar Design Dumont Printing, Fresno Thomas Hudelson Elizabeth Lippincott Mailing Address UC Merced External Relations 5200 Lake Road Merced CA 95343

This season’s arts offerings, Page 12

7

Three Essential Reads: Be inspired to explore the universe with books our astrophysicists love

ucmerced.edu | @ucmerced

Every effort has been made to contact copyright holders of any material printed in this magazine. Any omissions will be righted in subsequent issues if notice is given to the editors.

3

UC MERCED MAGAZINE // ucmerced.edu

DONOR IMPACT

Professor Abel Chuang, center, and his student research team

Boldly Forward Fundraising Campaign Tops $100 Million As the rst comprehensive fundraising campaign in UC Merced’s history approaches the halfway point, the tremendous generosity of the university’s donors has put the campus on pace to reach its $200 million goal by June 2030. Since the launch of Boldly Forward: T he Campaign for UC Merced, more than $102 million has been raised, representing more than 12,000 gi s from about 7,800 donors. By Francesca Dinglasan

is fundraising milestone is signi cant because the campaign’s silent planning phase began during the 2020 scal year and the height of the COVID-19 pandemic. Despite the economic challenges posed by the global crisis, philanthropic contributions from UC Merced supporters far exceeded expectations. e campaign’s public phase launched September 2022, and donor giving was so proli c that by the following month, UC Merced — encouraged by vast support that outpaced original projections — announced its intent to double the campaign’s fundraising goal from $100 million to $200 million. Funds raised through Boldly Forward bene t UC Merced’s students, faculty, programs and facilities. Each of the Boldly Forward pillars — Leading rough Discovery, Cultivating Healthy Societies and Empowering Student

Excellence — focuses on UC Merced’s vision to expand the university’s ongoing excellence and in uence not only in its native Central Valley but throughout California and beyond. “UC Merced is what it is today because of the unwavering belief and commitment of our donors,” said Vice Chancellor and Chief External Relations O cer E. Edward Klotzbier. “We are deeply grateful for their investment and look forward to their ongoing partnership.” While the UC Merced community will continue to promote the campaign and the university’s ambitious plans over the next ve years to raise the additional $98 million, gi s made during the rst half of Boldly Forward are already having a tangible impact by expanding and creating new avenues of programmatic, faculty and student support.

4

UC MERCED MAGAZINE // ucmerced.edu

model to improve communication and engagement among health professionals and the communities they serve in the San Joaquin Valley, particularly among the region’s Hispanic, Latinx, immigrant and rural communities. Local partner and campus supporter the Central Valley Community Foundation is also driving UC Merced’s e orts to address health disparities and reduce the region’s severe physician shortage. A recent grant from CVCF will assist undergraduates enrolled in the B.S.-to-M.D. Pathway by providing scholarships to these students, ensuring they can focus on their academic studies and ful ll the requirements of their rigorous coursework.

Leading Through Discovery

Empowering Student Excellence

Gi highlights in support of UC Merced’s research enterprise under the Leading rough Discovery pillar include several grants made by Portola Valley-based rm 1s1 Energy to School of Engineering Professor Abel Chuang’s work in thermal and electrochemical energy innovating fuel-cell technologies. Both Chuang and the students in his lab bene t from the funding, which provides the resources needed for their research engagement and activity. Private philanthropy in support of research also has been directed toward improving educational attainment in the Central Valley. A signi cant grant from the Bill & Melinda Gates Foundation is funding a project called Advancing Educational Opportunities in California. Led by Professor Zenaida Aguirre-Muñoz and interim Associate Vice Chancellor Orquidea Largo, the two-year study is analyzing how college-readiness data, policies and access disparities are impacting underrepresented and socioeconomically diverse student populations in San Joaquin Valley high schools. Cultivating Healthy Societies UC Merced’s determination to improve the well-being of communities through the Cultivating Healthy Societies pillar was boosted by organizational support focused on the Central Valley. Genentech, a member of the Roche group and a longtime supporter of UC Merced, provided a new round of funding to the university to create the UC Merced Center for Health Equity. Led by Project Scientist Rosa Manzo, the center is applying a community-engaged collaboration

Student support, as embodied by the Empowering Student Excellence pillar, is a top priority for UC Merced donors. Notable among the capstone gi s received during the campaign is multimillion-dollar support from a longtime anonymous donor, whose increasing philanthropy for the Gateway Scholars Program includes an endowment to provide scholarships every year to at least 36 incoming freshmen who attended regional high schools as well as direct assistance to other promising student populations, such as community college transfers.

Chuang’s lab is researching sustainable fuel-cell technology.

5

UC MERCED MAGAZINE // ucmerced.edu

AI in Agricultural Research: Cultivating a Smarter Future

OPINION

By Professor Teamrat Ghezzehei Imagine a world where every blade of grass tells a story, where the soil beneath our feet holds secrets waiting to be unlocked and where arti cial intelligence helps us feed billions. is isn't science ction — it's cutting-edge reality being forged in the labs and elds of UC Merced. But how exactly is AI revolutionizing how we understand and manage a precious resource — our agricultural lands? In recent years, the Soil & Environmental Physics research group at UC Merced has embarked on applying the latest tools in machine learning for agricultural and soil science applications. We are harnessing the power of AI and machine learning to tackle some of the most pressing challenges in agricultural research. From predicting soil moisture to unraveling the complex dance of water through earth, our work is rede ning the boundaries of what's possible in modern farming. Picture this: A drone soars over a vast grassland, its multispectral cameras capturing images invisible to the human eye. is is no ordinary survey — it's a key part of our goal to predict soil moisture in grasslands using AI. Unlike uniform crop elds, grasslands are a complex tapestry of diverse plants and uneven terrain, making traditional moisture measurement methods a real head-scratcher. Our solution? Turn the vegetation into a moisture sensor, using AI to decode

hidden messages in those aerial images. But why stop at the surface? Dive with us into the world beneath our feet, where we are using AI to predict crucial soil properties such

as saturated hydraulic conductivity. It's a mouthful, sure, but think of it as the soil's ability to play water slide — vital for understanding how water moves through the ground. Our AI models are cracking this code faster and more accurately than ever, giving us unprecedented insights into the secret life of soil. Now hold onto your lab coats, because we're about to venture into the realm of physics-informed neural networks (PINNs). Imagine AI that not only learns from data but also understands the laws of physics. It's like teaching a computer to think like a scientist and a mathematician. With PINNs, we model water ow through soils in ways once thought impossible, especially in tricky scenarios such as layered soils with poorly de ned conditions at their boundaries that are a pain to solve with conventional methods. We also are ipping the script with inverse modeling. Instead of just predicting what might happen, we use AI to work backward from observations to uncover hidden soil properties. We use our

physics-informed models to turn routine measurements of soil moisture into deeper insights into soil characteristics. As we push the boundaries of what's possible with AI in agriculture, we can't help but get excited about the future. So, what's on the horizon? Imagine farmers making pinpoint decisions about irrigation and using their limited and precious water resources more judiciously. We are setting our sights on forecasting crop water needs, integrating satellite data with our models and exploring new frontiers in AI-driven agricultural research. As we face the monumental challenge of feeding a growing global population in a changing climate, the insights we are uncovering with AI will be more crucial than ever. At UC Merced, we are not just cultivating crops; we're cultivating a smarter, more sustainable future for agriculture. Teamrat Ghezzehei is a professor of soil physics in the School of Natural Sciences.

6

UC MERCED MAGAZINE // ucmerced.edu

THREE ESSENTIAL READS | The Universe

“Contact” By Carl Sagan

“Lonely Hearts of the Cosmos: e Search for the Secret of the Universe” By Dennis Overbye Of all the books I have read, it best captures the emotion of how it feels to be a cosmologist and what motivates us to spend our lives trying to understand the universe. Also, it describes the mind-blowing discoveries made in the 20th century because of the single-minded focus of a small number of people determined to advance scienti c understanding by pushing the technologies of their day to their very limits — far beyond what most of their colleagues thought possible. ey knew they would likely not live to see the results of their labor but were driven all their lives to seek answers to some of the deepest philosophical questions. — Professor Gillian Wilson, vice chancellor for the Office of Research, Innovation and Economic Development, who studies galaxy evolution, clusters of galaxies and cosmology.

“ e Last Stargazers: e Enduring Story of Astronomy's Vanishing Explorers” By Emily Levesque e book gives a beautiful perspective on what it is like to observe the sky, in the past 100 years and at the present, and it looks forward to what we can expect in the next 100 years. I enjoyed it a lot; it inspired me because it describes what it is like to be an astronomer through some amazing stories. — Professor Sarah Loebman , whose main research focuses on galaxy evolution, clustered star formation and chemo-dynamics in the local universe.

“Contact” is an excellent portrayal of a scientist. I relate to how she thinks about the world, what her job is like and how she approaches problems. And through the ctional aspects, it creates a sense of wonder about the vast possibilities of what is out there in the universe and explores how these possibilities relate to our understanding of existence. — Professor Anna Nierenberg, whose research focuses on star formation and the nature of dark matter.

Editor’s note: For this subject, we asked the entirety of our young and emerging astrophysics faculty — all three of them — to recommend a book about science and the universe. Each one o ered a book that inspired, entertained and encouraged them.

7

UC MERCED MAGAZINE // ucmerced.edu

AI Could be a Useful Tool in Classrooms and Across Campus

By Sam Yniguez Arti cial intelligence use is booming. Generative AI, including programs such as ChatGPT, are everywhere, including in classrooms. While some industries quickly adapted to this shi , the education sector’s initial response to this phenomenon was tepid, said environmental science Professor Sylvain Masclin, who facilitated a recent AI introductory workshop for students. Many unanswered questions linger, and there’s no real way to police its use, prompting some schools to ban ChatGPT. But that might have pushed students toward using it, Masclin posited. He believes that, contrary to popular belief, most students want to learn how to correctly use AI tools.

“Of course, cheating is a big worry, but cheating has always been around and, on the whole, it’s only a very small percentage of students who cheat,” said Masclin. “A new tool doesn't automatically mean that all of a sudden every student will start cheating.” Masclin said instructors are poised to educate students on best practices, and if AI plays any role in their elds, instructors should try to include it in their courses. In presentations to students and faculty, Masclin stresses AI is only one tool students should be able to access. “Imagine giving a hammer to someone who has never seen one before and asking them to build something,” said Masclin. “If you don’t teach them how to properly use it and when it's the appropriate time to use it, the consequences will be bad.”

8

UC MERCED MAGAZINE // ucmerced.edu

If you don’t teach them how to properly use it and when it's the appropriate time to use it, the consequences will be bad.

— Professor Sylvain Masclin

Zanzucchi worries about AI being used experimentally in many areas and hopes its use will retreat to areas where it is actually useful. Masclin is working on a grant with the California Education Learning Lab to create generative AI toolkits that will follow students throughout their college careers, providing guided activities and virtual tutoring. e grant would also fund faculty AI workshops for teaching and work productivity. “ e goal is to use AI to improve access and provide better learning opportunity for all students,” said Masclin. “We’re always looking at how we can improve students' learning. If While faculty consider whether and how to incorporate AI into classrooms, a similar scene is playing out in campus administrative units. e O ce of Information Technology is building infrastructure to inform many of the processes that will be implemented, said Vice Chancellor and Chief Information O cer Nick Dugan, including a website that will house information on AI tools available to the campus community and important safeguards to consider when using them. Dugan and his team plan to host a monthly engagement series, and both the website and symposiums will highlight a range of topics, from best practices for AI tools in the classroom and academic integrity to more administrative applications such as achieving operational e ciency. e work is all being done with an eye on equity and inclusivity. “We want to make sure we are not widening any gaps as we employ these tools,” said Dugan. His team will solicit a community of interest among campus leadership, faculty and sta to help shape this initiative. “ e work that I’m hoping to do this academic year is to bring people together around a common topic and facilitate discussion and ideas on how we want to incorporate this into our institutional identity,” said Dugan. “We also need to discuss how we are going to continuously change and adapt because, as you know, this eld is evolving at a very rapid pace.” AI is a tool, then why not use it?” AI Tools for Campus Use

Don’t Fear It, Learn to Use it T h e hammer analogy is one shared by economics Professor Justin Hicks. He argues that instructors who are not exposing their students to AI and having real discussions about it are doing a disservice to their students. “AI is going to take over a lot of jobs, and a big part of what we do as a university is to prepare them for success in the future,” said Hicks. “Not only do I want them using it in my class, I want them to become pro cient enough with it to be the creators of the next iteration of AI.” A proud early adopter, Hicks understands some faculty members’ hesitancy to encourage students to use AI on assignments. He said no service can predict with 100% certainty whether a piece was completely created using the tool. ere are also well-documented biases in generative programs, and ethical issues. E orts to improve AI tools make him comfortable with students using ChatGPT on assignments. “A year and half ago, I wouldn't advise students to use ChatGPT with some of the coding we do,” said Hicks. “I want students to focus on the talents and expertise they already have that cannot be mimicked by generative AI. It’s important that they be creative, and understand how to do the analysis and the synthesis. “But now, I encourage it because of the improvements I’ve seen in the output it creates. I know they'll get a functioning Keeping up to date with the latest developments in AI can be taxing. For those still on the outside looking in, Associate Dean of the School of Social Sciences, Humanities and Arts Anne Zanzucchi can see how they may feel overwhelmed and not know where to start. But there are resources, such as department chairs, the library and the Center for Engaged Teaching and Learning. “I see faculty and students running into challenges all the time, so there is a real need for open and honest conversations about AI, especially for those of us in creative areas,” said Zanzucchi. “AI has attened the landscape, so we need to talk about how that changes things from before and what the expectations are now.” code and can move on with the project.” Open, Honest Conversations

9

UC MERCED MAGAZINE // UCMerced.edu UC MERCED MAGAZINE // ucmerced.edu

Digital Assistants (Siri)

Navigation Systems

Healthcare (Disease tracking)

Home Security

Online Shopping (Recommend products)

Home Assistants (Alexa)

Search Engines

Automated Robotics (Roomba)

Manufacturing

Far From Automated Answers: Why AI Ethics Remain Ambiguous

By Alyssa Johansen

Department of Philosophy. She said when it comes to AI, the greatest ethical concerns and questions remain under debate and largely unanswered. Laws, Regulations and Accountability As uses of AI expand across countless industries, Gunn said, there are large pro ts to be made and companies have strong incentives to avoid being regulated. Current laws regarding plagiarism, copyright and intellectual property were not written with AI in mind. “When computer so ware rst came out decades ago, there was a debate over whether it was a product or an idea,” Gunn said. “Would it fall under patent or copyright? e output of generative AI, led by large language models such as ChatGPT, faces similar legal debates that a ect regulation and moral accountability.”

It’s common to fear the unknown. But when it comes to artificial intelligence, philosophy Professor Hanna Kiri Gunn said it’s what people think they know that makes them nervous. “What people are most concerned with is what they are most familiar with,” she said. “And that has a lot to do with science ction.” Movies such as “Her,” “Terminator” and “I, Robot” present audiences with highly capable, autonomous intelligences that exhibit human-like emotions and the capability for manipulative and even malevolent actions. Gunn conducts technology-related ethics research in the

10

UC MERCED MAGAZINE // ucmerced.edu

Not a lot of people are going to care if their weather report is written by a computer, but I would expect that a lot of people would not be OK with wedding vows being written by AI. — Professor Hanna Kiri Gunn

e new AI frontier means companies have weak obligations to disclose information to the government — and no obligation to disclose to the public. Furthermore, workers at AI companies are o en bound by strong con dentiality agreements and are only protected as whistleblowers when they report illegal conduct. Employees who raise red ags over ethical concerns have no protections, Gunn said, largely because there hasn’t been enough time to build legal guardrails around generative AI. For instance, many workers have said they worry their products manipulate users with authoritative-sounding responses, but these interactions are not de ned as human research or experimentation. But a psychology study that aims to change your mind in the same ways would be subject to ethical oversight, Gunn noted. AI products are immune from these research safeguards because they fall under “product development.” e question remains: Who or what is responsible when the consequences of AI aren’t as intended? “It’s one thing when a computer system messes up. It’s another when a person does,” Gunn said. “You can hold a person accountable.” In 2021, the University of California became the rst 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 sta , faculty and students judge the technology's opportunities and risks. ese resources will be available later this year. The Hidden Environmental Cost of AI An o en-overlooked consequence of AI is the environmental footprint. Gunn said water — a precious resource — is critical for cooling servers in huge data centers that provide the massive computing power AI requires. Gunn points to a study out of UC Riverside that found in 2022, the AI sectors of Meta, Microso and Google used

the equivalent of one year of Denmark’s population’s water consumption — about 2 billion cubic meters. Researchers estimate AI’s need for water could reach up to 12 billion cubic meters by 2027. It is estimated that for every 10 to 15 ChatGPT responses, a computer system “drinks” a 500-ml bottle of water. A Need for Transparency? Gunn said when considering ethics, honesty and transparency aren’t the “end-all, be-all.” However, when it comes to AI and plagiarism, “it matters.” “Not a lot of people are going to care if their weather report is written by a computer, but I would expect that a lot of people would not be OK with wedding vows being written by AI,” Gunn said. “If you ask somebody, ‘Why do you love me?’ and they ask a computer to reply, that person is not telling you what they value about you,” she said. “What that model will give is some assortment of what it’s been trained to o er as responses to questions about why someone deserves to be loved.” Generative AI products learn by scraping the massive amounts of information available on the internet, everything from newspaper articles to “Pride and Prejudice” to Yelp reviews. Whether AI companies have a right to that information is a huge area of contention. Journalists, authors and musicians are examples of professions that are paid for authentic writing and composition. e New York Times, Getty Images, Universal Music and a group of authors that includes John Grisham and George R.R. Martin have led lawsuits claiming copyright infringement by AI companies. Gunn said she asks her philosophy students not to use AI for class assignments. “I want students to write essays so they can work through the di culty of developing ideas, developing arguments and working on their ability to explain,” Gunn said. “Using something like ChatGPT misses the entire learning opportunity.”

7 11

UC MERCED MAGAZINE // UCMerced.edu UC MERCED MAGAZINE // ucmerced.edu

24 25

— Danay Suárez

Season at a Glance

Happening Now

10/27 UpstART: Danay Suárez

La Galería & UC Merced Art Gallery: Alma, Corazón y Vida: Latinx Art Legends

from the Mike “Surito” Echeverría Collection Through 12/6 Spanish Program and Merced Shakespearefest : “La vida es sueño” Through 10/27

JANUARY 2025 01/30 La Galería: Opening Reception for Actions for the Earth: Art, Care, & Ecology Through 4/19

For more information, visit arts.ucmerced.edu/season

12

UC MERCED MAGAZINE // ucmerced.edu

— Merced Symphony

— Shakespeare in Yosemite

— Okorie ‘OkCello’ Johnson

— Las Cafeteras

03/15 UpstART: Las Cafeteras

04/25-27 Shakespeare in Yosemite: “As You Like It” (park performances)

FEBRUARY 2025 02/01 UpstART: OkCello MARCH 2025 03/01-10 Todo Cambia: Merced Human Rights Film Festival

APRIL 2025 04/24 Shakespeare in Yosemite: “As You Like It” (campus performance)

MAY 2025 05/31

UC Merced Children’s Opera: “Treble Trouble”

13

UC MERCED MAGAZINE // ucmerced.edu

By Brenda Ortiz Using AI to Solve Societal Problems

PROFILE

Making the World a Smarter Place

One of the PANS Lab’s projects integrates an origami metastructure into ambient elastic surfaces and turns everyday objects (such as a mat, sheet or shelf liner) into pressure sensors.

14

UC MERCED MAGAZINE // ucmerced.edu

Members of the PANS Lab: (back row left to right) Ph.D. students Shubham Rohal, Dong Yoon Lee and Shangjie Du and high school intern Joshua Zhang; (front row left to right) undergraduate student intern Victoria Zepeda and Professor Shijia Pan

One research group at UC Merced is using arti cial intelligence to make technology more powerful and unobtrusive. e Pervasive Autonomous Networked Systems Lab at UC Merced, led by computer science and engineering Professor Shijia Pan, focuses on using machine learning to change how society addresses challenges for healthier and more sustainable conditions that make life easier. Her goal, according to her lab site, is to “make the world a smarter place.” Over the past couple of years, the PANS Lab’s research has evolved into incorporating AI into embedded systems around us, a concept the researchers call the Arti cial Intelligence of ings (AIoT). “When we design sensors, we try to make them a seamless part of the objects that people use,” Pan said. Her philosophy follows the vision of Mark Weiser, former chief technology o cer at Xerox, o en called the father of ubiquitous computing, who said, “ e most profound technologies are those that disappear.”

More Efficient Shopping One of the projects Pan and her students are working on borrows from origami — the Japanese art of paper folding. It integrates origami metastructure made from conductive materials into commonly used elastic surfaces, such as mats and liners, to transform these surfaces into sensors.

(Continued on page 16)

A plug-and-play sensor used to remotely detect movements of older adults

15

UC MERCED MAGAZINE // ucmerced.edu

(Continued from page 15) “Using a speci c origami structure pattern, we created a sensor surface that can detect the shape of an object as you press it down on it,” electrical engineering and computer science Ph.D. student Dong Yoon Lee said. is technology could be integrated into autonomous retail stores. Imagine a customer takes a product o a shelf and it is instantly added to their digital cart. ey are charged automatically when they exit the store, streamlining the checkout process. To transition to autonomous retail, Ph.D. student Shubham Rohal said stores would need to change their shelves and add weight sensors or cameras, which aren’t as e ective with smaller products that can be covered by a shopper’s hand. “Creating a sensor as a shelf liner would mean not needing to replace hardware; it would reduce the cost,” he explained.

While holding a small, white, square device in her hand, Pan explained the plug-and-play device captures vibrations induced by elderly patients when they interact with surfaces such as oor tiles and tables. “We develop AI models to interpret information of the patient's activity from these ambient vibrations, which in a certain way turns your home into sensors,” she said. “A caregiver can receive real-time reports on what’s going on in the home without setting up a camera.” Lee is designing an AI model that can e ciently interpret the human activity context behind vibration signals generated through I-CARE sensors. “It helps them be independent and live their lives while helping professionals keep track of their health,” Lee said.

Sensing Health Rohal said the technology could be applied to other areas such as elderly care. “Imagine there’s a rug in your house with a sensor embedded in it that can seamlessly track your movements and gait information.” Although surface matrix sensing produces higher false positive rates that reduce the scalability of such sensors, Rohal said they are incorporating AI to ensure they collect accurate readings. With the help of Pan and her students, researchers at UC Davis Health have developed a visualization system called the Interactive Care (I-CARE) Platform designed to connect older adults with cognitive impairment to their family members who live apart from them.

A low-cost air quality sensor from AirGradient is used to improve air quality mapping resolution and accuracy.

16

UC MERCED MAGAZINE // ucmerced.edu

limited resources. e lab is collaborating on a project with Central California Asthma Collaborative, SocioEnvironmental and Education Network and UC Merced Science and Conservation Field Station to monitor air quality in the San Joaquin Valley. “We have very bad air quality in the Valley. Considering the socioeconomic disadvantages in the region, it’s di cult to have sensors as dense as would be possible in the

Silicon Valley,” Pan said. “Incorporating AI will allow us to have a more accurate geospatial mapping even with very sparse sensor measurements.” Small (resource constrained) devices still pose a challenge, Pan said. “We need to come up with a smarter way to deploy AI in many small devices at scale, and this will unlock more of AI’s potential for social good,” she said. “ at’s why we focus on AI of ings that people use in their lives.”

With powerful AI models, signals from these sensors can be used for tracking medicine intake and other daily activities. However, the embedded systems o en have limited computational power and memory, so Pan and her team are exploring e cient on-board inference. Tracking Air Quality e PANS Lab is also using AI to better characterize the impacts of air pollution, especially from short-term events caused by wild res, with

7KH &DPSDLJQ IRU 7KH 8QLYHUVLW\ RI &DOLIRUQLD 0HUFHG

8& 0HUFHGǽV KLVWRULF HȯRUW WR UDLVH PLOOLRQ E\ ZLOO HQVXUH RXU RQJRLQJ DELOLW\ WR PDNH D GLȯHUHQFH WKURXJK RXU HQGXULQJ SULRULWLHV of Leading through Discovery , Cultivating Healthy Societies DQG Empowering Student Excellence 2XU SXUVXLWV RI UHVHDUFK IRFXVHG RQ UHDO ZRUOG LPSDFW VROXWLRQV DGGUHVVLQJ GLVSDULWLHV WKDW DȯHFW LQGLYLGXDOV DQG FRPPXQLWLHV DQG WKH DFDGHPLF DFKLHYHPHQW DQG VXFFHVV RI RXU VWXGHQWV DUH LPSURYLQJ OLYHV HYHU\ZKHUH 7KH &DPSDLJQ IRU 8& 0HUFHG LV DQ RSSRUWXQLW\ WR EXLOG RQ RXU PRPHQWXP DQG FUHDWH D EULJKWHU IXWXUH IRU ERWK RXU LQVWLWXWLRQ DQG WKH UHJLRQ ZH FDOO KRPH 7KDQN \RX IRU BUILDING THE FUTURE LOCALLY & GLOBALLY

FRQWULEXWLQJ DQG KHOSLQJ XV UHDOL]H RXU VKDUHG SURPLVH Together we are moving our world Boldly Forward.

7R OHDUQ PRUH SOHDVH YLVLW boldlyforward.ucmerced.edu

17

UC MERCED MAGAZINE // ucmerced.edu

Alumni

Here’s what some Bobcat graduates have been up to: • A er seven years with CNN, Oliver Darcy, '11, has embarked on a new venture with the launch of his own independent media publication, Status. • Pin Lyu, Ph.D. ’23, got his dream job as an assistant professor in the Department of Chemistry and Biochemistry at the University of North Carolina, Asheville. His advisor was Professor Son Nguyen. • Duy Nguyen, Ph.D. ’23, is entering the career she had been aiming for as a process engineer at Western Digital in Fremont. Her advisor was Professor Son Nguyen. • Hanbo Hong, Ph.D. ’24, is now a tenure-track assistant professor at Truman State University. His advisor was Professor Liang Shi. • David Moses, Ph.D. ’23, is now a postdoctoral fellow at Scripps. His advisor was Professor Shahar Sukenik. • Feng Yu, Ph.D. ’23, is now a postdoctoral associate at Lawrence Berkeley National Lab. His advisor was Professor Shahar Sukenik.

• Ajay Khanna, Ph.D. ’24, has accepted a postdoctoral position at Los Alamos National Laboratory. His advisor was Professor Christine Isborn.

• Janna Glaze, ’12, and husband, Grant, welcomed their rst child, Summit, in early 2024. Despite his

age, Summit is already making his way around the world with his parents.

• Having reconnected at UC Merced’s orientation, Sharanjit, ’13, and Aarti Gill, ’13, were married in December 2023. A er UC Merced, they continued their education with Sharanjit recently graduating from UC Irvine with an MBA, and Aarti earning an MSW from California State University, Bakers eld. • Hugo Lopez Chavolla, Ph.D.

’24, M.A. ’23, B.A. ’16, and

Destinee Baker, ’17, were engaged in January 2024 and will be married later this fall.

18

UC MERCED MAGAZINE // ucmerced.edu

Oliver Darcy

What Artificial Intelligence Tells Us About Ourselves

COVER STORY

UC Merced student Kahilan Skiba, manager of Professor Colin Holbrook's EvoLab, with one of the lab's robots, Ameca

By Jody Murray One of the remarkable things about arti cial intelligence is how it can provide a deeper understanding of natural intelligence: ours. We talked to faculty in UC Merced’s Department of Cognitive and Information Sciences about the intersection of

AI and “cog sci.” Department faculty teach and study how people process information, combining aspects of psychology, computer science, neuroscience, linguistics and philosophy. Here are some examples of what we are learning about ourselves and about our evolving relationships with AI.

(Continued on page 20)

19

UC MERCED MAGAZINE // ucmerced.edu

(Continued from page 19)

Why Do We Trust AI So Much?

“We must remember these are still devices with

As much chatter in the public square about AI leans toward concern, perhaps even fear, of its expanding presence, research indicates we put far too much trust in it, especially in moments where lives hang in the balance. Experiments conducted by a team led by Professor Colin Holbrook found that subjects in life-or-death warfare scenarios were swayed by information from robots. e subjects were told the robots had limited abilities and were giving advice that could be wrong. In reality, the robots’ advice was completely random. Yet if the subject made a choice — one that, in the scenario, would kill either questioned the decision, the subject changed their mind two-thirds of the time. A human-looking robot with an expressive face and moving arms and torso was marginally more in uential than a ’bot that was simply a box with two cameras mounted on top. But no matter what the AI looked like, and no matter how sure the subject may have been about their rst decision, a machine had a powerful in uence on their resolve. Holbrook stressed that the research’s war-like setup was a means to a broader end. “Our project was about life-or-death decisions made under uncertainty, when the AI is unreliable,” he said, enemy combatants or civilians — if the robot

limited abilities.” — Professor Colin Holbrook

Professor Colin Holbrook

citing situations like emergency medical care or police calls. e project’s ndings raise larger concerns, Holbrook said. Despite the stunning advancements in AI, the “intelligence” part does not include ethical values or an awareness of the world in which we co-exist. We must be careful every time we hand AI another key to our lives, he said. “We must remember these are still devices with limited abilities,” Holbrook said. “ e example I o en give is: Would you take relationship advice from your dishwasher? No. But what if it had an appealing face and spoke to you as a caring friend? Many people might be in uenced by what this dishwasher buddy said.” Sharing (Maybe) the Theory of Mind Nearly a century ago, psychologists started kicking around ideas about social interaction that, in the late 1970s, would crystallize into a concept called the theory of mind. In brief, the theory is the ability to understand the thoughts, feelings and perspectives of others, and to use this understanding to interpret and even predict others' behavior. “It’s an important part of what it means to be human,” Professor Chris Kello said.

20

UC MERCED MAGAZINE // ucmerced.edu

It’s not exclusive to humans, though. Some primates have this ability, and researchers only recently have been able to ask whether AI might also exhibit the trait. “Do large language models have theory of mind?” Kello asked, referring to the neural networks that underlie soaringly popular chatbots like ChatGPT. “Well, they exhibit many of the behaviors taken as evidence for it.” Studies have subjected LLMs to a benchmark test that measures theory of mind. By some measures, the AIs passed many of the tests “with ying colors,” Kello said. e implications of studies showing similarities between humans and AI are leading researchers down numerous avenues of inquiry. Kello and Rachel Ryskin, another UC Merced cognitive science professor, are working on a study that examines the degree to which LLMs can adjust their responses based on an understanding of human memory and attention — one of the many earmarks of theory of mind. Results show as they are trained on massive collections of data, LLMs learn about human limits of memory and attention and can adjust responses accordingly, even though LLMs hardly share those limitations. For Kello, these results refer to a vitally important issue called alignment, which is a desire for AI to align with human intentions, limits and values. It has been known for years that AI algorithms learn human biases through their training data. Extraordinary e orts have been made to root out these biases in chatbots and other AI technology. Some researchers, however, believe this problem may become existential as AI’s capabilities evolve. “We’ve come so far so fast, it doesn’t take a lot to imagine wiring up a robot with some reinforcement learning so they can function in the world but behave in a way we can’t directly control,” Kello said. “ is is where the existential danger comes in.

“Part of my value system is grounded in my experiences over time, of being in a culture. We call that conceptual alignment. We need to be sure AIs are grounded in the same conceptual framework we are.”

Professor Chris Kello

Professor David Noelle

We’ve come so far so fast, it doesn’t take a lot to imagine wiring up a robot with some reinforcement learning so they can function in the world... — Professor Chris Kello

(Continued on page 22)

21

UC MERCED MAGAZINE // ucmerced.edu

(Continued from page 21)

Surprise, Surprise Language — how we acquire it, how we use it, how we understand it — is a big part of cognitive information research. Ryskin and Professor David Noelle are using LLMs to estimate, as a person listens to a sentence, how surprising the next word is. e research uses LLMs, trained on tens of billions of words and phrases harvested from the internet, to estimate the probabilities of which words can follow another. at information is compared to how a subject responds to a sentence like “I’ll take my co ee with non-fat dog.” at last word triggers pulses that are picked up by sensors on the subject’s scalp. “It turns out there are various parts of the brain where activity seems to be sensitive to that level of surprise,” Noelle said. Furthermore, the processing power of LLMs allows researchers to draw parallels between machine and human language processing. “In English, word order matters. ese learning systems are powerful enough that they can pull out these irregularities just from being forced to predict, predict, predict,” Noelle said. What’s Next? How About a Teamup? Looking to the future, imagine a team of LLMs working together to solve a big problem. One is trained in chemistry, another in engineering, a third in biology. Link them up with some human researchers. Give them all a task. Aim high. “So the prompt would be like, ‘We’re trying to cure cancer. Get to work on it. Talk to us,’” said Kello, imagining a machine-human hybrid team tackling some of our biggest issues. “LLMs could provide the thinking but we would provide the intelligence,” Kello said. “We don’t seem far from this happening. And that just blows my mind.”

6WXG\LQJb KRZ $, processes knowledge can provide insights into how the human mind works.

22

UC MERCED MAGAZINE // ucmerced.edu

Impacts of AI on Elections Can Be Bad... and Good

By Patty Guerra

AI also can have a positive impact by helping voters access information on how to cast ballots in their precincts. “AI can help people participate,” Ojeda said. “Participation is crucial to democracy.” Misjudgment or Malicious? AI is only as good as the information put into it. “AI can provide good information to voters, but it can also provide bad information,” Ojeda said. “ at is where it gets a little concerning.” And it doesn’t have to be nefarious, he added. “Maybe it just mischaracterizes what the candidate believes or what their background is and then I come to some conclusion.

When imagining the role of arti cial intelligence in elections, it’s easy to envision an AI-manipulated video of a candidate saying something shocking or incendiary. And that certainly has happened. But AI is also having a more subtle impact on elections. And misinformation isn’t always created by someone intentionally trying to put one over on voters — in fact, you’ve probably contributed to it. “When thinking about AI, the first thing we think about is the bad stuff,” said Christopher Ojeda, a political science professor at UC Merced. “But there are some good ways people can use AI. If you need information about candidates or about referendums or issues on the ballot, there’s no reason why you couldn’t turn to ChatGPT or Google Gemini to provide you with good information.”

(Continued on page 24)

23

UC MERCED MAGAZINE // ucmerced.edu

(Continued from page 23)

less about the candidates, it’s easier for misinformation to have an impact.” e issue has become so concerning that lawmakers are searching for ways to ght misinformation legislatively. Gov. Gavin Newsom in September signed three measures aimed at removing deceptive content from large online platforms, increasing accountability and better informing voters. “Safeguarding the integrity of elections is essential to democracy, and it’s critical that we ensure AI is not deployed to undermine the public’s trust through disinformation — especially in today’s fraught political climate," Newsom said. " ese measures will help to combat the harmful use of deepfakes in political ads and other content, one of several areas in which the state is being proactive to foster transparent and trustworthy AI.” e California Secretary of State has an o ce of elections cybersecurity, which monitors false or misleading information about elections published online or elsewhere. But some lawmakers fear that e orts to weed out fake information could infringe on people’s First Amendment rights. Legislating AI is challenging but not impossible, Ojeda said. “You can see a world in which companies such as X and Facebook take more ownership over the information they carry, that users

don’t have to gure it out for themselves,” he said. “It’s hard work but I think it’s possible.” Reliable Sources Fortunately, there are ways voters can distinguish good information from bad on their own, Ojeda said. One is fact-checking, which involves going to known reputable sources to clarify whether some claim oating around the internet is true. Fact-checking websites such as Snopes, PolitiFact and FactCheck.org can be tools in this e ort. Other processes can be less laborious, Ojeda said. “You could just ask a friend, ‘What do you think?’ e friend may not know but at least you’re getting another opinion,” he said. “ ey can make you feel more skeptical.” Also, voters should keep an open mind. “It’s a challenge for sure,” Ojeda said. “If I support this candidate, I want to believe that they are good and the other candidate is bad.” He recommended getting information from a wide variety of reputable sources. “Information literacy is so important,” he said. “Knowing where to nd credible information. Knowing what apps and websites are more likely to have misinformation. If you have that sort of literacy, you’re already in a good position to begin to gure out or navigate AI as you’re thinking about the elections.”

It wasn’t like there was some bad actor out there.” Concerns about unreliable sources and misinformation have been growing, particularly as social media and other online avenues increasingly replace traditional sources for news. A recent poll from the UC Berkeley Institute of Governmental Studies found that California registered voters get most of their election information from voter guides and news organizations but they use Google and other search engines almost as much as newspapers and magazines. According to the Los Angeles Times, citing a 2023 Pew Research Center survey, four in 10 Americans who get news from social media say they dislike the inaccuracy, up from three in 10 in 2018. A er the 2016 presidential election, about a quarter of Americans said they shared fabricated news stories, knowingly or unknowingly. Legislating a New Frontier ough many people immediately think of the top-of-the-ticket races as rife with misinformation, the bigger risk lies with smaller races, Ojeda said. “We all focus on the presidential election,” he said. “It’s the biggest political o ce in the world.” But most people already know who will get their vote for president. “It’s harder for misinformation to really sway the presidential election because we already know so much about it,” Ojeda said. “For local races, for congressional races, where we know

Information literacy is so important. — Professor Christopher Ojeda

24

UC MERCED MAGAZINE // ucmerced.edu

ONLINE-ONLY CONTENT: California has problems. AI can help solve them.

As AI gets faster, better and easier to use, UC faculty are putting it to use to solve some of the biggest problems facing California, including spotting and responding to wildfires before they spread out of control.

Since the University of California’s earliest days, its researchers have put their expertise and combined years of study to the tasks at hand. But even the nimblest human brain can only crunch so many numbers in a day. That’s where AI comes in. It can sift through massive troves of data, picking out patterns and making predictions that put vital insights within reach. As AI becomes more powerful and easier to use, UC researchers are finding ways to integrate these emerging technologies into their work, shortening the distance to solutions that benefit every Californian. Across science, medicine and engineering, as well as humanities and the law, UC researchers, including many

from UC Merced, are showing how AI can amplify human potential, rather than substitute it, and enhance our creativity and capacity for innovation. From preventing homelessness and spotting wildfires to early identification of Alzheimer's disease, UC writer Julia Busiek describes how as AI gets faster, better and easier to use, faculty are using it to solve some of California’s biggest problems.

Read the story online.

25

UC MERCED MAGAZINE // ucmerced.edu