Bill Swartout has spent decades building systems where people talk to virtual humans. In 2010, he and his team created Ada and Grace, virtual museum guides named after computing pioneers Ada Lovelace and Grace Hopper, for the Boston Museum of Science. The AI-powered twins could answer questions about computers, technology, and what visitors could see in the museum.

Families would come in, parents fiddling with their smartphones while their kids interacted with the virtual guides. But after a few exchanges, something would shift. “The phones would go in their pockets,” Swartout recalls, “and they’d get interested, too.” Over 250,000 visitors met Ada and Grace during their time at the museum.

With the New Dimensions in Testimony project, which preserved Holocaust survivors as interactive holograms, the reactions ran even deeper. It’s the only project Swartout has worked on in 50 years that regularly moved people to tears.

After half a century in artificial intelligence, a field he fell in love with as a fourth grader reading Danny Dunn and the Homework Machine, Swartout* retired on February 1 from his role as Chief Science Officer at USC’s Institute for Creative Technologies.

“It was a much longer journey than I thought it would be,” he says. “When I read that Danny Dunn book in the early 1960s, I thought we’d have this figured out by the time I got to college.”

Swartout has written extensively about his 50-year career in AI, including a detailed essay on ICT’s website. This story focuses on the man behind the technical milestones, how he navigated choices, what moved him, and what he’s taking with him into retirement.

The Road Not Taken

Swartout didn’t set out to become one of the world’s leading experts in virtual humans and explainable AI. Growing up in St. Louis, Missouri, he was torn between two passions as a teenager.

During his high school years, computers were rare and access meant improvising. His boys’ high school didn’t have one, but the girls’ high school next door did. “Some of us guys managed to get some access to that computer,” he recalls. As for whether he also got dates out of the arrangement? “Unfortunately,” Swartout says with a laugh, “it was mostly the computer.”

“I was really interested in computers, but I was also passionate about photography and, in particular, filmmaking,” he says. One summer, he volunteered as an assistant cameraman at the local PBS station. “That was a really fun experience, and I could sort of see myself doing that.”

Ultimately, two things settled it. Working at the station, he recognized he was around genuinely talented filmmakers. “In computers, I felt like I had a real knack for it,” he says. His father’s advice didn’t hurt either. Practical Midwestern wisdom: “Bill, if you go into computers, you’ll probably always have a job.” Swartout laughs. “Not an incorrect insight, I’d say.”

Building AI at USC

After getting his bachelor’s degree from Stanford in 1974, Swartout earned his master’s and PhD in computer science from MIT, graduating in 1981. He joined USC’s Information Sciences Institute, one of two research homes that would define his career, along with ICT, both part of the USC Viterbi School of Engineering. There he worked with Bill Mark and Bob Balzer on automatic programming and explainable AI systems.

In 1989, he became Director of the Intelligent Systems Division at ISI. The AI winter that hit at the end of that decade, when expert systems failed to live up to their hype and the market collapsed, might have derailed lesser operations. But Swartout’s group had strong DARPA funding and a particular advantage: a generation of talented PhD students had entered AI programs in the early 1980s, expecting a boom. When they graduated into a bust, ISI was ready to take them on. “We were able to hire some of these people,” Swartout says. Among them was Craig Knoblock, who would become ISI’s executive director decades later, and an AAAI Fellow. By the time he left the institute, five researchers from the division had been elected AAAI Fellows, a remarkable concentration of talent.

By the late 1990s, after a decade of building the division, Swartout was ready for something new. Then came an unexpected opportunity.

When USC was selected to create the Institute for Creative Technologies—an unprecedented collaboration between Hollywood, the military, and academia—Swartout jumped at the chance. The kid who’d once agonized between computers and filmmaking was about to work at their intersection. “It was like getting to do both,” he says.

Working alongside Hollywood professionals turned out to be its own education. Swartout recalls a conversation with a former entertainment executive who had moved to a tech startup and found himself managing engineers. Swartout had expected him to say the experience was easier, less emotional, more rational. Instead, the man said engineers were the hardest people he’d ever managed. It took Swartout some time to figure out why. In engineering, agreed-upon models, physics, algorithms, analysis, ultimately make the decisions. “If someone shows you an algorithm that’s three times faster and you overrule it anyway, you immediately lose credibility,” he says. In Hollywood, no such models exist. Nobody can prove a script will succeed. Decisiveness is the currency, and being wrong later is forgivable. Hesitating is not. “It’s almost a 180-degree mind shift.”

The Sunday Everything Went Wrong

As the ICT was getting underway, the grand opening was set for Tuesday, September 26, 2000. The centerpiece was the Mission Rehearsal Exercise, an immersive AI-driven simulation that would prove ICT’s worth to the Army, Hollywood, and the university. After months of hard work, on the Saturday just before the opening, they’d finally gotten it to run all the way through without crashing. “I slept great that night,” Swartout remembers. “Best sleep I’d had in weeks.”

Then Sunday arrived. A power glitch had knocked out systems overnight. The elevators were down. The main integrating programmer called from the hospital on crutches. And the fiber-optic controller connecting the computers upstairs to the displays below needed a password nobody had—the installers had flown back to the Midwest. “Sunday afternoon in the Midwest,” Swartout says. “We’re calling around trying to track these guys down.”

They got the password. Swartout spent the entire opening in the theater running demonstrations. The system held. There was only one crash across dozens of demos, and ICT’s reputation was safe. When it was finally over, he walked out to the remnants of a party scattered everywhere. “I was thinking, wow, this must have been a pretty good party. I wish I’d been there.”

The Project That Made People Cry

Of all Swartout’s projects—the expert systems that could explain their reasoning, the virtual humans that trained soldiers, the educational tools that helped sailors retain knowledge—New Dimensions in Testimony marks him differently.

The collaboration with USC’s Shoah Foundation, the ICT Natural Language Group, headed by David Traum, and the ICT Graphics Lab, headed by Paul Debevec, created interactive holographic-like displays of Holocaust survivors. Using thousands of pre-recorded responses to anticipated questions, the system could engage in natural conversations about their experiences.

“We recorded Pinchas Gutter,” Swartout recalls. “Remarkable man. Remarkable story. We spent days with him, asking questions, recording his responses.”

What stayed with Swartout wasn’t the technical feat. It was Pinchas’s answer when asked why he tells his story. “He talks about how he tells his story so that people will learn from the Holocaust, so that what he and so many others suffered will never be forgotten, and never repeated.” And when visitors ask Pinchas whether he hates Germans, his answer is no. “There’s just an amazing amount of wisdom that comes out of that,” Swartout says.

“People would cry,” Swartout says quietly. “Regularly. It’s the only project I’ve worked on that consistently moved people to tears. And the thing is, it wasn’t about AI. After a few interactions, people forgot about the AI, and the stories took over. People were having conversations with Pinchas. With a person.” The technology worked so well it disappeared.

Optimist With a Long Memory

Swartout has watched AI cycle through enthusiasm and disappointment more than once. He started working on explainable AI, systems that could justify their reasoning, in the late 1970s. The field lost interest. Now, 45 years later, as large language models make decisions we don’t understand, explainability is urgent again.

But he’s quick to note it’s a different problem now. Back then, his systems were built with explicit rules and concepts — if a patient’s temperature exceeds 99 degrees, conclude fever. You could read the code and understand exactly why it did what it did. Today’s AI works differently: neural networks learn by processing vast amounts of data and adjusting billions of numerical relationships, none of which correspond to anything a human would recognize as a rule or a reason. The system arrives at an answer, but there’s no trail of logic to follow back.

Sixty years after reading about Danny Dunn’s homework machine, the technology is real. But Swartout isn’t particularly worried about doomsday scenarios. “If something goes seriously wrong with AI, it won’t be from AI becoming these super-smart overlords,” he says. “Without regulation, which is essentially where we’re at right now, we can do a lot of damage to ourselves.”

What gives him hope is education, the work he’s continuing in retirement through ICT’s AI Research Center of Excellence for Education. He’s particularly interested in tools that help students learn without doing the work for them. Systems that can facilitate brainstorming, sharpen critical thinking, give feedback on writing, and use AI to enhance students’ ability to think rather than delegate thinking to the machine. The key is tracking the student’s process at a fine-grained level, so teachers can evaluate the work they did, not just the artifact they produced.

“That’s the Danny Dunn connection again,” he says. “The homework machine was supposed to help kids learn.”

The Post-Production Years

Swartout’s retirement is now effective, but he’s not disappearing. He’ll continue as Research Professor Emeritus in the Thomas Lord Department of Computer Science, serving on USC President Beong-Soo Kim’s AI Strategy Committee and working on AI and education projects as funding comes through. He’ll probably trade his current office for something smaller, or use hoteling space on campus.

The career came with trade-offs he still thinks about. When ICT was launching, the most demanding stretch of his professional life, his kids were teenagers. “I think I missed out on being more involved at the time,” he says quietly. “I kind of regret that.”

He’s been getting more serious about his life-long interest in photography, ranging from landscapes to street scenes. It’s a return to that early love from his teenage years, the road not quite taken. For street photography, he often uses his smartphone. “Nobody thinks twice about somebody taking a picture with a smartphone,” he says. “That opens up a lot of possibilities.” Though he notes, with a laugh, that AI is now doing heavy processing on every photo a smartphone takes. “Maybe I can’t escape it after all.” He’s also got a high-end Nikon for when he wants more control.

He has two granddaughters, one four years old, one just a few months old. He’s thinking about when the time comes to talk to them about using AI tools thoughtfully, without letting the tools do their thinking for them. “My son does a really good job limiting screen time,” he notes.

For now, though, retirement means time. Time for photography. Time to travel, to Europe, maybe the Middle East someday. His wife speaks French, he speaks some German. “A lot of places to go,” he says.

The conversations will continue. The Holocaust survivors’ testimonies will survive, teaching future generations long after everyone involved in creating them is gone. “Without that technology, without that work, we wouldn’t be able to recreate it,” Swartout says. “That’s what I’m most proud of. Not the technical achievement.”

He pauses. “That’s how you know it’s time. When the work can go on without you, and you’re happy about that instead of threatened by it.”

The homework machine his fourth-grade self dreamed about is real now. It took sixty years longer than he expected. But Bill Swartout helped build it, and the conversations he enabled will continue long after he’s stepped away from his office at ICT: between students and teachers, between visitors and Holocaust survivors, between past and future.

*William (Bill) Swartout was the director of technology at the USC Institute for Creative Technologies and a research professor in the Thomas Lord Department of Computer Science at the USC Viterbi School of Engineering and the USC School of Advanced Computing.

Published on March 27th, 2026

Last updated on March 27th, 2026