When Jonathan May was completing his undergraduate education at the University of Pennsylvania, he realized he wanted to find “the universal truths in human language.”

Now, as an associate professor of computer science at the Viterbi School of Engineering and the director of the artificial intelligence division at the USC Information Sciences Institute, May is developing technologies with a goal of actually helping humans.

Daily Trojan Magazine editor Maria Laguna sat down with May in mid-April to talk about the ways technologies have rapidly scaled, what motivates him during research projects and the revelations that have changed his mentality.

This interview has been edited for length and clarity.

DAILY TROJAN: I was interested in hearing about this blend of health communication and artificial intelligence technologies, how this process really took shape with your team and what revelations you guys experienced during this blend of studies that you created.

JONATHAN MAY: I really like to seek out these kinds of cross disciplinary things.

We also do work in understanding bias in and the propagation of stereotypes in these language models. … Clearly there are biases baked into AI that then are getting used to make actual decisions that affect people’s lives. …

Xiaoyue Mona Guo, she’s a gynecological oncologist at Keck [School of Medicine] and she works a lot at County General … It’s like a hospital of last resort. It’s a hospital that serves the needs of people who don’t have insurance, who are prisoners, often are homeless. It’s a very overrun hospital, and there’s a lot of burnout of the doctors who work there.

The hope [for the initial project, Chat HPV] was to increase vaccine uptake, … and also reduce burnout of the doctors. … Basically, think of it as an AI-human hybrid, where you’re getting an early education about this vaccine in the waiting room, and then also having follow up conversations with the doctors that are perhaps less difficult conversations. That’s what we’re trying to do here.

One of the revelations we’ve had so far — it’s not a particularly surprising one — is how it takes a long time to get approval to do human subject studies with special populations.

DT: ​​With all the discussions about the research you’re doing, there’s this big emphasis on communication and dialogue, which makes it very fitting for DT dialogues.

But as you were saying earlier, with the development of this communication chatbot research, and also in your role as division director for the USC Information Science Institute, which is a world leader in research and the development of these advanced information processing systems, what keeps you and your team looking for answers in your research? Especially from an outsider perspective, tech is a field that appears to move so rapidly with emerging technologies. What pushes you guys to keep looking for these new developments?

MAY: There’s always really interesting challenges, things that, let’s say, don’t work as well as we want them to. …

My motivations originally come from science fiction. “Star Trek” is science fiction, and it’s still motivational. I was also a fan of “Star Trek: The Next Generation” (1987) and I want to talk with Data — if you’re familiar with “Star Trek: The Next Generation” (1987) at all — [or] talk to androids. I have a superhuman that I can communicate with. And, in a way, we’ve come a long way, and you can have a conversation with Claude, or Chat GPT or Gemini. …

There’s great promise [in] the idea that you could have interesting interactions and have the best of both worlds, that you could have an android that has a superhuman knowledge of all the specifications on the ship or can go into a dangerous situation, and that you can communicate with [the androids] as fluently as you’d like to with a human. We’re not there yet, we’d like to be there. I think there is good that can come of that.

DT: I’m also curious: has there been a moment in your career where something just “clicked” in your brain? Like if you were realizing something in the engineering or artificial intelligence industry that you either wish you knew sooner or you felt relieved to finally understand?

MAY: As a freshman in college and going through an intro to computer science class that was meant as a trial to see if people would survive, they taught us programming languages that were very different from what people used then, and certainly [from what] people use now.

It was very hard to get our minds wrapped around it. And I remember feeling very lost that fall semester and then, it’s one of those late nights in the lab, and you’re working through your problem sets, and there’s a thing a professor described as called the “Gestalt aha,” which is just that moment of revelation when suddenly things fall into place.

DT: It’s your revelation, and I feel like that’s what’s special about it. … Is there anything else you’d like to add?

MAY: It’s important to remember that the reason we’re doing this is because we don’t know the answer. …

Then when you learn something, it’s extremely rare that everything falls into place, and all the dominoes are perfect, and the results are all one directional, right? It’s always a bit of a fog that moves slightly in one direction, and I guess being able to get comfortable with the idea of that fog, that you are incrementally making progress, and that you’re all kind of working together and not knowing anything. It helps. …

I encourage you to not worry about … when things don’t work so well, and not worry if you have too many decisions to make. The revelations will come, and they will often be profound in retrospect.