|
The offices of USC’s new Stevens School of Computing and Artificial Intelligence still kind of looked like a building site when Marketplace’s Meghan McCarty Carino visited in June — there was scaffolding and lots of bare rooms.
“Some people haven't moved in yet,” said the school’s director Gaurav Sukhatme as he showed us around a lab for physical AI, filled with robotic arms.
“Those two robot arms are using modern sort of techniques to learn how to collaborate with each other to do bimanual tasks,” he said.
The new school recombines some existing programs within engineering and adds new tracks: a dedicated AI major and minors for non-STEM students to apply the tech in disciplines like art or history.
“We have a very project-driven curriculum, which allows us to refresh the projects that the students do as part of their degree programs, and we keep changing those projects and updating them,” said Sukhatme.
But moving at the speed of AI isn’t really what universities are built for. And the state of the art is expensive. Training one of today’s advanced models is estimated to cost hundreds of millions of dollars.
“One of the things a university is really good at is posing questions not about what's happening today, but about what might happen a decade from now,” he said. “It is having the freedom to think about entirely new ways of doing things.”
It also means having the freedom to keep humans at the center of that inquiry, according to Shri Narayanan. He runs USC’s Signal Analysis and Interpretation Lab, which uses AI to analyze patterns in human “signals” to tackle big clinical questions. Researchers at SAIL study beat boxers to understand how vocalization works to help preserve it in cancer patients, and identify early signs of autism or depression when treatment can make the biggest difference.
“For me, humans are at the center of everything, right? For humans, by humans,” said Narayanan.
But AI that can infer details about our health, mind or identity carries real risks, as well as real benefits.
“Can we do it in the right way? And this is not something just computer scientists can do,” he said, noting these questions demand expertise from clinicians, social scientists, even philosophers.
It’s this kind of cross pollination USC is aiming to foster with its focus on applying AI across disciplines. |