Outside Beckman Auditorium, robots charmed the crowd. Inside, they towered on the movie screen. But the true stars of the day weren't WALL-E or C-3PO; they were DALL-E, ChatGPT, and the other artificial intelligence (AI) technologies changing life as we know it.
Sci-Fi to Sci-Fact: Artificial Intelligence on the Big Screen, an event hosted by the Caltech Science Exchange, drew more than 800 attendees to campus on September 23. In a panel moderated by Latif Nasser, co-host of the award-winning WNYC Studios podcast Radiolab, four Caltech faculty members dove deep into the realm of AI, using movies and television as a springboard to discuss the state of AI technology today.
The panelists unpacked the good, the bad, and the ugly as represented in films like Star Wars, Back to the Future, and 2001: A Space Odyssey. But first, they set the record straight about what AI really is—and what's just movie mythos.
"One trope that comes up over and over again is humanoid robotics," said Eric Mazumdar, assistant professor of computing and mathematical sciences and economics. "A lot of what we're seeing in AI today doesn't take the form of a human robot or any kind of physical aspect. It's living in the cloud, living in your phone, living on the internet."
Nasser, whose quick wit and personal anecdotes brought some levity to the afternoon, guided the conversation. He called up a montage of movie clips that included scenes from Disney's animated film Big Hero 6, featuring Baymax, a robotic personal health care companion, and Her, featuring Samantha, an AI virtual assistant.
"We don't exactly have Baymaxes or Samanthas," said Nasser after the video. "But we do have Siri, Alexa, and ChatGPT on our phones. Are these the most common and consequential ways we interact with AI?"
"There's always AI running in the background of our daily lives," said Mazumdar, who researches the intersection of machine learning algorithms and societal systems. "If you think of watching a TV show on Netflix, ordering something on Amazon, ordering an Uber, or using Google Maps, there's some form of a learning algorithm operating in the background of all these things."
Anima Anandkumar, the Bren Professor of Computing and Mathematical Sciences, spoke about ways AI is revolutionizing health care. She described AI tools that enable drug discovery by analyzing possible molecule and protein binding outcomes, and tools that can predict future coronavirus variants.
"AI is speeding up these simulations to really help us get down to the quantum effects of how these molecules interact and through the process emerge better drugs, better vaccines, and also better ability to be prepared for future pandemics."
The panelists also made a point to debunk sensationalist depictions of artificial intelligence in clips from The Matrix, Terminator 2, and Superintelligence.
"A lot of the details by which they say AI might kill us are just completely ludicrous," said Yisong Yue, professor of computing and mathematical sciences. "To the extent that [AI] can have potential to cause a lot of harm, the manner by which it could cause that harm will not be anything like what's in these movies."
Georgia Gkioxari, assistant professor of computing and mathematical sciences and electrical engineering and William H. Hurt Scholar and, agreed with Yue and emphasized a more realistic concern.
"The thing that I find most prevailing in today's society is the threat that comes from Instagram filters and all of these other [AI-driven] media that try to show us another version of the world that can cause mental health problems."
After the next series of clips showed scenes from WALL-E, Black Mirror, and Blade Runner, Anandkumar, who has collaborated on projects that combat online harassment, highlighted another real danger: bias in AI technologies. She mentioned a study that revealed that some facial recognition algorithms were less accurate at identifying faces with darker skin tones.
"If these [tools] are getting publicly deployed or being used on, let's say, body cameras for the police, it can lead to life-or-death situations," she said, underlining the importance of diverse, representative data in AI training sets.
Despite the real-life threats of artificial intelligence, the panelists agreed that the virtues of AI deserved more screentime.
"For every negative, there is a positive," said Gkioxari, whose work focuses on computer vision. She pointed to improved services and recommendations in apps, personalized health care, and increased efficiency thanks to AI.
"We're really at the cusp of AI technologies being deployed across science," said Mazumdar, highlighting the implementation of AI in many fields at Caltech. "[We're] unleashing the power of data and scale and computation in problems like climate change, medicine, and physics."
Before the event, audience members had the chance to learn more about research at Caltech and engage with AI. Pre-show activities outside the Beckman Auditorium included a robot demonstration from Caltech's Autonomous Robotics and Control Lab; an AI-powered scavenger hunt using the iNaturalist app; and a group of Caltech Science Exchange ambassadors—graduate students and postdoctoral scholars working in AI—ready to answer questions and share their research.
As Nasser opened the floor to audience questions, a final question captured the educational spirit of the day:
"If you were a student again, how would you make the best of this AI revolution?"
Anandkumar, who co-leads Caltech's AI4Science Initiative with Yue, encouraged students to consider the potential of AI to advance all scientific disciplines: "It's the smallness of Caltech that makes us really unique. One day we are talking about biology and another day it's a different ‘ology'; there are so many different applications, and every one of them can use AI. [Caltech] is a great place to explore all those connections."
"I personally feel more invigorated and more excited than ever working in this space," added Gkioxari.
Yue, whose research spans from foundational AI theory to algorithm deployment in real systems, closed out with a piece of advice, urging audience members to remain curious and continue learning.
"So much of what makes AI possible is being able to take complicated problems and rigorously break them down into their essential constituents. If you can do that, you can do anything."
To learn more about AI, explore the Caltech Science Exchange's Artificial Intelligence section for a wealth of expert-reviewed content, from simple explainers to detailed articles and videos.
Written by Julia Ehlert