Ken Goldberg, Stuart Russell, and Trevor Darrell in conversation at TechCrunch Robotics 2020 at UC Berkeley
Robotics and AI experts from industry and academia presented research and tech innovations at the CITRIS co-sponsored event TechCrunch Sessions: Robotics + AI 2020 on March 3 at UC Berkeley’s Zellerbach Hall.
CITRIS People and Robots Director and EECS faculty Ken Goldberg moderated the afternoon panel on “The Next Century of Robo-Exoticism,” offering insight into how contemporary attitudes towards robotics intersect with a historical context of Orientalism, the concept of alienating and mystifying the Other first put forward by the late Palestinian academic Edward Said.
“We want to think about how these things are affected by this bigger historical and political context of robots,” Goldberg said, addressing growing levels of global xenophobia and populism.
Panelist Abigail De Kosnik, Director of the Berkeley Center for New Media, talked with Goldberg about how Chinese railroad workers in the 19th century were Orientalized as robotic machines impervious to pain, able to work in the mud and heat. Kosnik compared this stereotype with that of scheming “Asian masterminds.”
“Robots can either be this hyper-intelligent servant class that we don’t want or this servant class that has to be our slaves,” said De Kosnik.
Goldberg argues that both these stereotypes are not representative of reality. “In politics, Andrew Yang was saying that automation will wipe out all of our jobs, and I hear that trope so often. Most of us working in robotics are like, ‘What are you talking about?’ That is not imminent at all,” he said.
AI pioneer, EECS faculty, and CITRIS researcher Stuart Russell spoke on “Saving Humanity from AI” — asserting that AI must be fundamentally reformed in its construction, with humanity itself at stake. The Guardian has described his 2019 book Human Compatible as, “Surely the most important book on AI” of the year.
“The problem is, we don’t know how to specify [AI’s] objectives completely and correctly,” Russell said, addressing the dangers of black-box AI delegation. Contemporary AI research continues on from Enlightenment rationality that prioritizes objective-based decision-making.
Russell argues that algorithm development must be beholden to humans rather than just coded instructions. “The objectives [of AI] are really in us —it’s what we want the future to be like… Once AI is capable of running our civilization, we no longer have the need to teach our next generation about everything we know about our civilization.”
At the panel on “Opening the Black Box with Explainable AI,” CITRIS affiliate faculty and EECS faculty Trevor Darrell spoke on the practical need for open source software and how the choice between effectiveness and explainability is not reflective of the real world.
“When you train something as a black box… and it just doesn’t work, it’s very unsatisfying,” said Darrell. “At Berkeley, when [our AI systems] learn, we train them to essentially show their work, and open up their black box… enough so that humans can get in and have a more effective process of training them, using them, and trusting them.”
Find videos of all presentations at the TechCrunch Sessions: Robotics+AI 2020 website.
The Center for Information Technology Research in the Interest of Society (CITRIS) and the Banatao Institute drive interdisciplinary innovation for social good with faculty researchers and students from four University of California campuses – Berkeley, Davis, Merced, and Santa Cruz – along with public and private partners.
To learn more about CITRIS, sign up for our newsletter: bit.ly/SubscribeCITRIS