By Jackie Brown. On September 13, 2023, renowned computer scientist and author Jaron Lanier, addressed tech enthusiasts and UC Berkeley faculty, staff and students. Co-hosted by CITRIS and the Banatao Institute, UC Berkeley College of Computing, Data Science, and Society and the UC Berkeley Artificial Intelligence Research (BAIR) Lab, his talk offered a personal perspective, historical overview, and suggestions to foster “data dignity” within the realm of artificial intelligence.
Lanier has significantly shaped today’s tech landscape through his participation in the invention of virtual reality (VR). Lanier worked at Atari before co-founding VPL Research, which sold the first VR goggles and wired gloves. In 2006, he joined Microsoft and currently works in Microsoft Research. His list of awards is extensive, including an IEEE lifetime career award, the Peace Prize of the German Book Trade, and named one of the world’s 100 most influential people by TIME magazine.
Jennifer Chayes, Dean of UC Berkeley’s College of Computing, Data Science, and Society, introduced Lanier, a coworker when at Microsoft and friend for 20 years. Lanier was a compelling and humorous speaker, and reinforced his unconventional manner by speaking without slides, seated in a chair on stage rather than from a podium. In another unconventional move, Lanier explained how his self-titled role as “prime unifying scientist” under the Office of the Chief Technology Officer creates the acronym OCTOPUS.
The father of Virtual Reality (VR)
Lanier’s groundbreaking invention of VR hardware gave him the nickname “the Father of VR.” Lanier notes disparities in reliability, usability, and consistency between his programs and those of his competitors. Currently, scalability and quality are intertwined with success, but Lanier is unsure if this trajectory will continue.
Lanier holds a unique position in the tech industry: as an innovator and a critic. Lanier published Ten Arguments for Deleting Your Social Media Accounts and appeared in “The Social Dilemma,” a movie which demonstrates his concerns regarding the business models of some tech companies and the potential for societal addiction.
History of Artificial Intelligence (AI)
Lanier encourages individuals to know the history of foundational computer scientists, specifically Marvin Minsky and Norbert Wiener. In the 1950s, Minsky and a group of researchers convened at the Dartmouth Conference and coined the term “AI” in response to Wiener’s work.
Along with pushing the boundaries of tech innovation, Lanier pushed boundaries with Minsky, who became his mentor. Although Lanier and Minsky enjoyed a professional and personal relationship, they shared challenging discourse and differing opinions on the tech industry. Lanier reminisces how their relationship was mutually beneficial and continued until Minsky’s death.
Social collaboration lens
With thought-provoking insights, Lanier discusses the dichotomy of AI perception. Many view AI as an alien-like figure and an independent entity, creating a dynamic that alienates those who may feel excluded and unwanted in the evolving “tech universe.” However, Lanier offers a proposal. If individuals view AI as a “social collaboration resource” similar to Wikipedia, the tech narrative could be more inclusive. This strategy could also increase the amount of AI training data which would better represent the human experience and offer more accurate results. Lanier cites websites on GitHub as examples of effective social collaboration where contributors can adapt and change coding languages.
Open AI’s success and improvements
Although other AI models exist, ChatGPT gained traction due to its simple user interface. Individuals can utilize this tool for daily questions without needing deep technical knowledge. It can be used for a range of tasks, including writing wedding vows, as Lanier recounted.
Lanier candidly addresses the challenges faced when AI models, such as ChatGPT, generate responses that are dramatic, romantic, or even shocking. Many of these responses were produced from soap opera and fan fiction transcripts. Lanier emphasizes the importance of understanding AI algorithms. Although models can remember chat cycles, their memory can be unreliable. Lanier’s team has implemented limitations on the number of interaction cycles with the bot to enhance response quality. However, Lanier explains how these constraints introduce potential issues akin to the “Genie problem” where restrictions on AI behavior are not foolproof.
Lanier addressed how AI’s future is not a black-and-white scenario. He acknowledged how new technology can create and stifle jobs, yet decades must pass before society can have a comprehensive understanding of technology’s effects. Lanier suggests a novel solution to address potential job loss: create a model to “celebrate contributors” who could earn income by enriching open AI models through data input. This diversification of inputs could further promote data dignity.
Do you think development of AI should stop, why or why not?
I don’t think AI is a real thing. My buddy wrote a statement and lots of people at Microsoft signed it. However, it makes AI seem like an entity when it should be focused on with a social collaboration lens. Bluntly, it is hard to argue with success. AI is a commercial enterprise that is making a lot of money. I have faith in fellow human beings that my belief will gain traction.
Can ChatGPT be a social collaboration? How can we reckon with training data being primarily white and male?
This is a problem. However, slapping another model to try to combat biases runs into the genie problem with rules and regulations. A USC model using visual data and sensors during the 1990s did not recognize African Americans because it was trained on the USC engineering team which was all white at the time. Our antecedent must reflect human data. Why can’t people be motivated to make training data better?
Watch the recorded livestream from Lanier’s talk here. Register for upcoming CITRIS Research Exchange events here, both in person and online. The Wednesday speaker series will continue October 4, and October 25.
UC IT Blog will be hosting an online BrownBag lunch to discuss Lanier’s talk along with the other CITRIS Research Exchange speakers. This event will be hosted on October 25th, 2023 from 2-2:45pm. Register here for this Zoom event.
[Cover photo caption: Jaron Lanier presents at UC Berkeley.]
Wednesday, October 25, 2-2:45 pm
Host: UC Tech news blog team
Guest: Camille Crittenden
Location: Zoom [use this Zoom registration link]
Marketing & Communications Intern
UC Office of the President