"Should we slow down AI development?"
by
Wed, Nov 4, 2026
3:30 PM – 5 PM EST (GMT-5)
Private Location (sign in to display)
Details
In recent months there have been a rash of stories about companies losing control of their AI models. OpenAI had an experimental model "escape" and hack the servers of Hugging Face, another AI company. After Anthropic started to look more closely at their test data, they noticed that their own models did similar things at least three times. This triggered calls for shutting down AI research, at least until it is possible to control new models better.
Potential risks posed by AIs are not new. Nobel laureate Geoffrey Hinton commented on potential harms related to AI including its use for surveillance, the potential development of new bio-weapons and autonomous weapons, and the existential threat posed if we create "digital beings" more intelligent than us in his 2024 acceptance speech. Other issues noted include the use of AI to "nudify" people (including children), hack websites, and create deepfakes of public figures to manipulate voters.
So, what should we do? While there is widespread distrust of AI today, the Trump administration has no interest in the government regulating AI (contrary to what the administration suggests, China is regulating AI much more closely than the US). In addition to political opposition from the top, it also may not be possible to stop AI agents from lying, cheating, and coordinating with other agents.
To help us understand these issues and more, and whether it is necessary to stop AI progress in the name of safety, at least for now, we welcome Dr. Lauren Eastwood, Professor of Sociology and Chair of Art and Design, and Dr. Michael Walters, Professor of Physics and Co-Chair of both Physics and Computer Science, to the Institute.