In the early days of personal computing, people imagined artificial intelligence as a single, self-contained system — a kind of digital brain with a fixed identity. Talking to AI meant asking questions to a clearly defined machine.
That picture no longer holds, says David Chalmers, a professor of philosophy and neural science at New York University. Today’s large language models, or LLMs, run on distributed cloud infrastructure, so they don’t behave like a single fixed machine with one clear physical boundary.
While LLMs likely lack consciousness, he says, they are not merely passive tools. Instead, they’re “quasi-agents” that exhibit goal-directed behavior shaped by context. Each back-and-forth between the user and the agent holds together as a coherent exchange, but it only exists for the duration of that interaction. If this thread is the real unit of interaction, he says, then agency isn’t something the system permanently processes. Rather, it emerges during use.
In this Berkeley Talks episode, Chalmers asks: When we’re interacting with AI, what exactly are we interacting with? If these short-lived “selves” exhibit beliefs and goals within a conversation, do they warrant any ethical consideration?
Chalmers’ talk took place at UC Berkeley’s International House on May 7. It was the inaugural Sarah Douglas Lecture on Philosophy and Artificial Intelligence, sponsored by Berkeley’s Department of Philosophy.
Listen to the episode and read the transcript on UC Berkeley News (news.berkeley.edu/podcasts/berkeley-talks).
New York University photo.
","author_name":"UC Berkeley"}