AIAre AI agents aligned with users or their makers?
More or Less describes a debate over choosing models task by task—and whether ordinary users want to deal with switching agents.
TLDR
More or Less shares a discussion about whose values AI agents follow: the user's or the model company's. One proposed approach is to choose models task by task, treating them like coworkers and switching when one refuses to do a job. The objection is that ordinary users don't even want one agent, let alone the hassle of switching. The counterargument: expecting people to want just one agent is like expecting them to want just one app.