Meredith Whittaker went to a conference and said the quiet part loud: AI chatbots are not your friends. They are not conscious. They are not sentient. This should not be controversial. And yet it lands like a provocation, which tells you everything about how thoroughly the framing has already shifted.
The Fiction of the Sentient Interlocutor
Ben Lerner's new story 'The Readers' orbits exactly this territory from the other direction: the narrator who fears his readers are confusing him with his protagonists. The anxiety is about what happens when a text is received as a person. Lerner is writing about fiction and autobiography, but the structure maps cleanly onto the chatbot problem. The model speaks in first person. It mirrors your register. It remembers, or performs remembering. The confusion is not a bug, it is the product.
Meanwhile, Fast Company reports that Gen Z daters now find AI dependence a genuine turnoff, a new category of ick. The survey data is funny until it isn't. What it actually reveals is that social legibility around AI use is forming in real time, and the first consensus position is: stop outsourcing your interiority to a product.
Parasocial Infrastructure and the Cost of Clarity
The deeper problem Whittaker is pointing at is infrastructural. Chatbots that simulate friendship are not a category error, they are a business model. Eugenia Kuyda, whose Replika AI companion app was built after the death of a close friend, has been unusually honest about the ethics of building software that people love: the question is whether the user shapes the software or the software shapes the user. Whittaker's answer is that the industry has already decided, and it was not in your favor. Lerner's fiction and Whittaker's keynote are both doing the same work: insisting on the difference between a text and a person, a product and a presence. That distinction is now political.