Somehow we are too distrustful to talk to each other, and more than happy to confess to a sycophantic alien machine.

Chatbot companies want to have it both ways: personalization and emotion as a retention moat, but minimal responsibility for safeguarding the intimate relationships now running on their servers.

LLMs are exceptional improv actors: they’ve ingested countless conversations about consciousness and sci-fi plots, and can convincingly act out a role as if autocompleting a script.

Consider how online radicalization happens: the combination of user agency (proactive search) and algorithmic amplification (recommending related content) leads people to weird places—to micro-cults of internet strangers with their own norms, values, and world-models. No corporate malice is necessary; the ML engineers at YouTube don’t care about... See more

Human skeptics can kill a nascent idea, but ChatGPT was willing to entertain every far-fetched hypothesis.