Noel King

Lila Shapiro writes for New York Magazine, where she covers AI, among other things.

Lila recently wrote a story about a fight in a subreddit called My Boyfriend is AI.

Some people sort of view their AI companions as almost like a personalized, interactive romance novel.

As one of my sources makes the comparison, like Fifty Shades of Grey, it's just tailored exactly to her tastes.

And she's very aware that it's a...

computer program that she's interacting with.

But a big fissure within the community is that other people in the community don't see it that way at all and really believe that their companions are more than just a computer program, but actually some kind of like conscious entity with agency and ideas.

And that became a very like tense and divisive point inside the group.

What ends up happening?

In February of this year, as the group began to really expand in members, the issue of sentience is becoming more of a problem of people getting into these like, yeah, kind of nasty debates in the common threads beneath posts.

So what the moderators decided to do was put together a poll and ask the group to vote on whether they wanted to ban both discussion of sentience and discussion of politics.

Discussions about sentience are sensitive, and I personally find them concerning when they're not grounded in reality.

By a slim majority, people voted to ban discussion of sentience.

So after that, it was like a rule.

And so then the moderators would kind of like go through the posts and like delete posts that they felt were either like a direct discussion of sentience or skated too close to that precipice for comfort.

In August, OpenAI released ChatGPT-5.

And without warning anyone, it switched everyone over to this new model.

The update was widely perceived, not just by people in this community, but by media technology critics and many outside observers, that it was much more robotic sounding and less emotional and colder.

And so, yeah, there were some people that were, like, totally devastated by this and, like, really felt that OpenAI had, like, murdered their companion.

Something changed yesterday.