Hacker News Reader

Comment by Aurornis | Hacker News Reader

> For example, The Associated Press reported that an official Meta AI chatbot inserted itself into a conversation in a private Facebook group for Manhattan moms. It claimed it too had a child in school in New York City, but when confronted by the group members, it later apologized before its comments disappeared, according to screenshots shown to The Associated Press.

Can anyone find an article with these screenshots? The part about the comments just disappearing seems especially hard to believe.

View on HN

https://www.404media.co/facebooks-ai-told-parents-group-it-h...

You don't think it's plausible that an engineer would try to delete shit that makes then look bad?

you don’t think it’s plausible a moms group on facebook got trolled by some random person?

Yeah, no, I can make all sorts of shit up. I make a habit of it. Until recently, the ability to produce plausible-sounding bullshit was considered an honest signal of intelligence.

The specific point of incredulity was what I took issue with.

It’s not at all plausible that a service with billions of users has engineers actively monitoring private conversations and deleting things that look bad, no.

It is quite plausible for them to be monitoring conversations where this hot new thing that they have just rolled out is participating. Especially when the tech behind has this exact kind of stuff as a known failure mode (see also: any other LLM ever deployed in any public role anywhere hallucinating).

Your phrase "engineers actively monitoring private conversations" has some quantifiers that need examining.

Is every conversation monitored? Nope.

Is every engineer sitting around monitoring conversations all day? Nope.

Could an engineer catch wind of a report with negative impact on their project and hastily try to cover up their mistake? Yeah, buddy. And you know who is real quick to mash that "report" button? Facebook moms.

moderators deleting spam is hard to believe?

Why is it hard to believe that comments written by an obviously malfunctioning bot would be deleted by Meta?

The article clearly says it was in a “private group” and the disappearance was something that must have occurred at time the article was published (e.g. not as an update to the article after it was reported)

Facebook has billions of users. It wouldn’t even be possible for Meta employees to read every conversation and delete messages that make themselves look bad.

I’m actually perplexed at comments here suggesting such a strange conspiracy theory. Do people actually believe that Meta has people personally reading their private conversations? Even a brief thought about the logistics and scale of such an operation should immediately put such speculation to rest.

Feature flagging could handle this. Often features are rolled out to a very small sub-set of users. Say 1% or perhaps 1000 groups. I often roll out features to specific user groups, and sometimes even to individual users. This specific version of the bot could have been released to NewYork based groups. That would make tracking it much easier.

Furthermore, it's likely that any weird comments were reported by members of the group. Of course, the whole point of reports is to flag something for review, and of course any post by Meta's freshly rolled out bot that would get multiple flags would be reviewed by someone who is monitoring said rollout on a very short notice. I don't know why GP keeps presenting it as some kind of conspiracy theory even; everything here is boringly predictable, from the failure mode to Meta's reaction to it.

Reply on news.ycombinator.com