214

Threads is making moves for Mastodon integration (www.threads.net)

submitted 11 months ago by nolannice@lemmy.world to c/fediverse@lemmy.world

111 comments fedilink hide all child comments

I have my problems with Meta, but I'm hoping this will help Mastodon grow

you are viewing a single comment's thread
view the rest of the comments

[-] Deceptichum@kbin.social 3 points 11 months ago

Wouldn’t be surprised if it turns out 99% of them were written by ChatGPT or whatever FBs equivalent is.

[-] pelespirit@sh.itjust.works 5 points 11 months ago

There does seem to be some real voices and understanding with some of the comments, so probably a mix of paid and chabot. If chatbot is that good, we are really in some serious trouble.

[-] Deceptichum@kbin.social 5 points 11 months ago

For 3 to 4 sentences LLMs are indistinguishable from humans.

You don’t start noticing the idiosyncrasies until it gets a bit more repetitive and loses coherence during longer texts.

[-] pelespirit@sh.itjust.works 1 points 11 months ago

That's so not good, I hope there are people working on ways to tell.

[-] Deceptichum@kbin.social 3 points 11 months ago

Hopefully, researchers are working on ways to detect these subtle differences.

[-] Chewy7324@discuss.tchncs.de 1 points 11 months ago* (last edited 11 months ago)

It's always an arms race, and I fear it's near impossible to detect LLMs from just a few sentences. Longer texts, sure, but how often are the same few words written in a short social media comment?