Bonus issue:
This one is a little bit less obvious
Why do LLMs obsess over making numbered lists? They seem to do that constantly.
Oh, I can help! 🎉
- computers like lists, they organize things.
- itemized things are better when linked! 🔗
- I hate myself a little for writing this out 😐
My conspricy theory is that early LLMs have a hard time figuring out the logical relation between sentenses, hence do not generate good transitions between sentences.
I think bullet point might be manually tuned up by the developers, but not inheritly present in the model; because we don’t tend to see bullet points that much in normal human communications.
That’s not a bad theory especially since newer models don’t do it as often
- Honestly I don’t know
Well they are computers…
Late but I’m pretty sure it’s a byproduct of the RHLF process used to train these types of models. Basically, they have a bunch of humans look at multiple outputs from the LLM and rate the best ones, and it turns out people find lists easier to understand than other styles (alternatively, the poor souls slaving away in the AI mines rating responses all day find it faster to understand a list than a paragraph through the blurry lens of mental fatigue)
Lol, my brain is like, nope, I’m not even trying to read that.
I think I lost a few brain cells reading it all the way through.
Wow, this just hurts. The “twice, I might add!” is sooooo fucking bad. I don’t have any words for this.
aby
Checks out
god damn it i can’t type lmao
When your repository is on Facebook.
I wonder if they made chat gpt use an unnatural amount of emojis just to make it easier to spot
People often use a ridiculous amount of emoji’s in their readme, perhaps seeing it was a README triggered something in the LLM to talk like a readme?
The emoji littering in fastapi’s documentation actually drove me away from using it.
I mean, even if it’s annoying someone obviously used AI, they probably still have that problem and just suck at communicating that themselves
They don’t, because it’s not an actual issue for any human reading it. The README contains the data and the repo is just for coordination, but the LLM doesn’t understand that.
Then… that’s so fucking weird, why would someone make that issue? I genuinely lack the understanding for how this could have happened in that case.
I’m pretty sure it’s an automated system that makes these issues. The accounts looked like bots. However, that only makes it even weirder.
There have been so many people filing AI generated security vulnerabilities
