7 comments

[ 3.4 ms ] story [ 14.1 ms ] thread
(comment deleted)
This stuff worries me. I have already seen colleagues using LLMs and not checking the results. Really it just makes stupid people more dangerous. And there is a weird absolutist mechanical faith among certain people to the point even criticising the things is blasphemous and yes everyone is going to be replaced by LLMs and that's a good thing.

Real problem: crap is being normalised because no one wants to call the vendors out on it, because it'll devalue their business.

I figure I'm just going to sit this out at this point and watch everything burn.

What's a reliable benchmark for measuring "AI Distortion" in these models, allowing for consistent tracking over time and potential improvement?
The left wing biased bbc are guilty of manipulating the news.
I know it's easy to contuse these two, but this is not X.