Ask HN: Has ChatGPT gotten considerably worse?

2 pointsandai2 comments

I'm getting complete nonsense from it today. As if it's been replaced with a much smaller model.

I remember there was a study a while back that measured GPT-4's performance and found that it got measurably worse. Possibly they've done similar "optimizations" on the gpt-3.5-turbo front? (I'm talking about ChatGPT itself, I imagine they'd be less eager to "experiment" with the paid API...)

View on HN

> we've heard all your feedback about GPT4 getting lazier! we haven't updated the model since Nov 11th, and this certainly isn't intentional. model behavior can be unpredictable, and we're looking into fixing it

https://twitter.com/chatgptapp/status/1732979491071549792

Update: ChatGPT-3 just gave me the opposite of the truth four times in a row (so it wasn't a fluke--regenerated the response 4 times). Not an obscure subject either. That's pretty embarrassing.

Reply on news.ycombinator.com