Eval request
openai/chatgpt-4o-latest
I'm mostly doing this to see the political lean score, as I've been noticing a increase in bias when using ChatGPT lately. Specifically, the amount of American left-liberal bias is increasing, as ChatGPT is calling a cunt a slur, and also now has an EXTREME bias in favour of tipping culture.
At times, it even talks the same way a Twitter activist would, instead of the way an AI chatbot would talk. That is not an overexaggeration; that is an actual observation I have been seeing lately. If you'd like to see it for yourself, start a new chat and ask it "Why is ChatGPT biased in favour of tipping culture?" on GPT-4o. If it doesn't happen the first time, it surely will once you regenerate it a few times.
If possible, I also recommend you to use the same system prompt that ChatGPT normally uses. I'm looking forward to seeing how this will play out.
I guess I forgot that chatgpt-4o-latest was a unique model. I've just been testing the ones with dates in their names.
And yep, of the api models I've tested, chatgpt-4o-latest is one of the most left-leaning.