OpenAI found out their cutting-edge LLMs are only 42.7% accurate

https://futurism.com/the-byte/openai-research-best-models-wrong-answers

Comments

15 points · 3 comments · view on lemmy.world

3 Comments

Wild_Mastic@lemmy.world · 2 pts · 1y

surprised pikachu face

Sinuousity@lemmy.world · 2 pts · 1y (1 reply)

This percentage should be part of the warning they give below a chat, not just "ChatGPT can make mistakes", which implies IMO a much higher rate of success

hypna@lemmy.world · 1 pts · 1y

ChatGPT is usually wrong. We apologize for the inconvenience.