obsidian topaz
Both ChatGPT-4 and 3 appear to frequently make errors with simple calculations, something even a basic calculator can handle correctly. While occasional mistakes can be understandable, the consistent occurrence is concerning. Please address and rectify this issue at your earliest convenience.
https://chat.openai.com/share/11fad5ab-9a86-484a-8b33-dfb947d9ea6a
frank mortar
Both models are "only" language models, not math models. I'd recommend enabling the Wolfram plugin to handle math queries.
Longer response: ChatGPT can usually answer simple math problems like 1+1 or 2*2 because they arise relatively frequently in natural language. However, there's no "math mode" that vanilla ChatGPT can enter, so when the math becomes more than basic, the model is still treating and responding to it like language, which means it's going to be a little random each time, and only get you kind of close to your answer.
turbid prairie
Both ChatGPT-4 and 3 appear to frequently make errors with simple calculations, ...
I saw this too recently. it's like it doesn't even understand the math it is regurgitating. because it seems to make many errors and very consistently.