I primarily use GPT-4 for coding and questions related to scientific research. If I’m asking a question where I don’t foresee a need for GPT to use code, is there any downside to turning on Code Interpreter regardless?
I’ve found that it really wants to write code, so much so that it find reasons to even when it’s not strictly necessary.
My understanding is the underlying model is more-or-less the same as base GPT-4, so I think it’ll be fine and of comparable quality—you just might see it write a function or two you’re not expecting.
I’m hearing (Latent Space etc.)
- The model is different / better
- It’s worth using all the time