So they're controlling the output to make ChatGPT "better". They're not making a better model to make ChatGPT better. Isn't it a bit of a waste at this point t

So they're controlling the output to make ChatGPT "better". They're not making a better model to make ChatGPT better. Isn't it a bit of a waste at this point to spend time on doing that?