Perfect and Boring
If you have used ChatGPT and its colleagues quite a bit, you will ave noticed that its outputs end up being quite similar to one another in their structure, in their word choice, in their content: it’s many of the same, and in a way these outputs, while grammatically impeccable, are quite soulless and eventually some average, nobody’s language. (This is at least by default, one could try to push for some more creativity, but they’ll eventually be average-creative.) And this then begs the question: what is a good output? And thus: what is a good model?
