The reason I personally don’t like Copilot is the same reason I dislike text prediction: I already have planned what I am going to write, either in code or in prose, a little bit ahead. Seeing a prediction that may or may not even make sense given the context in my head breaks my train of thought and slows me down. This is pretty clearly a personal problem. I could see how if someone got stuck it could be useful, but even seeing that prediction means it takes me 2-3x as long to write the thing I actually want to write. The prediction doesn’t generally fit with what I want to do or say even though it could logically follow for an average of what all the training data did or said! I don’t even like predictive text or autocorrect when texting on my phone for the same reason. This isn’t an “stuff with AI” gripe, just the user experience of the way these things have been designed for about ten years does not work for the fact that I fully compose what I want to say or code beforehand in my head and then input it.
