I guess, with more and more powerful LLMs coming — we are the last generation of engineers who are manually typing the commit messages (and maybe typing at all lol)
And thus deprive future LLMs of training data. Or, rather, force future LLMs data pipelines to learn throwing away previous LLM-generated data from the training set, letting through only the human-generated commit messages. N-n-nice! This is how the new technology's tail strikes you hard at the forehead when you think you're running up front. Cool! That's something to philosophise about!
And thus deprive future LLMs of training data. Or, rather, force future LLMs data pipelines to learn throwing away previous LLM-generated data from the training set, letting through only the human-generated commit messages. N-n-nice! This is how the new technology's tail strikes you hard at the forehead when you think you're running up front. Cool! That's something to philosophise about!
hahah :)