• 0 Posts
  • 13 Comments
Joined 3 年前
cake
Cake day: 2023年7月4日

help-circle
  • The reality is that it is a next word prediction machine. There probably aren’t any examples in the training data where people are writing music reviews on something that isn’t music. It probably interprets the sound as best as it can as “music” (and the best it can do this is likely very bad in the first place), and then, since the prompt was about reviewing music, it uses next word prediction to write a music review, which of course turns out looking like a typical music review. It’s not really interpreting the sound as “not music” especially since you told it it IS music.









  • Look, I recognize that it’s possible for LLMs to produce code that is literally someone else’s copyrighted code. However, the way I use copilot is almost exclusively to autocomplete my thoughts. Like, I write enough code until it guesses what I was about to write next. If that happens to be open source code that someone else has written, then it is complete coincidence that I thought of writing that code. Not all thoughts are original.

    Further, whether I should be at fault for LLM vendors who may be breaking copyright law, is like trying to make a case for me being at fault for murder because I drive a car when car manufacturers lobby to the effect that people die more.