In praise of randomness

Sean has a great article here on the “randomness” of computers generally, with this observation on AI:

generative AI, too, can act as a force for homogenization. Nothing any model produces is truly random. After all, generative AI is only novel insofar as its training data allows, which means it can only “remix” that which it has already “seen” in the past. It feels novel to us because none of us has seen nor read the entirety of the Internet.