Mastodon Feed
Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
xgranade@wandering.shop ("Cassandra is only carbon now") wrote:
Maybe another way of putting all this is that if I look at something like an LLM and ask "how can I make it ethical," I wind up with something that is probably pretty ethical, but it no longer resembles an LLM at all. Like, you can't train one at the scale of an individual artist, or on work that you can meaningfully obtain consent for. The "large" in "large language model" is incredibly load-bearing.