On FLOSS and training LLMs
5 points by gerikson
5 points by gerikson
14th day of Chaos 3192 YOLD
Shame on the Discordian who defends copyright!
First of all, hit them where it hurts: deny them access to your work.
That's not where it hurts! What hurts a corporation is wasted money; cash is to businesses as blood is to humans. The reason that we have Llama is because it was exfiltrated from Meta and distributed via Bittorrent; Meta decided to try copyrighting the weights, immediately disqualifying them from trade-secret status and ensuring that pirates will always have a copy. To dramatically quote from the "we have no moat" memo:
Feb 24, 2023 - LLaMA is Launched: Meta launches LLaMA, open sourcing the code, but not the weights.
March 3, 2023 - The Inevitable Happens: Within a week, LLaMA is leaked to the public.
What hurts a corporation is wasted money
While this is true in general, it doesn't appear to be true of langlemangle companies. Because the entire US economy has been rearranged around them, losing unprecedented amounts of money doesn't seem to have an effect on them, similar to how losing a head doesn't necessarily cause a chicken to stop running around.