Ariadne Conill 🐰 on Nostr: npub17qejc…gv7la i think the GenAI issue is very complicated, the licenses clearly ...
npub17qejctd76jm8syfmpkde4nlur4avy2fhthwk75yce2ykes08zqjs3gv7la (npub17qe…v7la) i think the GenAI issue is very complicated, the licenses clearly allow anyone to study the source code, which certainly maps to using it as data to train ML models.
however, the people pushing back on GenAI training are largely doing so because they believe it to be a risk to their livelihood.
i’m not so certain of that. to me, the creative aspect of software development is figuring out the “what,” not the “how”. GenAI models can only generate what they are told to generate, which means that the actual value-creating part of the labor is not at risk.
on the other hand, training GenAI models with code without tracking where it came from or what license it is under carries significant legal risk. if it doesn’t carry legal risk, then copyright as it applies to software source code is useless and nobody has any intellectual property anymore in software. this is likely dangerous in its own right for developers’ livelihoods if true.
however, the people pushing back on GenAI training are largely doing so because they believe it to be a risk to their livelihood.
i’m not so certain of that. to me, the creative aspect of software development is figuring out the “what,” not the “how”. GenAI models can only generate what they are told to generate, which means that the actual value-creating part of the labor is not at risk.
on the other hand, training GenAI models with code without tracking where it came from or what license it is under carries significant legal risk. if it doesn’t carry legal risk, then copyright as it applies to software source code is useless and nobody has any intellectual property anymore in software. this is likely dangerous in its own right for developers’ livelihoods if true.