For the past few days I have been thinking if it theoretically makes sense to have a #LLM trained exclusively on #GPLv3 ...

For the past few days I have been thinking if it theoretically makes sense to have a #LLM trained exclusively on #GPLv3 code or even stricter, a specific language like C of Scheme or Common Lisp.Does that somehow "solve" the licensing issue if the generated code is also GPLv3?Can it be used for autocomplete, a.k.a tab completion, of that specific language in a GPLv3 project?How about MIT or BSD licenses?Note: This is just a thought experiment as I don't have any means to do any of these.

Read Original

Related