@joeyh we don't know yet about (1) and (2) (there is preliminary science about it, but results are inconclusive on pros/cons). What we know is that code LLMs are now a tool that developers use, together with IDEs, compilers, etc.
We don't want free software to be at disadvantage wrt proprietary software in not having access to them.
The interesting question is how do we build a FOSS-friendly code LLM. (Bonus point: how do we make it *disadvantage* proprietary software.) →
@zacchiro @[email protected]
Not all open source licences are compatible. Is there going to be a GPLv2-only LLM, an LGPLv2.1 LLM, a GPL3 LLM, an AGPL3 LLM, ... ?
Is the LLM going to declare a license on the output?
@vorlon with my teacher hat on, I'm also still quite skeptical on the net benefits and waiting out what empirical scientific results will tell us (as mentioned up thread, there are some, but they are not conclusive yet).
But boilerplate alone is a very interesting one. I think it has the potential of alleviating the most frustrating part of the work life of professional programmers, letting them focus on the most rewarding parts. (There too: science doesn't know yet.)
@zacchiro This is a bit of a tangent, but with my programming language person hat on, I think boilerplate in programs is a sign of poor programming language design.
Java IDEs were famous for helping with boilerplate; the question should have been: how can we design the language and tools to reduce or avoid that boilerplate in the first place? Extensible programming languages are very good at that.
@zacchiro Right, though it’s a matter of perspective. On one hand, we might want to help developers “where they are”—helping them write boilerplate because their tools require it anyway.
On the other hand, if we are to empower people through programming, we should probably produce and teach tools that avoid the problem entirely, rather than take away some of their autonomy by making them dependent on “boilerplate as a service”.
Soon after its release, GitHub #Copylot was caught distributing #GPLv3 code from Quake 3 Arena, with a wrong attribution and permissive MIT-like license. That's why I call it #CopyALot.
For a famous piece of copylefted code tht was recognized, the work of thousands of less known free software developer is going to be included in proprietary products without even the offending developers being aware of the theft.
#LLM "trained" on #OpemSource software can only be used to ethic-wash the practice, so that most of opensource developers won't realize how they are fooled and their work expropriated.
Stefano Zacchiroli
vorlon
Ludovic Courtès
Chris [list of emoji]
Shamar