The panel was working through Debian's decision to permit AI-assisted contributions, and a co-host had just said the vote left him feeling defeated. Morrison turns the complaint back on the show: the models were trained on the millions of lines of open source code they spent decades urging people to publish. He lands it as a joke at a colleague's expense and then calls it depressing himself.
With the decision to say you as a contributor are responsible and you can't say, oh, well, the AI did it. So it's not my responsibility. Are they including copyright in that? So you as a contributor are responsible?
…to it, it used to be that, right, fuck it, that's going on the shit list. But between Debian, the kernel, everything, it's fucking everywhere. And so what can I do? I can still moan about it, but I can't just not use any software that's got AI code in it, can I? That's just not realistic now.
I had a horrifying thought that, in fact, maybe all of this kind of open source advocacy we've been doing for decades, we've actually become part of the problem. If it wasn't for the readily available millions and millions of lines of open source code, we wouldn't have LLMs in the first place. So in fact, it's all your fault, Joe.
Yeah, it is. It's completely my fault.
Such a depressing take.
Oh, God. I think in Debian's case, as in lots of very popular open source projects, there is a choice between drown under the never-ending sea of pull requests and bug fixes and comments and not even bug fixes, but bugs being logged. I think that the answer for these projects is to fight fire with…