ESR has become a big fan of AI-assisted code writing, and as a result I’m a bit less “doom-y” about the potential uses of AI because ESR is a living legend in the software world. If he thinks it’s a good thing, then I’m inclined to believe him. That said, as he illustrates here, not everyone in the OS community agrees with him about code-writing clankers:
I find it bizarre that so many open-source projects have posted policies against AI-assisted contributions.
This will not last, because they’ll be outcompeted by projects not afraid to use power tools to debug the code and and features. In crucial cases, AI-hostile projects will be forked to remove that restriction.
How can anybody not see this coming?
This dynamic is so obvious – and anti-AI rules so patently futile and doomed – that I have to conclude there is something like a religious frenzy going on.
I’m going to bend over backwards to be fair about this. There was a point earlier in the development of AI at which such restrictions might have made temporary sense, because model outputs were prone to enough noise and hallucination to be a net drag on productivity.
We’re long past that now. And the obvious solution to merge-request queues flooded with AI-generated slop is to use AI to filter them so exercising human judgment on the remainder remains practical.
I’m not going to argue rights and wrongs here. One might just as well argue that it’s wrong for sandcastles to collapse when the tide comes in.
The religious frenzy does bother me, though. Maybe less than it did 5 minutes ago when the mind infection was “wokeness”, but a mob of crazies is never an edifying or reassuring sight.
I guess the most positive thing I can observe is that this phase of nuttery won’t last long, because it can’t. The selective pressure against it is too brutal.
It does leave me wondering what cause the nutters will attach themselves to next. Alas.
