> Unfortunately, all of these issues come from humans.

They are. They’ve always been there.

The problem is that LLMs are a MASSIVE force multiplier. That’s why they’re a problem all over the place.

We had something of a mechanism to gate the amount of trash on the internet: human availability. That no longer applies. SPAM, in the non-commercial sense of just noise that drowns out everything else, can now be generated thousands of times faster than real content ever could be. By a single individual.

It’s the same problem with open source. There was a limit to the number of people who knew how to program enough to make a PR, even if it was a terrible one. It took time to learn.

AI automated that. Now everyone can make massive piles of complicated plausible looking PRs as fast as they want.

To whatever degree AI has helped maintainers, it is not nearly as an effective a tool at helping them as it is helping others generate things to waste their time. Intentionally or otherwise.

You can’t just argue that AI can be a benefit therefore everything is fine. The externalities of it, in the digital world, are destroying things. And even if we develop mechanisms to handle the incredible volume will we have much of value left by the time we get there?

This is the reason I get so angry at every pro AI post I see. They never seem to discuss the possible downsides of what they’re doing. How it affects the whole instead of just the individual.

There are a lot of people dealing with those consequences today. This video/article is an example of it.