Thoughts on the recent Debian GR

If it’s not a complete ban of gen AI usage, I would prefer to do not have any specific policy, but I’m not sure if it’s possible to do that at this point.

A lot of people share the same arguments before and in other forums, but I will try to summarize why I’m against any usage of LLM and these kind of tools in any open source community (or any other group of people).

Personal

  • It removes the fun from all the processes. I’m participating in gnome mainly for fun, I like to write code by myself and with other people, and if the code is written with an LLM, I won’t have anything to do here. It’s like playing chess with cheaters using stockfish, not fun.
  • It degrades the ownership relation with the software, if you craft something from scratch, and it takes months to create, there’s a relationship with that piece of software that’s deep, it’s part of you and you are part of that project. But if something is “easy” and “cheap”, it’s easier to just forget about it and “buy” something new the next week.
  • Mentoring is nonsense nowadays. Mentoring newcomers is “fun” and useful when there’s another person on the other side. You help that person to grow and the other person contribute back and motivates you to be better. There’s a knowledge share that will last in time, it’s how the community grow and how the newcomers became the future members of the living community. The LLM usage breaks all this process, there’s no possibility for any community to grow and there’s no space for newcomers to become “experts” in the future.

Technical

  • It doesn’t work, LLM generates a lot of noise, it’s the worst way of solving almost any problem. It’s not accurate and it’s not reproducible. At the end is like trying to brute-forcing any problem with statistics. Looks like it’s working because there’s a lot of money and resources there, but I can’t use any of these fancy LLM locally in my cheap laptop with 8 or 16GB of RAM.
  • It floods any project, you just need to look to what’s happening in the Linux Kernel mailing list. It’s impossible to review all the text (or code) that’s generated. It’s “impossible” to argue with the LLM output, it always wins just generating more noise that you can handle. And all that noise will hide anything “important”. Allowing the usage just makes unusable any collaboration tool (how many projects are just disabling issues/PR in github?)

Political

  • All the companies (and promoters) behind these “tools” are the devil in earth. They do not hide or try to sell that they are good people. And just aligning with the devil, even if all we could ignore all the previous points, I won’t feel comfortable delegating the “thinking” to the worst people in the world. Or at least, if we remove the moral part (good or bad), I think that’s not intelligent to delegate the “thinking” to the people that have the opposite interests than me.
  • The ridiculous cost of this tech do not just destroy “virtual” communities, it also destroy the environment and places where the new data centers are built. It’s wrong to be part of that, and it’s even worst to be proud of it.
  • I’m a professional software engineer, the usage of these tools is devaluing my work directly. It’s something stupid to support something that makes you work more for less money.
  • It’s a direct attack to the humanism, the current trend is trying to center any creation act around the “final product”, without thinking about the process or the consequences. Accepting this is just accepting “The end justifies the means”

And I’m sure that just thinking a bit any of you will be able to find a lot more reasons to justify to do not use any LLM anymore. I personally can’t find any reason to use this tech, it’s not useful for me. Other people could find this useful, but as almost anything in our life, it’s not just a personal decision, every decision is political and affect other people around you.

So I will try to fight back against what I consider wrong, in every space where I have a voice, and I’m deeply sure that all of this new LLM is completely wrong, for learning, for coding, for designing, for drawing, for writing, for translating or for anything that you can do by yourself or with friends.

20 Likes