Post B5FUZoE6AWuZJKQDGS by libreleah@mas.to
 (DIR) More posts by libreleah@mas.to
 (DIR) Post #B5FTOTPZ7PKtc9ofM8 by libreleah@mas.to
       1 likes, 0 repeats
       
       FUN FACT: The Linux kernel now accepts AI-generated code:https://www.tomshardware.com/software/linux/linux-lays-down-the-law-on-ai-generated-code-yes-to-copilot-no-to-ai-slop-and-humans-take-the-fall-for-mistakes-after-months-of-fierce-debate-torvalds-and-maintainers-come-to-an-agreementhttps://docs.kernel.org/process/coding-assistants.htmlSo.... now, essentially, Linux cannot be trusted legally. Who knows what code the AIs were training on? It's almost certain that use of AI-generated code will violate someone's copyright.Good timing, since I've been planning to migrate my main system to OpenBSD. Now I definitely will.AI can be used productively for code auditing, but vibe coding the Linux kernel seems like a bad idea.
       
 (DIR) Post #B5FUGM1Ioc5kKP9xsu by cryptgoat@fedifreu.de
       0 likes, 0 repeats
       
       @libreleah Not a fan of LLM code at all, but this is pretty much the most negative interpretation you can have on this.Still, glad that OpenBSD has a very clear stance against AI/slop code.
       
 (DIR) Post #B5FUZoE6AWuZJKQDGS by libreleah@mas.to
       0 likes, 0 repeats
       
       @cryptgoat what would be a positive interpretation?i'm actually not anti-AI, i think AI is wonderful in fact, for auditing. publishing AI code is a legal minefield because you are essentially re-licensing someone else's code. the output itself is theoretically fine, since it's machine-generated, but you cannot avoid that you're going to be using someone elses's code. that's how AIs work. they're basically predictive text on steroids.use it to find use-after-frees. don't publish code with it.
       
 (DIR) Post #B5FVJMs8CguBnGs3jU by libreleah@mas.to
       0 likes, 0 repeats
       
       @cryptgoat even if it's all open-source code, who is the author? and this is linux, which is a GPLv2-only project. maybe the code in the AI output contains someones GPLv3-licensed work which would be legally incompatible.that's essentially the problem, with using AI code. even if the AI always producted the best possible patch (and it's getting pretty good nowadays), copyright law just isn't wired for it. the only way to fix it would be idk, abolish copyrights on software. another minefield :)
       
 (DIR) Post #B5FYUxBVpn8zSm9VfU by tom@tomrees.social
       0 likes, 0 repeats
       
       @libreleah yes I feel Linus' "It's just another tool" argument is a little reductive.
       
 (DIR) Post #B5FYmMYsjjxb5iNuOe by libreleah@mas.to
       0 likes, 0 repeats
       
       @tom unfortunately, it's not just linux. it's literally everything. it's impossible to know who used what code, in large projects. it was like this before AI, AI just made it much easier to infringe on copyrights at scale.i think that there needs to be a worldwide discussion at the political level, to reform copyright laws in a way that satisfies this issue.as i said, i'm not actually anti-AI. i'm not pro-AI either. AI *is* a tool - a legally dubious one, until the legal issues are resolved.
       
 (DIR) Post #B5FZgJu5o19743zKZk by libreleah@mas.to
       0 likes, 1 repeats
       
       @tom the legal issues (with using AI output) is one of two issues i have with AI.the other issue is that AI companies like e.g. openai have a monopoly on training / models. they are essentially bypassing national governments and controlling how people think at a global scale. the threat to democracy is extreme.so AI should be properly regulated, and control over it ought to be democratised. so many problems.we also need to build lots of solar/wind farms and not burn fossil fuels to run it.
       
 (DIR) Post #B5FjoVnMTbyvnsNuQC by kai@ajin.la
       0 likes, 0 repeats
       
       @libreleah https://hj.9fs.net/khm/p/1774962768.631573