Will FreeBSD adopt a No-AI policy or such?

Just finished reading an article about how AI cannot replace one's use of craftsmanship which, coincidentally, is a tag in my byline. A point briefly mentioned, but not dwelt on, made me ask, "Will AI stop us from asking questions? If you don't ask questions then you aren't exploring. If you aren't exploring how do you discover?"

Since AI's results only happen from past information, it doesn't seem it can discover new areas on its own. I don't mean finding bugs or a fancy algorithm. I mean could it discover digital logic before anyone knew what digital logic was? Could it develop FreeBSD before anyone knew what an operating system was? I don't think so.
 
The cost and accountability I mentioned earlier are not only the impact on craftmanship or the slop produced, but environmental impact. Also the price for memory and other hardware is spiking. That's what you are paying out of pocket. The other price you pay out of your time to live is still unknown. And the most funny thing is, you are paying it while someone pockets the proceeds. Wether you use AI or not.
 
As long as shareholders pay, with speculative money. No problem for me. But somehow this has relation with "quantitative easoning". Haha, catch you.
 
Alain De Vos said:
No problem for me.
Click to expand...
Good for you.
People are paying with increased energy costs, noise, pollution, hardware costs, ... Our children and grandchildren will pay for this with more global warming, storms, forced migration, ... It is not so much the AI slop that is bothering me (that can be avoided), it's these extra shite they put on all of us, no matter what we do or want.
 
You cannot escape. Its called "quantitative easining". Its money which does not exists goes to big banks then to small banks to investors. Who have big ideas but not for your good. The problem is inherit in economic system. And its not ai to blame. So question is how get ai get its money.
Even if there is no turnout or revenue. ? Data in star-trek would say illogical. :)
 
I will say more go to local bank. Ask a little bit of money, for house with your wife.
They will ask. What do you make net new on your account. Each month with proof.
If not ok, see you later alligator. Come back later.
But AI companies are above. They dont need to prove nothing.
Full details i do not understand. But like they say in English, "somethings wrotten in the state of Denmark".
 
Here would be my proposed policy: FreeBSD will only accept good quality code. And code that has a copyright and a license that allows FreeBSD to use it. Period, end of policy.

Whether that code is written by real men, real women, or real furry creatures from Alpha Centauri makes no difference at all. If an AI can write code every bit as good as that of other submitters, and the AI can make a reasonable (believable) argument that the copyright/license situation is no problem, that's fine with me. Exactly the same as if Marvin the Martian had written the code.
 
ralphbsz said:
Here would be my proposed policy: FreeBSD will only accept good quality code
Click to expand...
This misses multiple problems with AI/LLMS:

1. Skill erosion of AI users: https://www.nature.com/articles/d41586-026-01947-1

2. AI code leading to more bugs even after a human reviews it: https://www.coderabbit.ai/blog/state-of-ai-vs-human-code-generation-report If the people doing the reviewing are also using AI, as per point 1 the quality of that review is expected to drop.

3. Most FOSS projects never had a problem writing as much low quality code as they wanted, but rather a problem with focusing on actual high quality code. This is the one thing AI cannot do since it seems to be incapable of basic logical thought: https://www.forbes.com/sites/cornel...what-apples-ai-study-reveals-about-reasoning/

4. The whole copyright mess, I'm not a lawyer but the common "the courts already decided" arguments seem to focus on the US and here are some EU sources (this isn't legal advice, read for yourself): https://github.com/mastodon/mastodon/issues/38072#issuecomment-4105681567 https://www.twobirds.com/en/insight...-(gema-v-openai)-on-copyright-and-ai-training https://the-decoder.com/landmark-ge...-words-and-makes-it-liable-for-false-answers/

5. The addiction and vendor lock-in. While you could run local LLMs for coding, most heavy users seem to refuse to do so and want to use e.g. Claude Code instead. Now once the AI bubble bursts, you have a bunch of people that became dependent on it and are now hooked to a subscription just to be able to contribute to a free open-source project.

6. All the environmental and ecosystem effects already mentioned above.

7. The moral effect of the AI plagiarism discussion moving the window from "we shouldn't copy from other projects and how do we enforce this even though the typical project doesn't want to sue" to "how about we just copy from everywhere as long as it's short enough not to get sued". This seems bad for FOSS as a whole and devalues other projects and their consent, whether it's legal or not. (Yes it would be in theory possible to train a code model only on projects that opted in. But so far nobody could show me a model that did so.)

8. The effects of how AI coders will no longer search the web for snippets and libraries, instead generating their own code, and will never find out what upstream projects may have helped their problem and hence projects are isolated and the community suffers: https://www.404media.co/vibe-coding-is-killing-open-source-software-researchers-argue/

It seems to me like the only way to address most of these is to reject any LLM code contributions no matter how small or how well-reviewed, and to at least discourage LLM use for other tasks as well.
 
itstotallyme said:
1. Skill erosion of AI users: https://www.nature.com/articles/d41586-026-01947-1
Click to expand...
Kind-of raises a question of open-source vs proprietary: Being a dev and accepting contributions also needs checks to avoid questionable code, whereas the time to cross-check "other's" code could be spent on dev.

If I coded something, I could use AI. It feels odd to have an unrelated person try to submit code from an AI/LLM: What's their motivation? Do they think they're more-knowledgeable that AI exists than the person making software today where AI is well-known everywhere? Do they just want a contribution name tag with minimal effort? Why do users want to be middle-men for AI/LLM when projects/lead devs can use AI/LLM directly?
 
One question is what happens when person X gets their AI assisted (or generated) code, then moves on and this code has a bug or has to satisfy some new requirements? Another person Y would likely have problems understanding such code if it was not properly vetted in the first place. Y may use some other unrelated AI to grok the code and evolve it but overall I have my doubts that the quality of code would improve.....
 
You must log in or register to reply here.
Back
Top