FreeBSD's stance on generative AI

L

Levin

Hi everyone, I hope this is the right place for this.

I would like to preface this by saying that I'm here seeking information, not a debate on AI. I seem to be getting mixed answers online and I'm struggling to find proper answers on the website, so I would like to ask directly: What is the current stance of FreeBSD on its development in regards to AI (LLMs)?

I understand AI is used in the finding and the fixing of certain bugs and vulnerabilities at the moment, but is AI going to be involved further with time? Is there a general stance and philosophy regarding the use of AI in the development of FreeBSD as well as its use for purposes like, say, marketing?

Thanks.
 
Finding problems on software is in my opinion more a matter of repetition of a method that only requires computer logic. What's the "intelligent" part supposed to improve?
 
MG said:
Finding problems on software is in my opinion more a matter of repetition of a method that only requires computer logic. What's the "intelligent" part supposed to improve?
Click to expand...
Does it matter?

I have good results with LLM code reviews, whether I call them intelligent or not.
 
Thank you for your responses.
MG said:
Finding problems on software is in my opinion more a matter of repetition of a method that only requires computer logic. What's the "intelligent" part supposed to improve?
Click to expand...
The larger point for me is not necessarily whether it works as a tool nor the logistics accompanying the development, but rather the more political and philosophical stance being taken. It's a dividing subject, and I wanted to know more before committing to an install. Whether LLM output outweighs software development is another subject.
 
cracauer@ said:
Does it matter?

I have good results with LLM code reviews, whether I call them intelligent or not.
Click to expand...
I maintain cash register/financial systems running on FreeBSD using C and SDL and Qt. I really have no idea how any public, 0% transparent AI service could provide me valuable information to improve things. It's like someone online recommending programming books. Not using any of it until there's solid proof of usefulness.
Generative AI: what does it generate using what?
 
Charlie_ said:
The OP is seeking the FreeBSD Project's perspective, not your personal opinion.
Click to expand...
Likely the same? It has to contribute something visibly....
An AI service finding and solving software problems is nice. 1 that doesn't explain anything about it and could as well achieve the same result with real people has no value.
Any AI needs approved constructive capabilities to be able to contribute to a open source OS project.
 
While I prefaced my post saying I wished for information rather than a debate, I understand why it inevitably became one. 😄
I do think this is an important subject, and I myself have quite strong views against it, but since I doubt that a public debate on the forum would have much impact as to the later decisions being taken, I thought to mainly see about the project's perspective. I received an answer and I will wait for the official policy. Beyond that, this thread doesn't belong to me.
 
As always not a single mention of the massive theft of other peoples work by the AI companies as they vacuum up everything on the Internet to train their models. Once done not a single solitary attribution is given in the results they spit out for your use if you use the results for coding. I for one fail to see how any of that respects copyright and the policies of FreeBSD regarding it in their software development. I am certain they require proper provenance of where the code came from and the copyright of said code before it will ever be included in the OS.
 
I deem any policy largely irrelevant.
If you spam with bot or Ai bot or you give your authentication codes to retarded cousin or commit rubbish code under influence, it is still on your reputation.
Impossible to check or enforce if done by Ai, anyway.
These policies are just void posturing.
Data is either good, bad or ugly.
43e68206a40ffce61e1251fc6ada7598ef96dfceceb5638f98535c698721f7c0.gif
 
As far as BSD group policy, I've been wondering that myself but didn't want to poke the bear, since I have very strong opinions and beliefs on the subject,any policy that is incompatible with my ethics on the matter means that I would not continue to contribute. Of course that is a personal decision that each person is allowed to make for themself.
 
Levin said:
The larger point for me is not necessarily whether it works as a tool nor the logistics accompanying the development, but rather the more political and philosophical stance being taken.
Click to expand...

In the context of kernel development. If you're a kernel developer, ask yourself - would this (politics/philosophy) be of any importance at all to the development of an operating system? I think the former is where the tide meets the sand. AI is merely a tool. If you observe the BSD license - it's a very short and lenient license. I'm sure this paradigm will traverse many future "trends" in the market. From watching many BSD conferences; they (ie. /src commiters) don't wade through unnecessary politics or philosophy. What's best for the development of the operating system is what's relevant here. As you've mentioned; The Foundation sponsoring AI assisted bug/vulnerability discovery and fixing is one solid and conservative example. This ain't Linux, we Think Correctly™.

As a detractor, AI can't replace mindshare and community around intricate low level pieces of software IMO. Would you trusted your data with a vibe coded ZFS clone?
 
Beastie7 said:
ask yourself - would this (politics/philosophy) be of any importance at all to the development of an operating system?
Click to expand...
As i said in the post you've replied to, it is important. The subject is not to do with the performance of the tool (which is a debate of its own), but it is that I take a stance not to support AI for many reasons, some outlined by RedGreen925, others made evident by the state of certain industries and the current culture of AI. To say that AI is "merely a tool" in the context of its use is to disregard what it represents, what it impacts, its ramifications and the broader politics of it which matter to a very large number of people. There are clear ethical problems that come with it and I too am waiting to see whether FreeBSD is or isn't incompatible with my ethics on the matter. So far, NetBSD seems to be the only UNIX-like OS with a "No AI" stance.
 
It's just MY opinion (not FreeBSD project's nor FreeBSD Foundation's), but committing AI(I don't call it "SI")/LLM generated codes into base (src) still has legal issues.

"GPL cannot be inherited to AI/LLM generated codes even if the LLM uses GPL'ed codes for learning" needs to be clarified in international laws and dropped into each regional (each countries / organization like EU) laws.
Otherwise, LLMs that are infecting-license-free (i.e., GPL-free) on training is needed to use the generated codes in legally safe way. This is regardless the quality of the generated codes.

On the other hand, my current impression is that FreeBSD project / FreeBSD Foundation is going to use AI/LLMs to find issues by limited developers NOT to be "slop" but to prevent zero-day attacks.

And getting advice from AI/LLM such as
  • "this part would be better algorithm B instead of current A"
  • "these 2 variables are better merged into one"
or
  • "this function A is better splitting out into generic part and function A-specific part, so generic part can be used for function C, D, E, H and S to simplify and being maintainable"
would be relatively safe, if NOT ask for generating actual codes.

Anyway, the decision is on core team and we need to wait for the final decision.
 
Levin said:
but it is that I take a stance not to support AI for many reasons, some outlined by RedGreen925, others made evident by the state of certain industries and the current culture of AI.
Click to expand...

You're not alone in your concerns. But in open source, the playing field is a bit different. If a contributor decides to go rogue and not track -CURRENT, they lose quite a bit of benefits. Just look at like the whole Sony/Playstation debacle.

Levin said:
To say that AI is "merely a tool" in the context of its use is to disregard what it represents, what it impacts, its ramifications and the broader politics of it which matter to a very large number of people.
Click to expand...

I think you underestimate how tightly knit the FreeBSD community is. We're not just the sum of many different parts fighting to get along. We are one project, by one community.

You should get more involved. ;)
 
Levin said:
So far, NetBSD seems to be the only UNIX-like OS with a "No AI" stance.
Click to expand...

A bit of a history refresher. Both NetBSD and FreeBSD were birthed from 386BSD. OpenBSD came from a new perspective out on NetBSD focusing on ironclad system security.

I'd wager our stances on AI wouldn't differ much. ;)
 
Beastie7 said:
A bit of a history refresher. Both NetBSD and FreeBSD were birthed from 386BSD. OpenBSD came from a new perspective out on NetBSD focusing on ironclad system security.

I'd wager our stances on AI wouldn't differ much. ;)
Click to expand...
NetBSD is more of a hobby project/experimenting - not targeting servers or desktops, tho? They can afford to go Amish a bit more.
Wonder what OpenBSD is doing re security procedures.
 
Levin said:
As i said in the post you've replied to, it is important. The subject is not to do with the performance of the tool (which is a debate of its own), but it is that I take a stance not to support AI for many reasons, some outlined by RedGreen925, others made evident by the state of certain industries and the current culture of AI. To say that AI is "merely a tool" in the context of its use is to disregard what it represents, what it impacts, its ramifications and the broader politics of it which matter to a very large number of people. There are clear ethical problems that come with it and I too am waiting to see whether FreeBSD is or isn't incompatible with my ethics on the matter. So far, NetBSD seems to be the only UNIX-like OS with a "No AI" stance.
Click to expand...
Reality check:
Policy/politics - Culture - ethics - morality is a function of survival.
Example ad absurdum: If cannibalism is necessary for survival, eating people will be ethical/moral. Or your tribe perishes and with them their ethics.
 
DanielBowman said:
NetBSD is more of a hobby project/experimenting - not targeting servers or desktops, tho? They can afford to go Amish a bit more.
Wonder what OpenBSD is doing re security procedures.
Click to expand...
I think NetBSD is a kind of "academic project".
If I understand correctly, they prefer "technical correctness" and "sanity" over performance and "quickness in development".

FreeBSD project went different way. "Performance and stability first".
And as the project grows to multiple archs, seem to "lean to" the way NetBSD chose at the first place for maintainabilities. (Putting ports aside, scale of the in-base codes have grown, but the resources of the project haven't grew at the same pace, unfortunately.)
 
DanielBowman said:
Reality check:
Policy/politics - Culture - ethics - morality is a function of survival.
Example ad absurdum: If cannibalism is necessary for survival, eating people will be ethical/moral. Or your tribe perishes and with them their ethics.
Click to expand...
But who's facing a survival situation here? OP or FreeBSD?

If neither, aren't you trying to say that we shouldn't be concerned with ethics and morality altogether simply because we can imagine such a hypothetical case where we could overstep our beliefs?
 
sualbee said:
But who's facing a survival situation here? OP or FreeBSD?

If neither, aren't you trying to say that we shouldn't be concerned with ethics and morality altogether simply because we can imagine such a hypothetical case where we could overstep our beliefs?
Click to expand...
Both are facing survival situation. And even a meta one for all life on Earth, ultimately :)
That is also why the debate is often underlined by apocalyptic visions, btw.
[ Ai eats us VS Tech brakes keep us all, including flora, stuck in dying Solar system]

I am trying to say, that ethics are not set in stone...beyond continuation of existence, let's say, the objective, ultimate ethics measurement.

Concretely, regarding FreeBSD, to survive the competition, it should not limit itself by arbitrary proclamations about the use of a technology.
Refuse to utilize gun powder, because whatever, and your lofty ethics will bleed out with you. Just don't store it next to your stove.
 
You must log in or register to reply here.
Back
Top