FreeBSD's stance on generative AI

Status
Not open for further replies.
DanielBowman said:
Nice strawman you have there.
a] Ethics of a gun is pointless. Ethics of unnecessarily shooting people is paramount.
b] Promoting commercial big tech AT&T Unix systems, by advocating for not banning the use of computers....hmmm
Click to expand...
Are you sure I'm the one constructing a strawman? Obviously, as has been mentioned thousands of times everywhere, including in this thread, when we're talking about ethical concerns surrounding LLMs, we are not talking about the technology itself, because of course it'd be pointless. It simply exists. Just as when we are talking about weapons, it's not about its theoretical physical properties. What matters is whether we should build them, who controls them if they have been built and what the consequences of their (possible) use are. OP clearly expressed these concerns and more, but in return you've pointed them to the transience of ethics.

DanielBowman said:
But we can agree we have different values. I for example value the truth more than a fear. Therefore I prefer options to be open.
Click to expand...
This one is particularly nice, especially considering that I didn't say anything about banning the use of LLMs. And seriously, do you really think that I want to be controlled and value fear over truth? What is this, a kindergarten?
 
sualbee said:
Are you sure I'm the one constructing a strawman? Obviously, as has been mentioned thousands of times everywhere, including in this thread, when we're talking about ethical concerns surrounding LLMs, we are not talking about the technology itself, because of course it'd be pointless. It simply exists. Just as when we are talking about weapons, it's not about its theoretical physical properties. What matters is whether we should build them, who controls them if they have been built and what the consequences of their (possible) use are. OP clearly expressed these concerns and more, but in return you've pointed them to the transience of ethics.


This one is particularly nice, especially considering that I didn't say anything about banning the use of LLMs. And seriously, do you really think that I want to be controlled and value fear over truth? What is this, a kindergarten?
Click to expand...

You are talking about controlling the use of the technology, because of it's unethical origins. Matter of if we should even build them takes it to even higher level of control.
You are just contradiction yourself in the first paragraph.

I pointed out that ethics is derived from survival needs. And that limiting Ai in FreeBSD might disadvantage FreeBSD.
And apart from unnecessary harm, ethical rules are seldom unchanging.
Maybe re-read the thread again.

People who have desire for control, do that out of fear. I said nothing about you wanting to be controlled.
We can debate the merits of that fear but you already proclaimed them valid while dismissing mitigating qualities as practically null.
 
cracauer@ said:
The actual submission still has to come from a person who is responsible for it, and who understands the code submitted.
Click to expand...
Ah. "accountability". So if [you] understand the code then why not just write it [yourself] and eliminate the question?
...my point was more along the lines of defining what constitutes a 'contribution' (and how utterly asinine that slippery slope exercise is)? -ie. "speed" is often not necessary (or counter productive) when working with the complex.
-eg. change one `strcmp()` to `strncmp()` and you change the entire meaning/use.
 
cracauer@ said:
Question:

If we outlaw LLM contributions - what keeps people from submitting anyway without disclosing the LLM help?
Click to expand...
msplsh

Post in thread 'Will FreeBSD adopt a No-AI policy or such?'

cracauer@ said:
If you forbid that people will just stop disclosing that fact.
Click to expand...
Sure, but

1) if they do that, they'll have to cover for it, which is difficult, and why would you make your life more difficult for a project that doesn't want your contributions in the first place

2) let them ruin their own reputation for not being able to follow rules

and 3) as many people have said, you don't have to worry about them not disclosing it, they'll volunteer this information.

Should be self-selecting.
  • Thanks
 
JohnK said:
Ah. "accountability". So if [you] understand the code then why not just write it [yourself] and eliminate the question?
Click to expand...
It might be easier or quicker to review supplied code that purports to solve the problem than to create your own from scratch.

Today I read that some high level math problems are being solved but in ways that mathematicians don't understand. Is math or code supplied by AI but not understandable safe to use? I don't think so. Is understandable code reviewed by a knowledgeable FreeBSD developer that solves a problem a good thing? Sure. Why not?

Should AI be allowed to modify code on its own without review? Oh hell no!
 
Levin:
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?
Click to expand...
Going back to Multics and the early days of Unix. What is the original Unix philosophy?

I just wanted to add that context to this thread.
 
cracauer@ said:
Question:

If we outlaw LLM contributions - what keeps people from submitting anyway without disclosing the LLM help?
Click to expand...
Nothing, But taking a stand is saying that you don't condone this behaviour, even if it does happen, even if it is inevitable. Taking a stand means that you're in favour of human code, despite what people commit. It's to remove certain commitors the moment you learn without a doubt that they're submitting LLM-generated code. It can't just be all or nothing, you can't just say "Well, we can't help it, so we have to allow everything."
 
cracauer@ said:
If we outlaw LLM contributions - what keeps people from submitting anyway without disclosing the LLM help?
Click to expand...
I don't think it's consistent to discuss exactly how generated code could be reliably prevented from getting into the tree without agreeing whether it's necessary to do so in the first place and why, isn't it? To answer the question: probably nothing. Just as nothing kept people from submitting otherwise tainted or sloppy code all those years. This doesn't make such submissions any more or less welcome.
DanielBowman said:
You are just contradiction yourself in the first paragraph.
Click to expand...
Yeah, I see what you mean. And no, I don't think I am. But this conversation has become rather pointless and off-topic, so I'm out. At least we both seem to agree on the importance of ethics and that not causing unnecessary harm is paramount, which was my point. What we likely won't agree on is where the boundaries of such necessity lie, because I can hardly imagine any justification at all for even more environmental disruption at scale than has already been done. The mere possibility of better OS security, wifi drivers and whatnot is certainly not enough for me, which by the way is quite a noble application of the technology we're talking about compared to its overwhelmingly common use case.
 
Even AFTER legal issues are cleared ALL OVER THE WORLD WITHOUT SINGLE EXCEPTION, maybe any submitter of AI/LLM generated codes would need to ask AI/LLM something like below BEFORE SUBMITTING.
  1. Please confirm if the changes are easily maintainable by human maintainers.
  2. Please confirm if the changes are easily readable and understandable by human reviewers.
  3. Please fix the changes to be maintainable, readable and understantable for human.
And afterwards, confirm the person him/herself BEFORE AI/LLM FIXED THE CODES and there are no bug reports attempting to fix the same issue to avoid duplicates. If the final changes in hand looks better than already reported one(s), propose the fix to exactly same bugtracker (i.e., Bugzilla for FreeBSD) ticket.
 
Most intelligent people I know have changed their minds on AI. A rational opposition to AI is needed, sure, but it won't come from the grifters in 404Media & Tech Won't Us. To criticize something you must first understand it, which is not the case for the professional AI haters in the Fediverse. This technology is proving itself useful in so many ways that I don't feel the need to change anyone's mind.

What we need right now is open-weights & open-source models that we can run, and we just have to wait for hardware prices to come down. Most ethical & technical issues will be solved by then. Like every technology, it will create more jobs than it destroys.
 
rbranco said:
Most intelligent people I know have chang
Click to expand...
You should try a different flavor of kool-aid. I've heard grape is nice.
I am free to do/choose/use what I want so why do [you] often find it a requirement to push/sell/remind/etc. (the ai companies have their own marketing teams and they don't need you). Do you push this hard about which keyboard to use?
 
Status
Not open for further replies.
Back
Top