Will FreeBSD adopt a No-AI policy or such?

LibreQuest said:
I would disagree, the idea of forcing an idea or a program onto others is not ethical. Implying that one must or has to accept something is also typically "doomerism" inspiring.

Artificial Intelligence research and applications have been around for many decades represented in both hardware and software. The unregulated abuse of the open web, I believe, should not be tolerated. Technology should not be programmed to disregard the rights of the users, the license of different works, or the ability to choose not to participate. The alternative is force, and force always has resisitance.

Deciding what "we" do with it implies that "we" have a right to use the training data in these programs. And that is very questionable if not outright an abuse of the open internet.
Click to expand...
If we discuss the ethics:
1. We'll never agree on them pleasing everybody or even most people.
2. We genuinely risk that other countries develop more powerful models because they don't have the same moral itch as us. We must know how the models behave with all the data produced by mankind. It's a Faustian bargain that I completely understand.

The discussion is impossible right now because there's lots of noise and the extremes only see black & white.

We can only be pragmatic about this pushing for open models, open data and so on.
 
rbranco said:
If we discuss the ethics:
1. We'll never agree on them pleasing everybody or even most people.
2. We genuinely risk that other countries develop more powerful models because they don't have the same moral itch as us. We must know how the models behave with all the data produced by mankind. It's a Faustian bargain that I completely understand.

The discussion is impossible right now because there's lots of noise and the extremes only see black & white.

We can only be pragmatic about this pushing for open models, open data and so on.
Click to expand...
Well, I don't think the goal should be pleasing everyone but not harming anyone would be a fine goal. But the real questions is "What will the elected core team policy be?" :D

The rest of what you are saying is a geopolitical reference, in which case I don't have an opinion.

I think we can both agree that the tech is good. I guess we really only disagree on the current overall usage.
 
As long as it's opt-in and fully documented, no problem.
I'm not expecting they will allow supported hardware or programs in the base that do obscure data-exchange with online AI services. How is a firewall going to explain that?
 
rbranco said:
It's documented here: https://wiki.netbsd.org/zfs/

ZFS on root isn't supported in the installer yet, though, and I don't quite understand the wiki here:

Also interesting:

Re: ZFS status on NetBSD

mail-index.netbsd.org mail-index.netbsd.org
Click to expand...
Heh, even NetBSD is tipping its hat to FreeBSD, at least as far as ZFS goes...
😏

As for AI - yeah, it's possible to treat it as an advanced tool that can help, and it looks like the debate is really over what actually happens when AI usage gets out of hand - what THAT looks like, and what the consequences are. If you use a hammer for everything, eventually everything will appear to you as something that can be solved with a hammer - even if in reality, the hammer is not the best tool for the job at hand.
 
Espionage724 said:
in which case why should a project accept code with any hint of bad-intent?
Click to expand...
Like a headless browser getting scripted for a DDOS attack?

or the University of Minnesota debacle with the Linux Kernel?

or the xz/lzma code compromising the SSH server? See this Thread backdoor-in-upstream-xz-liblzma-leading-to-ssh-server-compromise.92922... If you read the conversation from beginning to end, there was some unfortunate politics involved among the human devs. Not the kind that break the Internet by unpublishing an npm package, but about someone's credibility. The unfortunate politics got a talented dev to just leave the project.

Any tool can be used with intent to do harm or good. And computer code is no exception.
 
MG said:
As long as it's opt-in and fully documented, no problem.
Click to expand...
I like the opt-in/opt-out idea. But to apply that at scale would take some time. Unless the FreeBSD community adopts a framework of some sort early.

Maybe the makings for innovation in the BSD community?
 
I don't think that the tech is a bad tech. I do think the license of the training data should be included in the output. It is a tool that is being used in an unethical way.

The current claw-code is a good example of how you can't have a license on something that is disregarded because it is visible. This is the legal argument from the owners of claude-code. I don't agree or disagree on their viewpoint.

You may have items in your home or business but the moment you take those items into the public where they are visible it does not change the status of the items into public property. The same is true for training data.

The training data status is going to continue to cause issues for the output of these programs.
 
rbranco said:
2. We genuinely risk that other countries develop more powerful models because they don't have the same moral itch as us. We must know how the models behave with all the data produced by mankind. It's a Faustian bargain that I completely understand.
Click to expand...

What?

Who are we and who gave them permission on the entire intellectual output of mankind?
 
What SHALL not be ignored.
Unless this kind if infringements are PERFECTLY, COMPLETELY, PEACEFULLY AND SATISFACTORY FOR COPYRIGTTS HOLDERS, AI/LLM shouldn't be used for generating something including codes, but for analyzing something (like finding bugs).

This is 100% (not lim100%!) clear fact, and doable for OSS/CCI matters by AI companies.
Just completely split out data sources for learning PER-LICENCING, NOT AT ALL MIXUP, clarify which licencing are user for the LLM.

Proprietary matters like sold books would be much harder.
AI company needs to charge "PER USER ACTIONS REQUIRING THE ORIGIN" and pay to copyright holders. How can it achieved, especially for locally executed cases? Maybe something like Private copying levy for CD-R?
 
Zare said:
What?

Who are we and who gave them permission on the entire intellectual output of mankind?
Click to expand...
Ideally it should be we, the people.

At least there's some consensus in that LLM output can't be copyrightable. That's why thorough human review is important. I myself don't believe in software patents, not so much because algorithms are math as Knuth said, but because they're a state-enforced monopoly.
 
rbranco said:
Ideally it should be we, the people.

At least there's some consensus in that LLM output can't be copyrightable. That's why thorough human review is important. I myself don't believe in software patents, not so much because algorithms are math as Knuth said, but because they're a state-enforced monopoly.
Click to expand...

There is no software patent problem here. It is all about data theft.

The algorithms and structures behind LLMs are free and open, it is the paid models and everything around them that is gaslit intentionally because the models are full of copyright breaches.

The AI companies are there to protect their illicit usage and practical ownership of this data, this is their strength, not some super duper soon-AGI neural network. Of course there is some consensus about it - because there is no push from other side, no AI vendor wants to own it. Anthropic doesn't go around trying to negotiate that output of Claude Code is theirs and therefore transferable to direct client licences.
 
Zare said:
There is no software patent problem here. It is all about data theft.

The algorithms and structures behind LLMs are free and open, it is the paid models and everything around them that is gaslit intentionally because the models are full of copyright breaches.

The AI companies are there to protect their illicit usage and practical ownership of this data, this is their strength, not some super duper soon-AGI neural network. Of course there is some consensus about it - because there is no push from other side, no AI vendor wants to own it. Anthropic doesn't go around trying to negotiate that output of Claude Code is theirs and therefore transferable to direct client licences.
Click to expand...
Every tech has its pros and cons. We can't highlight only the negative aspects. LLMs aren't special.

Lots of people with no programming background are happy that they can make stuff they couldn't do before. We must also see the positive aspects.
 
rbranco said:
We must also see the positive aspects.
Click to expand...
Like CEO admitting they are full of shit? WTF is wrong with people believing some stupid asshole liars? If the developers of FreeBSD need AI to program they should be axed.
PERIOD. This bubblegum bullshit is gotten so out of hand so bad it ruined the whole internet. Yet people are like DERP feed me some more of my own boogers.

fortune.com

Sam Altman and Dario Amodei are both walking back their AI jobs apocalypse prophecies as they eye blockbuster IPOs | Fortune

Some leaders like Goldman Sachs’s David Solomon and Box’s Aaron Levie have been saying all along that there won’t be a white-collar wipeout.
fortune.com fortune.com
 
My real concern is the level of accuracy. We expect five nines from our providers of utilities but what about AI.

Is its accuracy three nines? What level of precision is expected?

School children in Iran may have been bombed by AI targeting by my country. What level of accuracy would suit their parents?
 
Phishfry said:
We expect five nines from our providers of utilities
Click to expand...
Currently, AI looks like it delivers nine fives.

AI is not a person, it can not make a decision. It can only roll a dice with weighted outcomes. As this, it is absolutely NECESSARY that AI is checked by a human being, as it can not be resonsible for what it does. Whoever argues different is welcome to give a toddler a loaded gun. Bad idea? Sure. The toddler does not know what to do with it (and what not). No one can dodge resposibility by saying "it was the AI", that is right up with "only following orders" in my book. So whoever is pushing for AI, can he afford the cost on the scale that he is pushing it?
 
Dear Crivens, AI can be a nice tool. Given good ideas. But it us humans who need to check.
And there is the problem. Humans are lazy by nature.
And then there are persons receiving stupid AI answers by so called wise devs who did not the time to check. They are wasting there precious time.
So the problem is not AI. AI is a nice tool. But can we handle AI. Did we learned how to cope with AI at its worst & best.
 
Have spend many time with AI. It even has given me manipulating ideas, just like a politician would do. Even putting words i my mouth i never said.
So , cautious is not a bad attitude.
 
AI is just like a hammer. I can use this hammer at my own skull. But AI is new. We never learned to manage this hammer. And biologically this was never in our environment.
 
You must log in or register to reply here.
Back
Top