Ask HN: I don't want a single LLM I use to be involved in making war | Hacker News Reader

Ask HN: I don't want a single LLM I use to be involved in making war

2 pointsaledevv4 comments

In light of the recent disputes between Anthropic and the Pentagon, I'd like to know if anyone is aware of any initiatives (signature collections or otherwise) to ban or discourage the use of LLMs for war purposes.

This should be made clear by all researchers/founders/investors: "I don't want my AI used to facilitate war situations."

Is there a manifesto or similar on this matter?

View on HN

Do you not realize the existence of LLM's as all pretty much invalidated any practical recognition of there being a reciprocal acknowledgement of a creator's sovereign claim on how a thing is to be employed? That these models exist at all, is because everyone else's claims were not even asked about or inquired over; their work product was just taken and used. Now that the IP Santa machine is here, what in the name of all that is holy makes you think they're exempt from the same damn thing? Game Theory. Tit-for-Tat. It's over. You can at best argue we shouldn't be allowing governments to build their own infra to run such things, but compute is fungible. The cat is out, and if one didn't want it out, maybe there should have been more ethical outrage at building these damn things in the first place. Welcome to the wonderful world of Ethics, and what happens when you bite the forbidden Apple of Self-Referential Inconsistency.

In my opinion, the main issue isn't so much that data and content were 'stolen' to create AI.

It’s only right that the knowledge base remains in the public domain.

The real problem is that we must take a stand to establish proper guardrails for a technology that could become terribly disruptive and catastrophic.

If I were the creator of an LLM, I would forbid the use of 'my' technology for horrific purposes like warfare.

It’s fine to 'steal' knowledge to train models, but only if the goal is to build a better world.

>If I were the creator of an LLM, I would forbid the use of 'my' technology for horrific purposes like warfare.

Again. Cool. You'd use a license right? Or terms and conditions? Jussst like the ones in GPL, Apache, and most non-commercial licenses? Ya know, the ones that were blatantly ignored? Look, I'm very much of the Jeffersonian persuasion that information cannot be "stolen" per se. However, even when I've helped myself to the occasional apple off the tree of knowledge, I don't go around assuming I'm going to go and build a damn business around it without securing good faith terms of the creator first. I have historically honored an ethic whereby a creator does in fact have some claim on how they'd like something to be used. Golden rule/moral imperative, treat others as you would have them treat you, and I would like someone looking to build a business around something of mine to talk to me first, and at least do me the courtesy of not weaponizing things.

I'm not a consequentialist. I don't cut AI companies an ethical check because they think they're doing something for the public good; particularly when it is very clear they actually aren't.

>It’s fine to 'steal' knowledge to train models, but only if the goal is to build a better world.

Who gets to define better world? I might hypothetically think something of yours being weaponized is a better world. Therefore I'm justified taking it and free from recourse from you? I don't actually think that way mind, but do you see why that's easy to say, but terrible in practice?

The Department of Defense has more money than you and me together to pay for researchers/founders/investors.

Somewhat relevant xkcd https://xkcd.com/2128/

Reply on news.ycombinator.com