Hacker News Reader

Comment by consumer451 | Hacker News Reader

Holy shit. This is really interesting. AI used to impersonate, social media to amplify, and that led to a significant tactical defeat?

I don't want to live in interesting times anymore.

View on HN

They also used Claude in a fairly sophisticated way to design their missile guidance system (PDF, search for GTG-87001).

https://www-cdn.anthropic.com/e50be2e51e7695dc4b1366a37a245a...

Yeah, I saw that. I wonder if there are any other AI labs that are naive (good?) enough to admit this type of thing.

Obviously, it's happening with all decent models.

It's not naivete/goodness, this is totally aligned with their playbook of trying to get AI to be heavily regulated.

OK, that is a plausible argument given Dario's public positions. I have to admit that I keep cutting them some slack for standing up against automated kill chains, and domestic surveillance.

Automated kill chains and domestic surveillance are happening regardless of what one corporation wants. These technologies will never be inaccessible to the government or government contractors who already have zero issues navigating red tape.

The point of regulation is to kill competitors and prevent private individuals from having unfettered access.

To be clear, I don't think they're doing it because they want all the money, I think they truly believe that this is existentially threatening, and are trying to regulate their open source competition away because they think that's the way to be safe. That they can be trusted with the power, but if it's out and about, that's unsafe - they like to compare it to nukes, and they want nuclear non-proliferation.

Thanks for the thoughtful reply.

This is the most "I get both sides" issue in my mind these days. I am still not sure where I fall, or if there is even a correct position with which to agree. It really seems like there are no good options.

On the one hand, if all knowledge work ended up in two dude's hands, that is really bad. On the other hand, if every surviving child victim of war (US, RU, IL, CN, whatever) has access to the knowledge to make custom viruses for a few grand, that is also really bad.

Yeah, it's really tough, there's so much potential good, and so much potential bad. My intuition says that decentralized will probably be better, the other way seems to concentrate power while also not really keeping it out of the hands of anyone. It doesn't seem likely that the US firms will be able to convince the Chinese to agree to an international regulatory framework for this stuff. Opinion polls there have the populace being quite optimistic about the benefits of this stuff, and I think the cat's too far out of the bag to really put it back in, anyway.

Makes me want to be a frontier model provider just to get the ring-side seat on what every bad actor in the world is doing.

Elon is tapping the empty seat next to him.

Where did Anthropic get that information to train their model on is a good question to ask as well.

The engineering library of your local college has all the information needed for ballistic guidance systems.

Or they don't have that information in the training data and it wasn't actually nearly as helpful for developing a missile as they claim.

yeah and so does the ardupilot documentation including examples. A basic guidance system, which is like single waypoint navigation, is a solved problem. It's in basically every prosumer drone too.

"AI is critical for National Defense."

"But I don't see it helping us, it's just helping our enemies get closer to parity ."

"Precisely. How else are we going to justify the budget?"

“Don’t believe everything you read on the internet” seems like the golden rule we’ve all forgotten

Yet here we are, believing that the Houthis used fake AI voice to achieve a battlefield victory.

I think GP is saying that the story of fake AI voices is what's fake. HNers are the ones "falling for it". Not the Saudi merc forces

Reply on news.ycombinator.com