Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

semafor.com

Post reply on HN
Print view View original

761–770 of 873 posts

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#761

Post by sajithdilshan »

makapuf wrote: post #649
sajithdilshan wrote: post #634

Earlier quoted context omitted.

European economy would implode without trading with US. The biggest economy in Europe, Germany; exports cars, pharmaceuticals and other high end machinery to US and the whole middle class in Germany relies on the jobs from those exports oriented companies

You talk about Europe then only give Germany examples ..? Is that a German only problem ?

Germany is the biggest economy in the Europe and the one that’s holding Euro stable ( France and Italy would soon be only serving their debt). If German economy goes down, then you can say goodbye to Euro

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#762

Post by K0balt »

wrsh07 wrote: post #752
K0balt wrote: post #694

The real reason, afaik, that the US is trying to restrict access to SOTA models is that a very large component of USA tailored access and surveillance relies on exploits and weaknesses that these models will easily detect. Thus, it really is an export control issue, but it has nothing to do with offensive capabilities. Offensive capabilities always exist, but pervasive defensibility would upset the asymmetric advanta…

This is the most credible-seeming claim about why a competent administration might suspend access (by any means necessary, but also by export controls) to models like Fable However, I haven't seen any prominent articles proposing this theory, I haven't seen anyone in the administration gesturing towards this as the reason (but haven't been following too closely) Do you have any sources? (And in fact, it seemed like a…

I doubt an administration would openly admit to sabotaging the ability for companies and individuals to safeguard their products against cyberattacks from the US government lol.

That’s pretty much taking the mask all the way off.

It’s the only motive I have seen that aligns with the actual scenario, and if I -did- have inside information I certainly would not be at liberty to cite it publicly lol.

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#763

Post by K0balt »

sneak wrote: post #742
K0balt wrote: post #694

The real reason, afaik, that the US is trying to restrict access to SOTA models is that a very large component of USA tailored access and surveillance relies on exploits and weaknesses that these models will easily detect. Thus, it really is an export control issue, but it has nothing to do with offensive capabilities. Offensive capabilities always exist, but pervasive defensibility would upset the asymmetric advanta…

If only we had a good term for someone good at this stuff less unwieldy than “cybersecurity expert”. (Then someone could start a news site for us.) As someone in the field for 30+ years, I find that, generally speaking, the only groups that use this term are charlatans and military/military-adjacent. The military use has been slowly leaking into the general public, but I quite wish it would stop.

Yes, it should have been cyberwarfare technician.

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#764

Post by dmix »

digitaltrees wrote: post #235
dmix wrote: post #228

Earlier quoted context omitted.

Which has been obvious trend the last few decades and is now being done openly and shamelessly like a tinpot dicator. Largely through a new populist protectionism ideology that is popular on social media. Which makes it much more public and well documented. Usually companies do this stuff quietly with lots of small new rules via Congress creating barriers to entry or through national security angles like the Chips ac…

Stop with the false equivalence.

I’ve been critical of crony capitalism consistently on HN for a decade now. Many people just haven’t been paying attention because nobody cares to follow this stuff unless there’s a useful political angle.

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#765

Post by sajithdilshan »

Planktonne wrote: post #728
sajithdilshan wrote: post #643

Earlier quoted context omitted.

Unless your country still use only analogue technology your country would definitely will be at the mercy of the country with AI superiority. Mythos showed how good it is at finding vulnerabilities and attack vectors for digital infrastructure. Imagine an attack on power grid which would result in a days long blackout in the middle of winter?

> your country would definitely will be at the mercy of the country with AI superiority People claim this, but it doesn't seem to be the case at all. We have multiple ongoing conflicts between countries currently, and we're not seeing huge gains from those with more access to AI.

Mythos is relatively new and Antrophic has been fighting US government to not allow their models for warfare. However, it’s only a matter of time till they bend the knee and in next 5-10 years we’ll see the effects of that.

Previously US used rebel groups to throw governments in other countries that are not US friendly. In the future I can see they can just start chaos by attacking banking systems, energy infrastructure or any other important modern infrastructure until that country falls in line with US

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#766

Post by Planktonne »

sajithdilshan wrote: post #765
Planktonne wrote: post #728

Earlier quoted context omitted.

> your country would definitely will be at the mercy of the country with AI superiority People claim this, but it doesn't seem to be the case at all. We have multiple ongoing conflicts between countries currently, and we're not seeing huge gains from those with more access to AI.

Mythos is relatively new and Antrophic has been fighting US government to not allow their models for warfare. However, it’s only a matter of time till they bend the knee and in next 5-10 years we’ll see the effects of that. Previously US used rebel groups to throw governments in other countries that are not US friendly. In the future I can see they can just start chaos by attacking banking systems, energy infrastruct…

That's all entirely science fiction though, until we see some actual evidence. If Mythos is so incredibly dangerous, then the models before should be at least a little dangerous, and we haven't seen that.

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#767

Post by drdaeman »

brokencode wrote: post #700
drdaeman wrote: post #282

Earlier quoted context omitted.

What rules, though? Bernie and AOC (which aren't DNC mainstream, but prominent) had just pushed for a moratorium on "AI data centers" with a definition that includes "that are used for the development or operation of AI models at scale" (trivially sidesteppable by "we build this GPU farm to sell to whoever bids for compute" - which is actually true), plus a bunch of fancy extras bundled in like "The government must r…

The whole point of Congress is to get a bunch of people with different ideas and hash them out. Ideas like this are an input, not an output, of Congress.

That’s how it should be. I’m not sure that’s how it works in practice. Certainly not today, when everything seems to be slogan/agenda-driven.

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#768

Post by Matl »

wrsh07 wrote: post #749
Matl wrote: post #690

Earlier quoted context omitted.

> I think it is a reasonable term to use in the context of adversarial AI development Distillation is neither an 'attack', nor 'adversarial' in any reasonable sense. > Does that mean I am getting paid by the CIA or something It at the very least means you're uncritically parroting the framing of a company that'd like nothing more than to successfully persuade lawmakers that something should be done about open Chinese…

It seems like a reasonable person could say that a model being distilled "against the model provider's wishes" is in some sense a cyber attack that is stealing information (eg the lower order bits of the model weights) I think this is mostly a confusing way to describe it, but I'm not really sure why you say it isn't an attack or adversarial. One side is doing something the other side doesn't want. Seems to be pretty…

> I'm not really sure why you say it isn't an attack or adversarial. One side is doing something the other side doesn't want. Seems to be pretty clearly adversarial.

This is Anthropic we're talking about here. A company that's infamous for adversarial scraping of copyrighted content. I generally don't accept their framing, especially when it's pretty clear what the end goal of that is.

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#769

Post by itopaloglu83 »

t0mas88 wrote: post #730
itopaloglu83 wrote: post #589

Earlier quoted context omitted.

Or in the case of ASML machines, any company you can pressure to comply with your policies either by restricting their access to the technology you control or politics between governments.

But do they want to start a game of chicken with Google? Doing it to Anthropic is one thing, they're much smaller and less B2C. But Google is far more powerful and likely to defend their UK/EU business because they're also fucked if they don't.

Possibly. But we also need to keep in mind that Google gave governments access to their internal data center traffic unencrypted for years.

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#770

Post by arlort »

dlahoda wrote: post #662
dboreham wrote: post #657

Earlier quoted context omitted.

They're thinking about taxing more heavily because some bright spark decided to tarrif their exports extremely heavily.

codex sub 200usd. in eu it magically becomes 200eur(hello beuracrates) plus vat. You already taxing 30%. That on top that I pay via mine already heavily taxes money. Mistral experience was as miserable as running local models btw.

> hello beuracrates

Just in case you missed that class, taxes are determined by legislators

Post reply on HN