[bigyabai]

Okay, I disagree with most of Ed's takes but this is also a pretty wild refutation.

National security is a canard for covering up national investment. The US doesn't need to have it's hand in the pockets of private businesses, but by making a murky NATSEC claim they can manufacture interest for investments and under-the-table cooperation. This is exactly how we saw companies like Google, Apple and Microsoft get roped into PRISM, in exchange for alignment on political objectives and upholding the security theater of obviously-backdoored shitware like BitLocker. Departments like the NRO have had their own in-house AIs for decades, they're not beholden to any of these firms for access to cutting-edge inference. SENTIENT predates the conception of OpenAI as an organization.

If you disagree; look at China. Why would China Open Source their frontier models if there was a genuine national security threat in doing so? Because there is no threat, it's a tacit refutation of the United States' fearmongering and a direct attack on the credibility (and profitability) of the America's frontier labs like OpenAI and Anthropic. While Anthropic and OpenAI send representatives to grovel in the Oval Office to release a model, China greenlights anything, just like a real neoliberal free market would. The US' federal interventionism is causing China's AI to recoup all of the ground that America loses in credibility.

Fact is, China's hosting is no less trustworthy than America's hosting. OpenAI and Anthropic are certainly just as backdoored and data-hungry as China's hosted services are, I'd be shocked if both weren't populating a vector database of my characteristics for "marketing" purposes or somesuch. But I still trust China more as an American citizen, because their ability to prosecute, blackmail or manipulate me is much weaker than my domestic government. There's no fear when I send a request to Z.ai or Qwen because neither of them are begging for federal interventionism on a weekly basis. They have nothing to ask my government for in exchange for my data.

[CMay]

SENTIENT wasn't doing the same kind of inference that AI companies have now, so suggesting they have nothing new to offer is incorrect. Advanced AI is clearly dual-use technology that can be used for adversarial or defensive purposes in ways most people simply aren't used to thinking about.

Your lack of imagination is not proof of reality. The US releases open source models too. The primary concern is that there should be some limitations on the capability of models released publicly. When companies are hosting their models via API, they can have many other layers involved in serving those requests that can reduce abuse. When you release an open model, there is no layer and many protections built into the model itself can often be worked around.

China puts more restrictions on models domestically than they do internationally. They did the same thing with Tiktok, where their domestic version was more cleaned up, curated and censored. There, you can't download a model without basically sharing your identity. Huggingface is banned. It would also be illegal there to host the open weight models on a 3rd party inference provider. Kimi? Qwen? If it's not small enough that you can run it locally then you have to use it via API, which are all highly censored. Even the smaller models are censored, it's not only bigger models. The experience people can have outside of China with those models is not the same, but the point is that this directly refutes your suggestion that China doesn't think they are a threat.

China would love to maximize the abuse of open models in the west, because there are theories that it leads to more extreme authoritarian surveillance tendencies which benefits their Leninist world view. There is also the perspective that it can hurt investment into AI if people think it's inevitable that equivalent models will be open.

> China's hosting is no less trustworthy than America's hosting.

That's unlikely: https://nationalsecurity.virginia.edu/sites/default/files/fi...

[bigyabai]

> Advanced AI is clearly dual-use technology

It looks like a liability, if the US's assault on Iran is any basis to go off-of. All the "frontier" AI in the world isn't helping the US win a brushfire war, it's a bit embarrassing I'd say. What's the contribution of frontier LLMs here, you think? What's the dangerous dual-use case, versus traditional computer vision and multimodal data aggregation systems? I'm not really seeing it, SENTIENT is far more impressive.

> When companies are hosting their models via API, they can have many other layers involved in serving those requests that can reduce abuse

That's not a safety feature. API providers do that to reduce their risk exposure and limit the liability they could face if someone attempted to abuse their service. Without the legal liability incentive, service providers would gladly remove those limits. As you've noted, Open Weight models don't have this feature at all, because the user is subsuming the liability; there's zero reason to drop or refuse requests for cybersecurity, "bioweapon research" or whatever stupid boogeyman topic you want to use to justify censorship. Of course, none of these LLMs have posed a legitimately unique threat to anyone in the past 6 years of "dual-use" LLMs being available - but a lack of evidence sure isn't going to stop your imagination.

> China puts more restrictions on models domestically than they do internationally.

> but the point is that this directly refutes your suggestion that China doesn't think they are a threat.

You're just describing a typical surveillance architecture. The United States does the exact same thing for data retention purposes, except they extend their surveillance globally instead of just imposing it just domestically. With open-weight LLMs, you are free to avoid that type of federal intervention and run the model or finetune it on your own hardware.

> That's unlikely

Nothing in that packet is very scary, but I will admit that it's hilarious how poorly a document like that ages in as few as 12 months:

  China’s artificial intelligence program is centrally planned and ideologically engineered through layers of national policy, and implemented by a coordinated collaboration among government, SOEs, private tech companies, academia, and the military. 

As opposed to the US' AI policy, which is centrally planned and ideologically engineered through layers of national policy, and implimented by a coordinated collaboration among government, private tech companies and the military.

Can you explain actual dangers, or are we going to remain at this impasse? I'm not super worried, yet.

[CMay]

> It looks like a liability, if the US's assault on Iran is any basis to go off-of. All the "frontier" AI in the world isn't helping the US win a brushfire war, it's a bit embarrassing I'd say.

To put it into perspective, we didn't start the war over the Strait of Hormuz and it's correct to say that the US does not depend on the Strait nearly as much as other countries do. We aren't still there because _we_ depend on it so much as we're there to defend the usage of it by allies and defend the applicable international laws.

The US could largely walk away and be fine, but there are multiple benefits to staying around. One benefit is that it, combined with the strategic attacks against Russian oil infrastructure, puts indirect pressure and uncertainty on China's oil supply. China has been drawing from its huge oil reserves which it was building up not only to weather short term disruptions, but to sanction proof itself. That increases the threat against Taiwan, obviously. So encouraging them to consume their reserves is of strategic interest to us.

A higher price of oil also benefits us and also investment in Venezuela simultaneously, since the US is the largest oil producer in the world.

> That's not a safety feature. API providers do that to reduce their risk exposure and limit the liability they could face if someone attempted to abuse their service. Without the legal liability incentive, service providers would gladly remove those limits. As you've noted, Open Weight models don't have this feature at all, because the user is subsuming the liability;

False. Blatantly false. These companies pre-emptively implemented security and safety features for all kinds of scenarios, even subtle ones they aren't likely to be legally liable for. In addition, these open models that you say they don't care about as proof it's not about safety, they invest money in safety training the open weight models they release. Chinese AI companies aren't oblivious to this either and also do some safety training, but open weight models cannot be made completely safe and that's largely the argument. The more capable a model is, the more you rely on extra layers which offline AI usage typically runs without.

Moral aspects are regularly involved in hiring quality talent. If your company is obviously provably excessively immoral, disregarding human life or any sort of compassionate decency, you lose a lot of the best talent. Companies regularly attempt to align themselves with morals not only because it is the right thing (though that is often the primary reason), but the secondary hiring benefits further reinforce that. Companies are made of people and people do not suddenly become unfeeling as a result of being associated with a company. Real life does not map to Marxist caricature so easily.

> You're just describing a typical surveillance architecture. The United States does the exact same thing for data retention purposes

It's actually not typical and China is the only country that does this. It is law. You can even host models in Singapore or the EU, but not in China. It's illegal.

> As opposed to the US' AI policy, which is centrally planned and ideologically engineered through layers of national policy, and implimented by a coordinated collaboration among government, private tech companies and the military.

So, let me present a critical difference between the US and China. China does not have a military. The Chinese Communist Party has a military. Imagine if in the US, the republican and democrat parties had their own militaries and none of them existed to defend _you_, they existed to defend the party under the guise of defending the country, which they extend to suggest that overall defends you while excluding the fact that they will gladly kill you if you don't believe in their political party.

This political party then says, no matter what you believe politically, your company is required to implement our political ideology and place our political agents inside your company to maximize vision implementation and civil-military fusion. The US partners with private industry and there exists select strategic operation with the military, but it's nothing like the CCP does.

Perhaps they wouldn't care much about what you chat about with your AI on their platforms, because you seem to already be very sympathetic to the CCP. That is not a very useful perspective from which to convince other people that it is good and safe to use.

[bigyabai]

Pick a lane; can America walk away, or are we still leashed to the Levant?

The biggest flaw with this logic is that China can recover from an oil blockade much faster than America can regenerate interceptors and munitions. The longer the US enforces the blockade, the lower the magazine depth is; that gives China the advantage especially in a protracted blockade scenario. Their strategic reserve has been worth its weight in gold, and the US can't punish China with a blockade indefinitely.

Another aspect is the what the rest of the Middle East is feeling. They know that the US can't continue like this forever, so they'll have to appease Iran at some point to return the status quo. US military installations, American-run datacenters and US-administered airports all became targets instantly. Civil infrastructure like water treatment, power generation and bridges are still underneath the Sword of Damocles. If they assent to a wider war, both sides will bleed heavily, and neither will have much to gain in the end besides a destabilized region.

> False.

> they invest money in safety training the open weight models they release.

Ah, next you'll tell me that every model could be AGI if the owners didn't nerf it with safety finetunes.

Thankfully, we have abliterated and finetune-free base models to cross-reference, and these open models can remove refusals entirely. This goes for models like GPT-OSS as well, which can be made to do all sorts of horrible liability-inducing generations if you prompt it to. It's entirely possible to remove the majority of these refusals in even the most advanced reasoning models, you can't really argue that they're being adequately suppressed.

> Companies are made of people and people do not suddenly become unfeeling as a result of being associated with a company.

Correct, it is money that extricates people from moral responsibility, not proximity to a business. How do you think we got Jeffrey Epstein anyways, a financier got dropped off by a stork?

> It's actually not typical and China is the only country that does this.

They're the only ones open about it. I know you're a jingoist lackey, but even you can't refute NOBUS capabilities like Room 641A et. al. Those same crooks are breaking into US-owned servers and doing the exact same thing in 2026. You don't even have the decency to deny it.

> Imagine if in the US, the republican and democrat parties had their own militaries and none of them existed to defend _you_

That's basically what the purpose of the US Army is, post Civil War. Conservatism lost and submit to the auspices of federal administration. The republican identity of American politics was reformed into a democratic autocracy where each partisan leader is the Commander-in-Chief of the entire armed forces. The secular difference between both philosophies is moot when, for example, Crimea is invaded or Israel is attacked. All logic goes out the door to fulfill blithe political objectives.

> That is not a very useful perspective from which to convince other people that it is good and safe to use.

Sure it is. Trust nobody, use nobody's server. Buy your own hardware and administrate your own data. Anthropic and OpenAI will consume your data and sell it back to you; they are not your benefactor. Support open models and resist federal interventionism, exercise your constitutional right to free speech even when it pisses off Dario or Sam.

[csomar]

> The US could largely walk away and be fine

The US can't walk away, otherwise, Trump would have done so. That part of the world is part of the US system and "walking away" is basically giving it to anybody who is willing to exert control over it. They'll certainly be opposed to the US system (sanctions will be reverse imposed at that time).

And then what's next? Walk out of the pacific?

[CMay]

Some of your logic isn't exactly wrong, though just because someone can walk away doesn't mean they have to or want to. I simply mean that for the US alone from an isolationist stance, a lot of the primary and secondary goals are already achieved. If we walked away it would have already largely been worth it. Yes, Trump repeats something akin to this, but just because he said something doesn't automatically make it incorrect.

We didn't attack them _because_ of the Strait of Hormuz, not exclusively. We knew they could harass the strait, since we've been thinking about this for decades. The other person's argument was some kind of attempt at shaming AI and the US, with the perspective that with all the AI in the world we couldn't win a war with Iran. It's just politics, where a country losing as badly as Iran has to redefine what winning means to save face. Russia has to do the same with Ukraine. Inside their countries they largely have information control, but they still have to apply some strategy for external propaganda.

And honestly, to some extent when an opponent feels that they have some kind of leverage, that can be useful as a sort of offramp in negotiations.

Clearly there is an actual shift occurring in terms of water control, because Russia lost so many ships from drones alone. Most ships are not equipped to deal with these new threats the way the US carrier groups are and when you're talking about container ships or oil tankers it can be a big deal.

Iran relies on shipping too, though. Unfortunately, some countries don't mind hurting themselves. We see this with China and Russia a lot, where they are willing to sacrifice people, quality of life, trade, basics, relationships and more out of some kind of stubbornness or to immunize their country from a ruleset. Like if you think you can use this as leverage, we'll show you! Iran is being self-destructive in a similar way.

One comparison is what we see China doing harassing the Philippines in their own territorial waters and establishing fake islands to claim control over an expanded area of water which upsets many of their other neighbors.

[bigyabai]

> a lot of the primary and secondary goals are already achieved.

Which ones? FONAV is not restored to the Strait, attacks on shipping have not been eliminated, and Iran still possesses credible nuclear materiel according to recent Israeli and CIA reports. I'd like to hear you list explicitly what has been achieved from the United States' original goals.

> Iran has to redefine what winning means to save face. Russia has to do the same with Ukraine

The Russian Federation has a GDP that is ~10x larger than Iran, so you also have to redefine what "losing" means to make that analogy. America has a GDP that is ~100x larger than Iran, even more prestigious and strong, and it still cannot restore the status quo ante bellum. In terms of asymmetrical warfare, the US is much closer to Russia's strategy, and Iran much closer to Ukraine's tactics.

> and establishing fake islands to claim control over an expanded area of water which upsets many of their other neighbors.

It is awful. America has tried to escalate the issue in the UN Security Council, but their support for the military annexation of Syrian and Lebanese territories has cost them too much credibility to force the issue. Everyone knows that the US government doesn't care, including China. Such a shame that this double-standard costs so many people their livelihood and home.