Tag Archives: defense policy

Wrong Answer

Posted on by
Reply

The subheadline contains the wrong answer, for those astute enough to see it, especially those more so than this op-ed‘s opinion writer, Judd Rosenblatt, who is the CEO of AE Studio and president of the AI Alignment Foundation:

The key is “alignment,” which improves capability while cordoning off dangerous knowledge.

Rosenblatt expanded on that not too much later in his piece:

…build the world’s most capable open-weight model in America, with the dangerous capabilities stripped out before release

Because nothing says limiting innovation and truncating freedom like censoring knowledge and capability. Never mind that that will limit and truncate only American innovation and knowledge; it’ll have no effect on our enemies.

He went on:

Using GRAM [Gradient-Routed Auxiliary Modules] to train language models, we found that knowledge about bioweapons, offensive cyber, and nuclear capabilities, and even antisemitism, could be routed into compartments, then removed. The released model behaved as if it never learned that information.
…
What matters is a safety standard that causes a dangerous capability to stay absent under adversarial testing.

The bioweapons, offensive cyber, and nuclear capabilities bits are seemingly obvious targets for censorship, but that knowledge is too easily compiled via a variety of sources, including other AI models that aren’t restricted. The real kicker, though, is that last: who gets to define what is a dangerous capability? And who sets the threshold between random data to information to knowledge? Will these Know Better definers decide that identifying or not defining hate speech (whatever that is) is a dangerous capability and the thresholds for that? Providing arguments for, or against, human involvement in global warming? Men in girls’ sports and bath- and locker rooms? Defining “woman?” Criticizing government? This list is endless.

The blatant…foolishness…of Rosenblatt’s idea makes me wonder about trolls and moles.

Questions

Posted on by
Reply

The subheadline on an article about Ukrainian attacks on retail warehouses in Russia raises a couple for me.

Infernos burned through massive warehouses operated by Wildberries, causing billions of dollars in damages

Some of those fires burned for a couple of days before they were brought under control.

Thus: are the Ukrainians using fire accelerants in their attacks on the warehouses (good for them if they are; they should be adding them to their attacks on the barbarian’s fuel and refinery facilities, too), or does Wildberries really have such poor fire fighting systems and such poor flammables storage protocols?

Quitters?

Posted on by
Reply

American AI developers are all up in arms over the successes of People’s Republic of China AI models, especially those that are being released as open source. The American executives’ beef centers on how good those models are and how cheaply produced they are.

OpenAI and Anthropic executives are sounding the alarm about the rise of cheap AI, particularly powerful new models produced in China, suggesting they will lead to a “dystopian” AI future and present unacceptable security risks without regulation.

And

The emergence of highly capable, open autonomous AI systems—including Moonshot AI‘s Kimi K3 model and Alibaba’s Qwen 3.8 Max, which were released in recent days and viewed favorably by investors and users—has turned the AI race on its head once again. Kimi K3 also was competitive with US models on some benchmarks.
…
“One probable outcome of an open-weight-model-dominant world is full AI communism, which is precisely what China proposes: rather than a market product, AI is a ‘public good’ which will ultimately be provided by the state as a kind of ‘digital public infrastructure,'” Dean Ball, OpenAI’s head of strategic futures, said in an X post Friday.

The general thrust of these bleats and cries to Mommy Government is that these personages think that the PRC’s scientists and programmers are smarter than American scientists and programmers, and they want to cower inside a truly dystopian regulatory moat that would stifle American innovation in AI altogether.

That’s as unacceptable as it is despicable. AI development—American AI development, especially—enormously benefits from open competition. Competition is what drives at an accelerating rate the quality of any product class, and especially AI. Especially AI, too, is a competition we must win—where winning isn’t across the finish line first, since with all product development, there is no finish line. Winning here means maintaining a significant (not just by a nose) lead.

Thus, American AI development must proceed apace, without regulatory restriction. We have enough laws on the books to handle its misuse, just as we have for any tool. American execs need to stop bleating and get competitive, on quality, on price, on rate of improvement.

The sole exception to largely unfettered competition should be AI models from companies domiciled in, or associated with companies domiciled in, the People’s Republic of China.

That exception, though, has nothing to do with those models’ costs or capabilities or open source availability or nearly anything else. This exception is driven solely by the national and corporate security threat those companies and their models represent. Those companies, via the PRC’s 2017 intelligence law, are entirely beholden to the PRC intelligence community for conducting espionage on any subject, in any venue, and against any potential source that the intelligence arm of the PRC government wishes. These companies should be barred from the US, and other nations wishing to remain free from PRC…influence…should do the same.

And this central truth: American scientists and programmers are every bit as good as, if not better than, PRC scientists and programmers. As are the scientists and programmers of other nations wishing to remain free.

Iran, the Strait, and Nuclear Weapons

Posted on by
Reply

In Michael Gordon’s piece on The Battle for Hormuz, he wrote that

A stable peace appears elusive, some former officials say, as each side is calculating that it can win the long game in a test of wills….

A stable peace with terrorists is an oxymoron, and it’s time to put down that pipe dream. Green then closed his piece with this:

Former military officials say the US can weaken the regime if it is persistent but warn there is a long road ahead.
“The US will need to identify and pursue a longer term strategy that addresses Iran’s focus on controlling the Strait of Hormuz, threatening its neighbors with missiles and drones, and the nuclear weapons program,” said Joseph Votel, the retired Army general who led Central Command from 2016 to 2019. “This is going to be a long-term effort.”

The leaders of the Iranian government, the civilians in Tehran and the leaders of the Islamic Revolutionary Guard Corps, will never agree to limits on their nuclear weapons program or to giving up their aspiration to control the Strait of Hormuz. That doesn’t mean we should quit the struggle.

In fact, we can shorten it well within Votel’s implied time frame with a change in emphasis and a firmer drive to achieve the fight’s primary aim, the end of Iran’s nuclear weapons acquisition capability. Rather than seeking to “influence” or “coerce” those leadership teams, or “weaken” them into agreeing to a deal, it’s time to accelerate and broaden—escalate, if you will—the attacks on Iran, focusing on three target sets in all their breadth and (occasionally literal) depth: destroy Iran’s ability to fight at all by destroying not just some weapon sites and caches, but the means of producing replacements; destroying Iran’s nuclear facilities wherever they are along with their caches of purified uranium, including deepening the burial of those caches already hit; and killing Iran’s leaders, especially those running the IRGC. The latter has dispersed its command and control facilities and personnel, so that will be especially…tedious. Israel, though, has a demonstrated facility at this sort of activity; it would be good if we could enlist that nation’s aid here.

Only by completely disarming Iran and killing the terrorists reigning over it can there be a lasting, stable peace and one that includes a restoration of the Strait, and of the Persian Gulf and Gulf of Oman, to their status as international waters freely usable by commercial shipping.

Hitting Back

Posted on by
Reply

After 15 years of being hacked, often severely, by Russian government sanctioned, if not government owned operators, the nations of Europe has finally gotten around to responding. With sanctions.

The European Union and the UK said the sanctions would target Russian military intelligence officers, hackers, and private companies that support the Kremlin’s cyberattacks.

As if the sanctions’ targets would care. Those folks, their companies, and their backers will just find other ways to operate. Money isn’t all that much of a critical item for their operations, and the resources they really need are largely untouched by sanctions.

We do get this sort of thing, though, in addition to those sanctions: after years of Russian hacks against the French Ministers of Defense and Foreign Affairs and their networks,

French Foreign Minister Jean-Noël Barrot said he would summon Russia’s ambassador in France in the coming days.

For what? Le goutier sans vin? No rush in any event; whenever is fine.

How about, instead, you know—work with me on this; it’s a new concept for some—maybe fighting back and hitting directly Russia’s hack networks, and the computer networks of Russia’s intelligence agencies and defense establishment?

Sanctions alone accomplish very little, as years of them already applied against Iran, northern Korea, the People’s Republic of China—and Russia—clearly demonstrate. Those nations and street gang-run areas have not changed their behavior in the slightest, for all that their pace has been reduced.

What’s necessary is direct responses targeting their ability to operate at all. Get into their networks and erase data in some, insert false data in others, and inject sleeperware in all of them, the last to be triggered at a useful time in the future.