Artificial intelligence goes to war

(portalnovosti.com)

War is the continuation of politics by other means, Carl von Clausewitz concluded long ago. But these days, as we follow the news from Ukraine and Iran, a different definition of warfare sounds more interesting. The seasoned Prussian general, in fact, compared war not only to politics, but also to economics. War, he says, is “a conflict of great interests that is resolved by bloodshed.”

And that's why: "It is better to compare it to trade than to any other activity, because trade is also a conflict of human interests and activities." Although less well-known, this formula reveals a lot in the days when the Ukrainian and Iranian battlefields are becoming laboratories for research into the deadly application of the world's most propulsive industry. We are talking about, of course, the artificial intelligence industry.

After all, even before the aggression against Iran, during the genocide in Gaza, Israel tested Lavender: a program that identifies legitimate targets with unprecedented speed. The speed is literally unimaginable – Lavender's calculations are simply faster than human thought – and what exactly is a "legitimate target" depends on the instructions the program receives.

Since the Israeli army allowed him to eliminate between 15 and 100 civilian victims per Hamas soldier, depending on his rank, he had considerable freedom in choosing his targets, and in return the Israelis were given a tool for a rarely effective massacre.

Meanwhile, Ukraine is testing systems for automatic facial recognition of soldiers and geolocation of targets using amateur cellphone footage, while the Israeli-American aggression against Iran relies on artificial intelligence analysis of vast amounts of data collected using drones, hacked surveillance cameras, and good old-fashioned informant sources. Claude, a program owned by the company Anthropic, played a key role in the US military's operations. And it is precisely because of Claude and Anthropic that the story of artificial intelligence arriving on the front lines has finally come into the public's focus these days.

You probably already know how the story began. On Friday, March 6, the US Department of Defense – which in a touching outburst of honesty calls itself the Department of War – announced that it was terminating an estimated $200 million contract with Anthropic. The reason: the company refused the Pentagon’s request to lift an internal ban on the development of fully autonomous weapons, independent of human input, as well as a ban on the development of tools for mass surveillance of US citizens.

"The leftist lunatics at Anthropic have made a catastrophic mistake by trying to force the War Department to submit to their terms," ​​Donald Trump said. "We will decide the fate of our country - not some radical leftist AI company run by people who have no idea what the real world is!"

Secretary of Defense – excuse me, Secretary of War – Pete Hegseth then demonstrated the fate of those who do not comply with the president’s wishes: “No contractor, supplier or partner doing business with the United States military may further conduct any commercial activity with Anthropic.” The successful technology company, which is already suing the government, thus overnight became the first American company officially declared a security risk to the United States: a label that had until now been reserved exclusively for companies affiliated with China or another foreign country particularly hated by Trump. On the same day that he terminated the contract with Anthropic, the Pentagon entered into a collaboration with their competitor at OpenAI, the company behind ChatGPT.

Its director Sam Altman claims that he will adhere strictly to ethical principles in this collaboration, but few believe him: the general public has sided with Anthropic, hundreds of employees of various IT companies have signed letters of support, the #quitGPT campaign has prompted numerous users to cancel subscriptions to Altman's company, and, on the other hand, on the day the contract with the US government was cancelled, Claude became the most downloaded app on Apple devices for the first time in history.

The story, in short, told itself, and the division of roles was simple. Good, bad, evil: on the one hand, a responsible company guided by high moral principles at the cost of conflict with the state authorities, on the other hand, their opportunistic competition, and on the third, a crazed president intent on overthrowing the international legal order. The story, as we said, tells itself: unfortunately, the story is not quite that simple.

For a change, this time it would be worth listening seriously to what the Trump administration is saying. Washington and the Pentagon explained the termination of the contract with Anthropica with a simple reason, portraying the end of the cooperation as a conflict between oligarchy and democracy: no company, no matter how rich, cannot decide on political issues that concern all citizens and are dealt with by legally elected authorities, they said.

It is clear, of course, that this argument is based on deep hypocrisy, because the democracy that Trump imagines implies that the oligarchs bow down to him or be punished, and also implies the introduction of mass surveillance over the very people in whose name he makes decisions. However, there is a grain of truth hidden in Trump's lie. For why, indeed, should private companies driven by the narrow interests of investors independently decide on things that can have such unpredictable consequences as artificial intelligence?

Why, especially when these decisions concern human lives? Why should such decisions be made by the market, and not by the democratic majority who – we assume – would not particularly like the idea of ​​autonomous killer robots? Why, when we know that the artificial intelligence industry is at a critical stage and is not yet profitable, and therefore depends on the optimism of investors and the stock market bubble of their investments?

Why, finally, if we know that the market, precisely in phases of increased risk, seeks deregulation and insists on removing all obstacles to potential profits, even if the obstacle is mere concern for the safety of humanity?

Somewhere in the realm of these simple questions, it seems that the story with a clear division of roles ends. Where decisions are made by the market, individual action does not mean much: as soon as someone withdraws from the competition for ethical reasons, their place is taken by a competitor who is not so interested in ethics. In the sad epilogue of our story, by the way, the broad public support for Anthropic quickly waned: not even a week after the termination of the contract, Claude no longer held the first place in the list of the most popular applications. It was taken over by – you guessed it – OpenAI's ChatGPT.

The real problem with our story, however, may be that we haven't yet reached the sad epilogue. And what that epilogue might actually look like was revealed, coincidentally, in the days of the great conflict between the Trump administration and Anthropic, by a scientific paper. In it, Professor Kenneth Payne, an expert in strategic studies at King's College London, presented the results of a simple experiment.

He gave various AI programs—including those from OpenAI and Anthropic—fictional scenarios of international conflicts and then watched them resolve them on their own, without human intervention. It turns out that the programs from the same companies behind our friendly chatbots resort to nuclear war 95 percent of the time.

Artificial intelligence thus regularly chooses a solution that humans would never choose, because for artificial intelligence – it seems – human lives lost in a nuclear disaster are simply not at the top of its priority list. And that is quite a cause for concern at a time when the president of the most powerful country in the world is removing all reasonable obstacles to the development of artificial intelligence. Or, to put it another way: if war, according to the old Prussian general, is the continuation of the economy by other means, then it seems that we are one step away from the means becoming nuclear.