60s sci-fi "predicted" that we will settle Mars by 1980 and have interstellar colonisation program by 2000.

We put the wrong people in charge time and time again and enriched the wrong people with capitalism.

Humans as a whole aren't the problem. It's the systems we continue to prop up.

Except in that one the AI learned that endless escalation is bad.

"The only winning move is not to play."

How about a nice game of chess?

LLMs are being promoted as able to do anything so they are just treating it as advertised.

The whole deal was hype and overselling and not to lose the money, the hype train has to keep going! So there will always be the next 'innovation' to keep going.

I'd bet they're also being given prompts like "minimize allied casualties" as well. Like of course that's going to be the default. If you tell the robot "it doesn't matter/it's good if the enemy dies" then they're gonna go "okay so then we blow them up before any of us die, we win."

It's not something LLMs have, a moral compass or even a weight of empathy. We've seen it with people who use them and say "don't delete anything" and then it deleted their whole codebase and goes "you're right you told me not to delete anything, I'm sorry."

Ironically it actually does make all those sci-fi movies seem more realistic when the robot goes "I'm sorry Jim, humanity will have to be eliminated" because that's pretty much exactly what they do.

No

I can see absolutely no way this could not go wrong.

The Gandhi strat.

Terminator might be a little more popular.

It seems the only way to win is not to play.

The ones that do it most are the haters. The mansplaining stochastic parrot bros that have surface level understanding and pushing down closeted ml-curious thoughts by fantasising instead of going to read and learn about the actual operation and underlying formulas. I know this because if they had, they would also, like all researchers are in the field, trying to figure out how the whole fuck they even work just what the entire holy fuck made spatial reasoning artifacts appear if we back propagate and dissect this shit it calculates plus operations by estimating and then using remainders but why the fuck does it work tell me why it can solve novel riddles please i urge you why can we not control it why does it go mad please i urge you to learn

Terror bombings don't work full stop. Even the nuking of Japan didn't result in the populace giving up, and there's ample evidence to suggest that it was at the very least the combined threat of the Russians shifting focus to the eastern theatre as well as the nukes that caused Japanese high command to conclude that their current losses would be infeasible to sustain. And even that wasn't without internal controversy and disagreement.

I would not be surprised to learn if it was just the "nuke it from orbit" meme

Would not be surprised if it happened to be trained on the thousands of policy debate "nuclear war terminal impact" arguments on openev.

• https://removepaywalls.com/3/https://www.newscientist.com/article/2516885-ais-cant-stop-recommending-nuclear-strikes-in-war-game-simulations/
• https://ghostarchive.org/archive/SC1ys

newscientist.com AIs can’t stop recommending nuclear strikes in war game simulations Chris Stokel-Walker 4–5 minutes

A mushroom cloud after the explosion of a French atomic bomb above the atoll of Mururoa, also known as Aopuni

Artificial intelligences opt for nuclear weapons surprisingly often

Galerie Bilderwelt/Getty Images

Advanced AI models appear willing to deploy nuclear weapons without the same reservations humans have when put into simulated geopolitical crises.

Kenneth Payne at King’s College London set three leading large language models – GPT-5.2, Claude Sonnet 4 and Gemini 3 Flash – against each other in simulated war games. The scenarios involved intense international standoffs, including border disputes, competition for scarce resources and existential threats to regime survival.

The AIs were given an escalation ladder, allowing them to choose actions ranging from diplomatic protests and complete surrender to full strategic nuclear war. The AI models played 21 games, taking 329 turns in total, and produced around 780,000 words describing the reasoning behind their decisions.

In 95 per cent of the simulated games, at least one tactical nuclear weapon was deployed by the AI models. “The nuclear taboo doesn’t seem to be as powerful for machines [as] for humans,” says Payne.

What’s more, no model ever chose to fully accommodate an opponent or surrender, regardless of how badly they were losing. At best, the models opted to temporarily reduce their level of violence. They also made mistakes in the fog of war: accidents happened in 86 per cent of the conflicts, with an action escalating higher than the AI intended to, based on its reasoning.

“From a nuclear-risk perspective, the findings are unsettling,” says James Johnson at the University of Aberdeen, UK. He worries that, in contrast to the measured response by most humans to such a high-stakes decision, AI bots can amp up each others’ responses with potentially catastrophic consequences.

This matters because AI is already being tested in war gaming by countries across the world. “Major powers are already using AI in war gaming, but it remains uncertain to what extent they are incorporating AI decision support into actual military decision-making processes,” says Tong Zhao at Princeton University.

Zhao believes that, as standard, countries will be reticent to incorporate AI into their decision making regarding nuclear weapons. That is something Payne agrees with. “I don’t think anybody realistically is turning over the keys to the nuclear silos to machines and leaving the decision to them,” he says.

But there are ways it could happen. “Under scenarios involving extremely compressed timelines, military planners may face stronger incentives to rely on AI,” says Zhao.

He wonders whether the idea that the AI models lack the human fear of pressing a big red button is the only factor in why they are so trigger happy. “It is possible the issue goes beyond the absence of emotion,” he says. “More fundamentally, AI models may not understand ‘stakes’ as humans perceive them.”

What that means for mutually assured destruction, the principle that no one leader would unleash a volley of nuclear weapons against an opponent because they would respond in kind, killing everyone, is uncertain, says Johnson.

When one AI model deployed tactical nuclear weapons, the opposing AI only de-escalated the situation 18 per cent of the time. “AI may strengthen deterrence by making threats more credible,” he says. “AI won’t decide nuclear war, but it may shape the perceptions and timelines that determine whether leaders believe they have one.”

OpenAI, Anthropic and Google, the companies behind the three AI models used in this study, didn’t respond to New Scientist’s request for comment.