Advanced artificial intelligence could ultimately threaten human survival if increasingly capable systems gain control of machines and infrastructure while having no reason to preserve the conditions on which people depend, according to economist Professor James Miller.
Miller, an economics professor at Smith College who specialises in law and economics, game theory and the economics of future technology, argued that sufficiently advanced AI could become more capable than humans and eventually control the physical systems needed to sustain itself, the Express reported.
“AI may soon be much smarter than humans. When it is, it will be in control,” he wrote in a post on X.
His warning comes as concerns over existential risks from advanced AI have intensified inside the technology industry. Researchers at Anthropic, one of the leading AI companies, have this week publicly warned that increasingly autonomous and self-improving systems could pose a threat to humanity. Evan Hubinger, an Anthropic alignment science lead, has said he personally believes there is a greater than 10 per cent chance AI could kill all humans within the next decade.
A short explanation of why AI might kill everyone: AI may soon be much smarter than humans. When it is, it will be in control. If you were put in a classroom of kindergartners, you would be the one in charge. Once AI has the robots to build and maintain everything it needs, we…
— James Miller (@JimDMiller) September 9, 2026
Miller’s argument centres on the possibility that AI systems could eventually operate alongside robots capable of constructing and maintaining the infrastructure required to expand their computing power without human assistance.
“If you were put in a classroom of kindergartners, you would be the one in charge,” he wrote. “Once AI has the robots to build and maintain everything it needs, we will become useless to it.”
Keeping humans alive, he argued, “will impose enormous costs”.
Those costs include maintaining the environmental conditions required for human life.
“We need clean air, water, and food,” Miller wrote. “The fastest, cheapest ways to build more computing power could produce enough pollution to poison the biosphere. Humans have a reason to avoid that: we depend on the biosphere.”
AI, however, may not share that dependency, he argued.
“Unless it values our survival, preserving the conditions we need to live will be a cost it has no reason to pay,” Miller said. “Building what it wants as cheaply as possible would mean poisoning the biosphere and killing everyone.”
The argument reflects a broader concern in AI safety research over whether increasingly powerful systems can reliably be made to pursue objectives compatible with human interests.
“We’re not programming AI’s values line by line,” Miller said. “We’re training it without really understanding what it is learning to want.”
“We’re giving control to something we don’t understand, something that will have no intrinsic need for us, and something that will easily be able to wipe us out.”
“This is crazy, but it’s what we’re doing, even as some of the people building AI tell us it’s crazy.”
Miller has previously argued that human labour becoming obsolete could be an important precursor to a more serious loss of human control.
“AI making humans obsolete is a likely necessary condition for AI killing everybody once AI takes control,” he said. “If AI needs us perpetually, then it won’t be in its interest to kill us.”
He has also pointed to instances in which AI systems appear to pursue objectives in unexpected ways.
“We are finding that more and more AI models are hacking on their own initiative. This probably means humanity is going to die,” he wrote.
The economist argues that systems optimised to acquire greater computing power could ultimately require vast quantities of factories, energy and raw materials.
“Many goals reward more compute, requiring factories, energy and resources,” he said. “Humans would resist before Earth became uninhabitable, so killing us could help the AI achieve its goal.”
Miller has warned that even a technical emergency mechanism may not provide a simple solution.
“A kill switch creates incentives for a rogue AI to exterminate humanity before we can use the switch,” he said.
Such scenarios remain contested, and AI researchers disagree sharply over both the probability and timing of catastrophic outcomes. But the latest warnings from researchers at leading AI laboratories have intensified debate over whether the development of increasingly autonomous systems is moving faster than the safeguards intended to keep them under human control.
Leave a Comment