It could launch biological weapons, or start a global drone war. It could detonate nuclear bombs. Or it might destroy human society in a more gradual way — by hoarding all the world's economic resources, leaving us without food, shelter, or medicine.
These are some of the dark imaginings now swirling around artificial intelligence. The technology has proven to be a monumental advance in human achievement, yet some fear it could just as easily spell our extinction. This week, the anxiety made global headlines when an Anthropic researcher left the industry, saying he feared his lab and others like it were building superintelligent machines that humans cannot control.
It's understandable to fear what AI might do. But the modern world is full of risks that could threaten human survival. Humanity faces annihilation from killer asteroids, pandemics, thermonuclear war, climate change, massive blobs of hot material ejected from the sun, ecosystem collapse, and a sudden ice age.
Scientists and public policy experts say the problem is less about the threats themselves — AI being merely the latest — than about humanity's innate difficulty in seeing and weighing them properly. "We can tolerate risks that we have gotten used to, and we are afraid of risks that are new," said James Hammitt, director of the Harvard Center for Risk Analysis. We drive more than we fly, he noted, so "people consider air travel to be more dangerous than motor vehicles, when in fact it isn't." AI, he pointed out, "is clearly not yet well understood, and not within what we believe to be our own control, so it raises enormous social concern."
Among the tech executives and researchers building AI systems, a faction has long warned that AI could destroy humanity. Some argue that in the short term, AI could help malicious people without specialized skills create viruses or toxic substances, causing mass disease and death. Just this Thursday, Anthropic said it had thwarted several attempts by scientists conducting research that could lead to biological weapons development.
Over the longer term, those warning of AI's dangers argue that tech companies may give AI systems more autonomy, connect them to the internet, and even link them to critical infrastructure including power grids, stock markets, and military weapons. Once connected, they believe, AI systems could cause problems in unforeseen ways. Today, if an AI system goes rogue, companies like OpenAI can simply turn it off. But some experts worry future systems might resist human attempts to shut them down.
For now, these are hypothetical concerns. But this debate about "AI doomsday" is less about predicting destruction than about how we face the unknown: when a technology grows powerful enough to exceed our understanding and control, the rational response is neither blind panic nor casual dismissal — but to acknowledge it, study it, and prepare for it, as we do with every other existential threat.