A college student confessed a prank that panicked a room full of people, and he now sees the same dynamic playing out in AI research. The comparison is not abstract. It is personal.
The author, who has never shared this story before, recounts a night in his first year at university when he and a group of fellow students decided to spread a rumor. They walked into a general computer lab next door and loudly announced that a virus had escaped the computer science lab. The students there should eject their floppy disks to stop it from spreading.
The panic was immediate. People powered off machines, pulled cables from sockets, screamed, and ran out of the room. The pranksters tried to stop the panic. They could not. Some students responded with fury. Others refused to believe it was a joke. Once the fear took hold, it could not be erased.
The aftermath was bad. It was finals week. Students who had powered down their computers lost work they had been writing. They were furious. The pranksters were in serious trouble.
The next morning, they met the facilities director. He lectured them sternly, made them write letters of apology, and warned that another repeat would mean leaving the university entirely. When the author graduated four years later, the director returned the letter. The author remains grateful for the gravity of how the whole thing was handled.
The story is now public because the author believes he has seen fear spread among technologists in a way that feels irresponsible. He points to recent claims about AI causing human extinction.
The AI Extinction Claim
Ex-Anthropic employee Jacob Coxon claimed that the probability AI will “kill all humans” is “>10% in the next decade.” Anthropic Alignment Science lead Evan Hubinger agreed with the figure. The author describes this as a claim that, if true, means a newborn child has a greater than ten percent chance of dying before reaching middle school.
The claim is not that AI could kill thousands or millions of people. It is that there is a greater than ten percent chance AI will kill all humans. The author calls these ghoulish claims. Given the obvious importance of AI, they have leapt into the mainstream. People naturally ask how that could happen.
The answers always rely on hand-wavy extrapolation into the future. Coxon cites “hacking critical infrastructure” and “extinction-level bioweapons” without further elaboration. The author notes that Coxon is not an expert on critical infrastructure, bioweapons, or extinction.
Fear Spreads Faster Than Evidence
The author draws a direct parallel between the prank and the current situation. In both cases, a small group of people sowed fear among a larger group. In the prank, the fear was contained to a single room. In the AI case, it has spread across the field.
The author quotes Jonathan Swift, who said, “Falsehood flies, and the truth comes limping after it.” When domain experts sow fear, it takes root more readily than any evidence that follows. The author describes Coxon as more vector than index case. The fears of extinction risk are widespread, and Coxon likely has these fears because he has heard them from someone else.
As the fear takes hold, the sheer number of frightened experts becomes its own kind of evidence. Contrary voices are drowned out by the apparent consensus.
Building Computers, Not Robots
The author offers a rebuttal from his own narrow domain of expertise: building computers. He reflects on his 2023 Monktoberfest talk, “Intelligence is not Enough,” where he argued that acts of engineering are not acts of intelligence alone. They require human action in the physical world.
Those who are resolute in their fear like to gloss over the physical reality. “Robots will do this!” they say. But robots cannot do this today. They do not seem to be doing so on a time horizon consistent with these extinction fears.
More generally, to the degree that digital systems have agency in the physical world, it is agency that humans grant them. Humans build the machines, program them, and control their interactions. The author argues that the physical reality matters.
The Comparison Holds
The author’s central argument is that fear spreads faster than evidence. In the prank, a few students created chaos by lying about a virus. In the AI case, experts are creating chaos by claiming a high probability of human extinction.
The parallels are striking:
- A small group creates fear in a larger group
- The fear spreads quickly and cannot be stopped
- Contrary voices are drowned out by the apparent consensus
- The fear takes root before any evidence can catch up
The author’s worry is that the fear is becoming well established. He wants to offer a rebuttal from his own domain of expertise. He wants to remind people that robots cannot do this today.
The comparison is uncomfortable. The author admits that he could have kept the prank to himself. He chose to share it now because he sees the same dynamic playing out in AI research.
The fear is spreading. The author hopes his voice can be part of the pushback.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.