CEOs of companies developing artificial intelligence have warned about its risks many times before. But no one has ever told people as directly as Paul Christiano that they will most likely die because of AI. The OpenAI board member, who sits on its safety committee, gave a lecture in Toronto and told the audience that they were likely to die in an accident caused by artificial intelligence.
“For most people in this room, the most likely cause of death will be an accident caused by artificial intelligence,” Christiano said during this year’s Hinton Lectures, held annually by the AI Safety Foundation. The event was moderated by Nobel Prize winner Geoffrey Hinton, who is known as the godfather of artificial intelligence. Christiano delivered the opening address.
When machines make decisions about us
Christiano did not mean his statement as an exaggeration. He was talking about a so-called AI takeover—a situation in which AI systems pursue their own goals and make decisions about events on Earth just as humans do today.
“If you, I, or perhaps Donald Trump wanted to stop them, we would have the same chance as if we were trying to stop the U.S. military,” the former OpenAI researcher explained.
During the lecture, he also suggested how quickly this could happen. According to him, it could occur within the next six months, or within three years under a more moderate scenario. “If we developed AI as quickly as possible and took the simplest path, a takeover would be roughly a coin flip,” he explained.
Companies are slowing down, yet the risk is still growing
Christiano acknowledged that companies are currently avoiding this coin flip because they are deliberately slowing the development of the most powerful models. Anthropic CEO Dario Amodei recommended this approach after his former employee accused both Anthropic and OpenAI of gambling with our lives. But according to Christiano, even that is not enough. Even if politicians respond sensibly, he says the risk remains extraordinarily high.
His tone has grown more severe in recent months. When he joined OpenAI’s board in September, he estimated a four percent chance that humans would lose control of AI within a year and a fifteen percent chance within three years. At the time, he acknowledged that these were his personal estimates. In a statement published when he joined, he wrote that if humanity builds superintelligence without more reliably teaching it to follow human intentions, he expects a permanent loss of control. “If that happens, most people may die,” he added. According to him, only cooperation within individual countries and between nations would reduce the risk.
He spoke similarly in an interview for a popular technology podcast, where he said that the probability of completely losing control over artificial intelligence within the next year is rising rapidly. He also mentioned a non-negligible chance that the world could find itself in a very dangerous situation within six months.
Fear of AI has been growing for months
Concerns about where artificial intelligence is heading have been growing for some time, and executives at the largest companies are only fueling them. OpenAI CEO Sam Altman has acknowledged that humanity could lose control of the technology. Anthropic, meanwhile, has warned that someone could misuse its models to produce chemical weapons.
Not everyone sees the future in such bleak terms, however. Researcher and author Gary Marcus previously warned that AI could amplify serious dangers such as cyberattacks, biological weapons, and the spread of disinformation. But his view of the threat is considerably more measured than Christiano’s warning that most of those present in the room will one day be killed in an accident caused by artificial intelligence.
Sources: theinformation.com, fastcompany.com and businessinsider.com



