When someone says something that simultaneously seems completely plausible and ridiculous, a certain kind of silence descends upon the room. In 2022, OpenAI CEO Sam Altman created that silence by outlining two potential futures for artificial intelligence: one so good it sounds fantastical and one he referred to as “lights out for all of us.” He wasn’t giving a performance.
He was serious. And that’s what makes the present feel so weird—possibly more than any benchmark or product launch.
| Category | Details |
|---|---|
| Key Figure | Alan Turing — Mathematician, pioneer of computer science and artificial intelligence |
| Foundational Paper | “Computing Machinery and Intelligence” — Alan Turing, 1950 |
| Key Concept | The Technological Singularity — a point where AI surpasses human intelligence and begins self-improvement |
| Notable Warning (2024) | Geoffrey Hinton — Nobel Prize in Physics, 2024. Estimated 10–20% chance AI leads to human extinction within 30 years |
| Expert Survey (2023) | 2,778 AI researchers surveyed by AI Impacts — 50% chance of human-level AI by 2047 |
| Related Company | OpenAI — founded 2015, CEO Sam Altman. One of the fastest-growing companies in modern capitalism |
| Term Coined By | I. J. Good (1965) — “Intelligence Explosion”; Vernor Vinge — popularized “Singularity” in 1983 |
| Further Reading | The Singularity Is Near — Ray Kurzweil, 2005. Predicted singularity by 2045. |
In a way, we’ve been here before. Fears of machines surpassing human intelligence date back many years, but the general public has largely written them off as paranoia and science fiction. However, something has changed. These days, it’s not dystopian novelists or fringe bloggers who are sounding the alarm. They have won the Nobel Prize.
Geoffrey Hinton, a recent recipient of the Nobel Prize in Physics and one of the architects of modern deep learning, estimated in December 2024 that there was a 10–20% chance that AI would wipe out humanity within the next 30 years. It’s not a fringe estimate. It’s difficult to know what to do with a serious man giving a serious number.
Most people are unaware of the deeper causes of this anxiety. The first serious scientific paper on artificial intelligence was published in 1950 by Alan Turing, a man who, in this writer’s opinion, belongs alongside Newton and Darwin rather than just in the background of computing history.
A year later, he projected that machines “would not take long to outstrip our feeble powers” once they could learn from experience. He believed that machines would eventually take over. In 1951, he made this statement. Perhaps the only reason we didn’t pay attention was that it seemed too far away.
Another pioneer of the field, Marvin Minsky, was equally direct in a 1970 interview with LIFE magazine. If we were fortunate, he thought, superintelligent machines might choose to keep us as pets in the future. Now, it seems darkly humorous, but Minsky wasn’t kidding. He was using mathematical extrapolation. There’s a distinction.
The mathematician I. J. Good coined the term “intelligence explosion” to describe the theory he and others were discussing: that a sufficiently powerful machine would start to improve itself, setting off a chain reaction of increasing intelligence. The singularity is a more evocative name for it nowadays.
In a nutshell, the singularity is the hypothetical point at which machine intelligence becomes self-sustaining and self-improving, surpassing human cognition so rapidly that it is practically impossible to predict what comes next.
In a 1983 article, Vernor Vinge likened it to the physics of a black hole, where the rules of everyday life cease to apply. In his 2005 book, Ray Kurzweil projected its arrival by 2045. That sounded speculative at the time. 2045 may be more optimistic than alarmist, according to an increasing number of researchers.
Nearly 2,800 researchers were polled by AI Impacts in 2023; all of them had previously published peer-reviewed work at significant AI conferences. They were asked a specific question: when will machines be able to do all tasks more effectively and affordably than human workers on their own? According to the median response, there is a 50% chance that this will occur by 2047. a 10% chance by 2027. The source of those figures is what makes them remarkable.
This was not a survey of venture capitalists or tech enthusiasts. These scientists were in the workforce, and their predictions were much more optimistic than they had been a year ago. The way experts interpret the trajectory has obviously changed.
Nevertheless, uncertainty permeates everything. Even among experts, forecasting is infamously unreliable. After compiling expert forecasts for almost 20 years, social scientist Philip Tetlock discovered that well-known experts were frequently no more accurate than chance. Whether any of these timelines—2027, 2047, and 2100—are anything more than educated guesses wrapped in probability language is a legitimate question.
According to some serious researchers, such as Paul Allen and Steven Pinker, the development of AI is more likely to follow an S-curve rather than a hyperbolic spike: quick early gains followed by a leveling off. They contend that optimistic singularity theorists often underestimate the limits imposed by the laws of physics.
Observing all of this, it’s not only the technical argument that stands out, but also its emotional register. The people who create these systems are frequently the ones who worry about them the most. In part, Hinton left Google to voice his concerns more openly. While in charge of one of the companies most actively advancing AI, Altman has advocated for worldwide regulation of the technology.
Exactly, that contradiction isn’t hypocrisy. It’s more akin to driving a car and discovering halfway around a curve that the brakes might not be what you thought they were.
What does that mean for the rest of us? The institutional and political framework needed to effectively regulate a technology that is evolving more quickly than lawmakers can keep up is still up for debate. Governments that can hardly agree on trade policy are being asked to work together on a matter that affects every facet of cognitive labor and every sector of the economy. It’s a big request. Additionally, if the researchers are correct, the timeline might not be generous.
Turing once stated that it wasn’t a question of if, but rather when, machines would surpass human intelligence. He believed it would occur gradually enough for him to handle. Regarding the first part, he might have been correct. The question of whether he was correct about the second seems to be keeping many highly intelligent people up at night, and it might now keep the rest of us up as well.
