The Technological Singularity

The technological singularity is the idea that technology, and especially machine intelligence, could one day improve so fast that the future beyond that point becomes impossible to predict. Some serious thinkers expect it within decades. Others think it rests on a misunderstanding. This essay covers where the idea came from, what its main voices have claimed and the strongest arguments on each side.

Where the word comes from

In mathematics, a singularity is a point where a function stops behaving normally, often by growing without limit. In physics, the center of a black hole is described as a singularity because our current theories break down there. Applied to technology, the word means a point past which our models of the world stop being useful.

The first recorded use in this sense comes from the mathematician Stanislaw Ulam. In a 1958 tribute to John von Neumann, published after von Neumann's death, Ulam recalled a conversation about the ever-accelerating progress of technology. Von Neumann had said it seemed to be approaching an "essential singularity" in human history, beyond which human affairs as we know them could not continue.

The intelligence explosion

In 1965 the British statistician I. J. Good, who had worked with Alan Turing on codebreaking during the Second World War, published "Speculations Concerning the First Ultraintelligent Machine." His argument was simple. Designing machines is one of the things intelligent beings do. So a machine that is better than any human at intellectual work would also be better at designing machines. It could build a still better successor, which could build a better one again.

Good called the result an "intelligence explosion," and he concluded that the first such machine might be the last invention humanity would ever need to make, provided it remained under our control. That final condition is the root of today's debate about AI safety.

Vinge makes it famous

In March 1993, the mathematician and science-fiction author Vernor Vinge presented "The Coming Technological Singularity: How to Survive in the Post-Human Era" at a symposium sponsored by NASA's Lewis Research Center and the Ohio Aerospace Institute. He argued that within about thirty years we would have the means to create intelligence greater than our own, and that shortly afterward the human era would end in the form we know it.

Vinge was careful to describe several paths: computers that "wake up," large networks that become intelligent as a whole, computer interfaces so close that users become superhumanly capable, and biological improvements to human intellect. He wrote that he would be surprised if the event came before 2005 or after 2030.

Kurzweil puts a date on it

The inventor and futurist Ray Kurzweil brought the idea to a mass audience with The Singularity Is Near (2005). His argument rests on what he calls the law of accelerating returns: information technologies improve exponentially, and each generation of tools is used to build the next one faster. Kurzweil predicted that computers would match human intelligence by 2029 and that the singularity, when people merge with the technology they build and expand their intelligence many times over, would arrive around 2045.

In The Singularity Is Nearer (2024), he kept both dates and pointed to the progress of large language models as evidence that the 2029 milestone is on track.

The case for taking it seriously

• The curves are real. Computing power, data and spending on AI have grown at exponential rates for years, as the first essay describes.
• Recent progress surprised experts. Systems now do things, such as writing working code or explaining complex topics, that many researchers expected to take much longer.
• AI is already used to build AI. Machine learning helps design chips, write software and tune other models, which is a mild version of Good's feedback loop.
• The stakes are high. The philosopher Nick Bostrom argues in Superintelligence (2014) that even a modest chance of a machine far smarter than us deserves serious preparation, because such a system might be very hard to control once it exists.

The case against

• Exponentials end. Most real-world growth curves flatten into S-shapes as they hit physical, economic or practical limits.
• Intelligence is not one dial. Critics argue that intelligence is many different abilities, not one quantity that can simply be turned up.
• The complexity brake. In a 2011 MIT Technology Review essay, "The Singularity Isn't Near," Microsoft co-founder Paul Allen and computer scientist Mark Greaves argued that the deeper science goes into the brain, the more complexity it finds. Understanding that complexity, they said, slows progress instead of speeding it up.
• The real world is slow. Even a brilliant system must run experiments, build factories and win the trust of people and regulators. Those steps take time no matter how fast the thinking is.

Where the debate stands

The question has moved from science fiction into boardrooms, laboratories and government hearings. Leaders of major AI labs now speak openly about building systems that match or exceed human ability across most tasks, and governments have begun to write rules for the most capable models. At the same time, many researchers remain doubtful about dates, and some doubt the concept itself.

What nearly everyone agrees on is the narrower point behind it. Technology compounds, intelligence is the most powerful thing it could amplify, and how that goes depends on choices people make now.

Next: How Technology Compounds