A spiral timeline depicting the entire history of human technology was created by Max Roser at Our World in Data. Stone tools at the center mark the beginning of it 3.4 million years ago. The spiral’s rings each stand for 200,000 years.
Before our ancestors discovered how to control fire for cooking, it took twelve complete turns, or twelve sweeps over 200,000 years each. 2.4 million years, twelve rings. One finding. When you look at it that way, the rate of change over the past 200 years no longer seems like progress but rather something completely different. Something more difficult to identify.
| Category | Details |
|---|---|
| Core concept | Accelerating change — the observed exponential nature of technological progress; each paradigm shift occurs faster than the last |
| Stone tools to controlled fire | 2.4 million years — the longest gap between major human technological milestones |
| First flight to Moon landing | 66 years (1903–1969) — many people witnessed both within a single lifetime |
| Knowledge doubling rate (Buckminster Fuller) | Doubled every 1,500 years in 1 CE; every 250 years by 1750; every 150 years by 1900 — acceleration measurable across centuries |
| Kurzweil’s Law of Accelerating Returns | Predicts 20,000 years of progress at today’s rate during the 21st century; technological singularity projected ~2045 contested |
| Economic doubling rate (Industrial Revolution) | World economic output doubles every ~15 years — 60× faster than the agricultural era (every ~900 years) |
| AI as an accelerant | Max Roser (Our World in Data): if AI drives innovation, change compressed over decades may happen within a single year concern |
| Regulation gap | UNCTAD (2019): technology has “raced ahead of attempts to regulate it” — genomics, AI, and social platforms all cited |
| Counter-argument | Theodore Modis and Jonathan Huebner argue the rate of innovation may have peaked and begun declining in the late 2010s |
| Official reference | Our World in Data: Technology over the long run (ourworldindata.org) |
Roser must unroll the spiral into a straight line and then extend it further by the year 1800 because the inventions are coming in so fast that they can no longer fit the original format. In less time than the period between the invention of writing and the printing press, the telegraph, telephone, internal combustion engine, powered flight, penicillin, nuclear fission, transistor, and internet were all developed. When Neil Armstrong set foot on the moon, Orville Wright was still alive. Every time you take a moment to consider that fact, it is shocking.
For many years, Ray Kurzweil has worked to provide a mathematical framework for what that spiral reveals. His Law of Accelerating Returns, which was developed during the 1990s and has since been improved, contends that technological change is exponential rather than linear and that the exponent is increasing. According to his perspective, the twenty-first century will not witness a century of advancement.
If we measure progress at the current rate, it will be closer to 20,000 years. Although Kurzweil is a divisive figure and his particular predictions have drawn harsh criticism, the fundamental finding—that every technological advancement accelerates the development of subsequent generations—is not particularly contentious among scientific historians. Even those who disagree with Kurzweil’s timeline typically support the course.
Long before Kurzweil formalized the pattern, Buckminster Fuller saw it. According to Fuller, it took roughly 1,500 years for the entire corpus of human knowledge to double by the year 1 CE. It took 250 years for the next doubling, from two to four units. The doubling time had shrunk to 150 years by 1900. In the middle of the 20th century, before the internet and personal computers were invented, Fuller saw a curve that didn’t appear to be flattening. More knowledge produces more tools for learning, which accelerates the production of more knowledge. The feedback loop intensifies.
The fact that the institutions created to manage the risks were built for a much slower world and that the benefits and risks arrive at the same speed makes this genuinely challenging to think about. Jacob Corn, a professor of genome biology at ETH Zürich, described the state of gene-editing technology at a UN commission session in Geneva in 2019 with a controlled excitement that comes from working on something extraordinary.
He stated that interventions that alter the relevant DNA could address genetic diseases that are currently incurable, such as muscular dystrophy, malaria susceptibility, and conditions that are currently only treated symptomatically. There is technology available to accomplish this. With every year that goes by, it gets more affordable and easier to obtain. Additionally, he admitted that it has outpaced all significant attempts to create governance around it because it is so inexpensive and moves so swiftly. The frameworks were not necessary for the technology to advance. Seldom does it.
The most recent example of this tension is artificial intelligence, but it’s important to keep in mind that AI is merely the most recent manifestation of the pattern rather than an exception. Roser at Our World in Data has been careful to point out that what makes AI unique is that it’s not just a new technology. Since intelligence has been the driving force behind the acceleration throughout, it has the potential to be a technology that speeds up the development of other technologies.
The shift that now takes ten years could occur in a year if AI significantly increases the capacity for scientific discovery and technical problem-solving. Maybe more quickly. Although that assertion is conjectural, those who make it are not crazy. They are engineers working at large research centers, observing and calculating the capability curves of existing systems.
There are opposing viewpoints that should be taken seriously. According to Theodore Modis and Jonathan Huebner, since the late 2010s, there may have been a decline in the rate of true innovation, which is determined by transformative breakthroughs per capita rather than by patent filings or published papers.
They argue that some of the most productive periods of innovation may already be behind us and that the sheer amount of scientific activity can be mistaken for the depth of advancement. They might be correct. Technological change forecasting has a long history of accurate forecasts that needed to be revised.
However, considering the lengthy spiral—it took 2.4 million years to master fire, 66 years for the first flight to the moon, and now AI systems that can write code and conduct research in the time it takes to get coffee—it appears that the pace is not slowing down. In research labs, regulatory offices, and the kind of late-night discussions that don’t make headlines, the question isn’t whether the change is happening quickly. It’s whether those in charge of it have a genuine chance of staying up to date.
