The Singularity Is Nearer — Acceleration, Optimism, and Uneasy Futures

Ray Kurzweil’s The Singularity Is Nearer is an easy book to misread before even opening it. It is tempting to dismiss it as the wishful thinking of an ageing technologist doubling down on ideas he has promoted for decades.

That would be too easy. Kurzweil’s optimism is not vague: it is built from long-running trends in computing, economics, medicine and human development. The more difficult question is how much those trends actually prove. The book is strongest when it shows technological capabilities becoming cheaper and more powerful. It is less convincing when those curves are asked to carry precise timelines or predictions about society as a whole.

Cover image: The Singularity Is Nearer by Ray Kurzweil, published by Viking. Used here for purposes of review and commentary.

Acceleration as a Thesis, Not a Law

At the heart of Kurzweil’s worldview lies what he calls the Law of Accelerating Returns. Each generation of technology provides tools for developing the next generation, creating a compounding process. Individual technologies eventually encounter limits, but Kurzweil argues that new paradigms arrive to continue the broader exponential trend.

Moore’s Law is the best-known example, although it applies narrowly to transistor density rather than technological progress in general. Traditional transistor scaling has slowed, but other improvements—in specialised processors, parallel computing, algorithms and infrastructure—continue to increase the capability available at a given cost.

Recent AI development offers striking support for the practical part of Kurzweil’s argument. According to the 2025 Stanford AI Index, the cost of obtaining performance equivalent to GPT-3.5 fell more than 280-fold between November 2022 and October 2024.

That is an extraordinary change. It helps explain why AI capabilities that recently required a large company’s infrastructure can rapidly become available through an ordinary application or device.

It does not, however, prove that all technological development shares one universal curve, or that improvements in computation translate automatically into equivalent advances in medicine, energy, institutions or human welfare. The Law of Accelerating Returns is best understood as an ambitious model of technological history—not a physical law guaranteeing the future.

A Broader Book Than Expected

The singularity is the book’s organising destination, but it is not its only subject. Kurzweil’s vision involves more than machines simply becoming smarter than humans. He imagines human intelligence expanding through increasingly intimate integration with artificial systems: first through external assistants and eventually through direct cognitive augmentation.

Much of the book therefore reads as a broad survey of technological and social progress. Artificial intelligence provides the connective tissue, touching medicine, energy, manufacturing, education and cognition.

In that respect, the book revisits territory explored in works such as Yuval Noah Harari’s Homo Deus and Kelly and Zach Weinersmith’s Soonish. Kurzweil’s approach is more cumulative. Each chapter adds trends intended to support not merely the continuation of progress, but its acceleration.

The early sections also work as a rebuttal to cultural pessimism. Kurzweil draws on long-term changes in life expectancy, literacy, poverty and violence to argue that human conditions have improved far more than daily news and political rhetoric suggest.

There is considerable truth in that. Dramatic events dominate attention, while gradual improvements rarely generate headlines. A disaster is visible in a way that a slow decline in mortality or poverty is not.

The difficulty is that global averages can conceal as much as they reveal. Progress may be real without being evenly distributed, irreversible or automatic. A statistic showing improvement across two centuries does not tell us that every region is improving now, or that the institutions responsible for earlier gains will continue to function.

Kurzweil is most persuasive when he reminds us that pessimism can be statistically illiterate. He is less persuasive when historical improvement begins to look like evidence that future improvement is the natural condition of technological society.

Automation, Work and Disruption

Kurzweil expects automation and AI to eliminate some roles while creating new ones. Previous technological revolutions support the general possibility that employment can grow even while particular occupations disappear.

Yet occupations are bundles of tasks, and technologies rarely automate every task at once. Current research suggests that most jobs exposed to generative AI are more likely to be transformed than eliminated, at least in the near term.

That may sound reassuring, but transformation can still be painful. Work can become more closely monitored, less autonomous or divided between a smaller group of highly rewarded specialists and a larger group expected to follow machine-generated instructions.

Reskilling is also easier to recommend in the abstract than to accomplish in practice. A worker displaced after decades in one occupation cannot necessarily move smoothly into a newly created technical role. Age, education, geography and family obligations do not disappear because the economy has produced more jobs in aggregate.

Technological progress does not occur in a social vacuum. Even when it raises total productivity, it can distribute its benefits and losses in ways that generate political anger and genuine insecurity.

Measuring Value in a Digital Economy

One of the book’s more interesting discussions concerns the limits of conventional economic statistics. Free or inexpensive digital services can create enormous value without generating a corresponding monetary transaction.

Wikipedia is an obvious example. Its value to readers is not reflected by a market price because access is free. Research into the consumer welfare created by digital goods suggests that national accounts omit a substantial part of what users gain from such services.

This does not make GDP meaningless. GDP measures market production, not happiness, usefulness or every form of consumer benefit. The problem begins when a measure designed for one purpose is treated as a complete measure of progress.

Kurzweil also touches on containerisation as an example of infrastructure whose transformative impact was easy to overlook. That aside reminded me of Marc Levinson’s The Box, which shows how a standardised metal container quietly reorganised ports, factories, transport networks and global trade.

Some of the most consequential innovations do not initially look revolutionary. They become visible only after they have rebuilt the systems around them.

Extending and Copying the Mind

Kurzweil’s route to the singularity ultimately runs through the extension of human cognition. AI assistants already function as external cognitive tools. Brain–computer interfaces are an active field of research, although their current capabilities remain far more limited than the seamless integration Kurzweil anticipates.

Fully digital or simulated minds belong to a much more speculative category. The fact that we can describe whole-brain emulation does not mean that we know how to perform it, what information would need to be preserved or whether the resulting entity would possess the consciousness and identity of the original person.

Robin Hanson’s The Age of Em explores the social consequences of one such scenario with unusual seriousness. I later returned to those implications in an essay about a future populated by copies of human minds.

These questions deserve more attention than Kurzweil can give them while surveying almost every major technology at once. Increasing intelligence is not the same problem as preserving identity. A digital mind may be technically possible without answering whether it is a continuation, a copy or an entirely new person.

Optimism and the Missing Step

I do not share Kurzweil’s confidence in his timelines or in the inevitability of the destination. I do find his general emphasis on compounding capability persuasive. Technologies can reinforce one another, and the interaction between cheaper computation, better algorithms, improved sensors and larger bodies of data can produce changes that look abrupt from the outside.

But capability is not outcome. A technology becoming possible does not tell us who will control it, how widely its benefits will be distributed or which uses institutions will encourage.

Kurzweil acknowledges risks from artificial intelligence, biotechnology and nanotechnology, but his optimism tends to absorb them. They become engineering or governance problems encountered along an otherwise rising curve. Authors such as Nick Bostrom and Olle Häggström are more willing to consider the possibility that increasing capability may outpace our ability to manage it.

The Singularity Is Nearer remains compelling because it asks readers to take exponential change seriously. Human intuition is poor at imagining repeated multiplication, and there is genuine danger in assuming that the next twenty years will resemble the previous twenty.

The opposite mistake is to treat an exponential graph as destiny.

Kurzweil may be broadly right that technological capability will continue to accelerate. That still leaves the questions that matter most unanswered. Acceleration can tell us that more futures are becoming possible. It cannot tell us which one we will choose—or whether the choice will belong to all of us.