Book Review: Singularity is Nearer | Ray Kurzweil
Ray Kurzweil has been wrong about half his predictions, and that’s exactly why this book is worth your time.
The vibe
Pacing: Uneven. Dense data chapters next to breathless speculation.
Drives the story: Ideas. Relentlessly.
“X meets Y”: TED talk meets actuarial table meets sermon.
Page 99 feel: You’re nodding at a chart, then he casually predicts you’ll live forever.
What works
The scaffolding, not the predictions. Kurzweil’s Law of Accelerating Returns isn’t useful because it nails AGI to 2029. It’s useful because it gives you a mental model for why your five-year roadmap is probably wrong. Computing power is 11,200x cheaper than when the original Singularity Is Near dropped in 2005. That’s not a trend line — that’s a different planet. Whether you buy his specific dates or not, the model forces you to stop pretending linear planning still works.
The updated receipts. The 2005 book felt like prophecy. This one reads as a progress report. Kurzweil moved his Turing test prediction from 2034 up to 2029. Given what GPT-4 and Claude showed by 2024, that bet doesn’t look crazy. He earns the right to update because he shows his work — Stuart Armstrong pegged his accuracy at 42%, far below Kurzweil’s self-reported 86%. Even critics admit his directional instincts hold. He saw the shape of the wave, even when he got the timing wrong.
The contrast with my last review. If AI Snake Oil (January’s review) is the cold shower on overhyped predictive AI, this book is the hot tub of generative AI optimism. Reading them together is the real move. Narayanan tells you what AI can’t do. Kurzweil tells you what it might. Neither is complete without the other.
That said — it earns some eye-rolls too.
What doesn’t work
The repetition tax. If you read the 2005 book, you’re paying full price for maybe 40% new material. Goodreads sits at 3.9/5 (68% positive), and the loudest complaint is “I already read this.” The Washington Post’s Becca Rothfeld called it “daunting to summarize because it is so careless and careening.” That tracks. Needed a heavier editorial hand.
The optimism without friction. Kurzweil treats existential risks like speed bumps — addressable, manageable, not worth losing sleep over. Nanotech reversing aging by the 2030s. Brain-cloud interfaces. “After Life” technology. Paul Davies warned that “exponential growth never lasts.” Zero airtime here. Shermer observed the Singularity has “evolved from event to metaphor in 20 years.” Kurzweil hasn’t absorbed that shift. He’s still writing like it’s an engineering milestone when the metaphor is the better lens.
The circular reasoning on timelines. When predictions miss, he updates the model to fit. When they hit, the model is vindicated. That’s not science — that’s storytelling.
The 3-question check
QuestionAnswerWhat was he trying to do?Update his 2005 thesis for the LLM era and prove the Singularity timeline is on trackDid he pull it off?Halfway. The LLM updates land. The timeline defense is circular.Was it worth doing?Yes — not for the predictions, but for making you sit with exponential change as a real thing.
Who should read this
Read this if: You make technology strategy decisions and need a model for thinking about non-linear change. Pair with AI Snake Oil for balance.
Skip this if: You read the 2005 original cover-to-cover and want genuinely new arguments. Or if utopian techno-optimism with no real risk analysis drives you nuts.
Reviewed for the next reader, not the author.
Amazon link: The Singularity Is Nearer: When We Merge with AI
