Kurzweil Scorecard: The Adoption Curve Collapsed. The Poverty Line Didn’t.

When Ray Kurzweil wrote The Singularity Is Near in 2005, he stacked up
ten factual claims and forecasts about how fast technology would spread,
how quickly poverty would fall, and how much noisier the backlash would
get. Two decades later, the headline: he was almost comically right
about adoption speed and B2B commerce, mostly right about education and
neuroscience, wrong about a specific World Bank number, and early on a
Luddite revival now showing up in strike votes and dumbphone sales.

Most singularity post-mortems fixate on AGI timelines. This batch is
quieter and more instructive — the part of the book where Kurzweil
calibrates the reader’s intuition about how fast the world moves.

The predictions

Ten claims, mostly made “circa 2005.” Some are historical assertions
(WWII civilian deaths, B2B e-commerce in 2002, college enrollment
growth). Others are forecasts about the mid-2010s and 2020s (adoption
compression, poverty decline, Luddite reactions). They sprawl across
economics, education, neuroscience, and sociology — the point being
that every axis is moving faster than intuition suggests.

Where we actually are

Adoption timelines. Kurzweil under-called his own prediction.

Kurzweil wrote that “the delay between early and late adoption of
technologies, about a decade in 2005, will shrink to about five years
in the mid-2010s and a couple of years in the mid-2020s” (ch. “on
Work”). The smartphone hit 50% of knowledge-worker use in roughly 60
months. ChatGPT reached 100 million monthly users in two months and
crossed 50% knowledge-worker penetration inside 36. Time to 70% US
household adoption fell from about 40 years around 1900 to 17 years by
2000; for generative AI the equivalent window is now measured in
quarters. On this one, Kurzweil wasn’t bold enough — the late-adopter
lag for software-only products has compressed past his “couple of
years” to something closer to months.

Digital divide. Right on direction, still honest about the gap.

Kurzweil argued the digital divide “was diminishing rather than growing
because digitally connected humans were increasing rapidly and many
regions were leapfrogging wired telephony via wireless Internet access”
(ch. “DNA Sequencing, Memory, Communications, the Internet, and
Miniaturization”). The ITU’s 2024 Facts and Figures reports 5.5 billion
people online, 68 percent of the world, up from roughly 17 percent in
2005. Mobile broadband now carries more subscriptions than wired voice
ever did; 51 percent of the planet is inside 5G coverage. The
leapfrog is real — and so is the caveat. Rural internet use (48%) still
trails urban (83%); the cost of a fixed-broadband subscription in
low-income countries is nearly a third of average monthly income. The
divide narrowed; it didn’t close.

East Asia Pacific poverty. Number missed, trend nailed.

Kurzweil cited the World Bank’s projection that extreme poverty in East
Asia and the Pacific would “fall to under 20 million by 2015” (ch.
“The Singularity as Economic Imperative”). The actual 2015 figure at
the $1.90-per-day line was in the 60–80 million range — far below the
500+ million of the mid-1990s, but nowhere near the sub-20 million
target. In The Singularity Is Nearer, Kurzweil reframes more
carefully: “From 1990 to 2013 there was an astonishing 95 percent drop
in East Asia’s extreme poverty, in a total population that grew from
1.6 billion to 2 billion during that period.” Big story correct, 2015
milestone missed.

B2B e-commerce. Passed the trillion mark, kept going, kept going.

Kurzweil cited smooth B2B e-commerce growth “from $56 billion in 1999
to $482 billion in 2002, approaching $1 trillion in 2004” (ch.
“Deflation … a Bad Thing?”). Twenty years on, analyst estimates for
global B2B e-commerce in 2024 cluster between $10 trillion and $19
trillion depending on methodology, with consensus 2030 projections
around $47 trillion. The trillion-dollar marker Kurzweil flagged as
imminent in 2004 became the new floor — about ten to twenty times
over by 2024.

Education investment. 120x growth on one axis, roughly 10x on the other.

Kurzweil wrote that “over the previous 120 years, investment in K-12
education per student had increased tenfold in constant dollars and the
number of college students had increased hundredfold” (ch. “The
Singularity as Economic Imperative”). US college enrollment was about
160,000 in 1900 and 19.28 million undergraduates in Fall 2024 — that’s
roughly 120x, beating the hundredfold claim. K-12 per-pupil spending in
real dollars rose from about $4,060 in 1970-71 to $16,526 in fiscal
2023, and earlier figures suggest Kurzweil’s tenfold-over-120-years
line is directionally right. Both claims verify.

Brain reverse-engineering patents. Right then, a much bigger story now.

Kurzweil noted that “patent applications have already been filed based
on brain reverse engineering” (ch. “Analytic Versus Neuromorphic
Modeling of the Brain”). In 2005 this was a modest claim, backed by
filings such as US H632 (“Optical neuromorphic embodiments of
self-regulating neural networks”) and US 4,874,963 (“Neuromorphic
learning networks”). It didn’t stay modest. 579 US grants now mention
neuromorphic hardware — 2 in 1989, 15 in 2015, 66 in 2020, 71 in 2023,
56 in the first nine-plus months of 2025. The claims have teeth. US
12,602,228, granted April 2026, specifies a message-exchange facility
where neural events carry accumulation or leakage instructions,
modifying destination-neuron state directly — a primitive straight out
of the spiking-neuron literature. US 12,598,855, granted the same
month, describes a memristor-based synaptic device with an Al2O3
insertion layer supporting both electrochemical-metallization and
valence-change filaments, giving one device two independent switching
mechanisms. These are manufacturable circuits, not architecture
cartoons.

50,000 neuroscientists, 300 journals. Past it now.

Kurzweil cited “more than 50,000 neuroscientists worldwide publishing
in more than 300 journals” (ch. “How Complex Is the Brain?”). The
Society for Neuroscience alone has grown from about 500 members in
1969 to nearly 35,000 across 80+ countries. Neuroscience-tagged papers
indexed in our literature corpus show 9,906 publications in 2024 alone
and 145,276 cumulatively. The broader journal ecosystem now reaches
well past 300. Verified and extended.

World War II civilian deaths. Historical fact.

Kurzweil wrote that “about fifty million civilians died in World War
II” (ch. “on Warfare”). Scholarly consensus puts civilian deaths
between 50 and 55 million out of total war-related deaths of roughly
70–85 million. Verified.

US Army web-based training. Overstated.

Kurzweil claimed “the U.S. Army already conducted all of its
nonphysical training using Web-based instruction” (ch. “on Learning”).
The Army Learning Management System was rolled out in the mid-2000s as
a centralized e-learning platform, seeded by the Advanced Distributed
Learning initiative from 1997. But the claim that all nonphysical
training was web-based in 2005 conflicts with the record. Classroom
instruction at the service schools, small-group exercises, and
classified briefings continued on-site. Direction correct, 2005
milestone overstated.

Luddite reactions. Arrived on schedule.

Kurzweil projected that “accelerating technological change will
provoke increasing fundamentalist and Luddite reactions” (ch. “on
Work”). In The Singularity Is Nearer he doubled down: “I do think
we also need to take seriously the misguided and increasingly strident
Luddite voices that advocate broad relinquishment of technological
progress.” The 2023 SAG-AFTRA and WGA strikes ended with contracts
restricting AI-generated scripts. In 2024, Amazon and Google staff
walked out over a $1.2 billion AI defense contract. The Luddite Club,
founded by Brooklyn teens rejecting smartphones, now has 20+
high-school and college chapters in the US. Dumbphone sales are up;
vinyl is back; Edelman’s Trust Barometer shows trust in AI companies
falling from 62% in 2019 to 54% in 2024.

The scorecard

What Kurzweil missed (and what he nailed)

The pattern is sharp: Kurzweil is excellent at direction and slope, bad
at specific thresholds, and almost always too cautious on the steep
part of the curve. When he puts a number on a forecast — under 20
million in East Asia poverty by 2015, a couple of years in adoption
lag by the mid-2020s — the number is the weakest part of the claim.
When he describes a trajectory, he’s usually right and often too
conservative. The B2B e-commerce line he flagged as “approaching $1
trillion” now measures in tens of trillions. The neuromorphic patent
trickle he pointed at in 2005 became a river between 2019 and 2023.
The Luddite backlash he projected to the 2030s is already here.

The forecasting lesson isn’t that Kurzweil was optimistic. It’s that
when he was wrong, he was too pessimistic on the exponential and
too specific on levels. The Kurzweil who would have scored best in
2005 hedged less on when, and more on what it would look like when it
arrived.

Method note

Verdicts came from a full-text patent corpus of 9.3M US utility
patents and a literature corpus of roughly 357M scientific works,
queried for neuromorphic and neuroscience terms; plus live web
searches for World Bank, ITU, SfN, and NCES figures, reporting on
neo-Luddite activity, and analyst ranges for B2B e-commerce. Source
texts: The Singularity Is Near (2005) and The Singularity Is
Nearer (2024).