Kurzweil Scorecard: The Network Kept Its Promise on Access. It Broke the One on Democracy.

In 2005, Ray Kurzweil made two bets about what cheap, decentralized communication would do to the world. The first was about reach: the Web would push high-quality education and medicine into the poorest corners of the planet. The second was about politics: the same networks that had already grown the world from twenty democracies to more than a hundred would keep grinding autocracy down.

Nineteen years later, the reach bet is largely paying off. The politics bet has not just failed — it has inverted. The level of democracy enjoyed by the average person on Earth is now back to where it stood in 1978, before the personal computer existed. The wires that Kurzweil expected to liberate turned out to be just as good at surveillance and manufactured consent as at dissent. And in his 2024 follow-up, Kurzweil quietly edited the record.

The predictions

This batch sits in Kurzweil’s chapter on the promise and peril of genetics, nanotechnology, and robotics — the part of The Singularity Is Near where he turns from silicon to society. He argued that technology is inherently democratizing, that governance would need occasional targeted brakes rather than blanket bans, and that humans would remake their own bodies through a series of small, individually unremarkable steps.

They are softer claims than his transistor-count forecasts, and harder to score. But several are testable, and one of them is testable against a number that has moved a very long way in the wrong direction.

Where we actually are

The democratization thesis. Kurzweil wrote that the world “had grown from about twenty democracies after World War II to more than one hundred by 2005, largely through decentralized electronic communication” (ch. “Promise and Peril of GNR”). The count is roughly right: by the mid-2000s, most tallies put the number of electoral democracies above a hundred, at or near its all-time peak. That was the good news, and it was real.

Then the curve bent. The V-Dem Institute’s 2026 Democracy Report — built on expert coding of every country in the world — finds that the average global citizen now lives under a level of democracy last seen in 1978. Seventy-four percent of the world’s population, around six billion people, live in autocracies. For the second year running, there are more autocracies (92) than democracies (87). Freedom of expression is worsening in a record share of countries. The report’s own subtitle asks whether we are unraveling the democratic era.

So the milestone Kurzweil cited was true — and turned out to be the peak. The mechanism he credited did not keep working. That is not a small miss, because the mechanism was the prediction.

Look at what the same networks actually did. The academic literature tells the story cleanly: papers on social media, disinformation, and democracy were a rounding error a decade ago — a single study in 2016 — and there were nearly 200 in 2025 alone, a curve that turns sharply upward right around the 2016 elections and the Cambridge Analytica revelations. The Arab Spring, which Kurzweil in 2024 still describes as “largely enabled and driven by social media,” mostly collapsed back into autocracy or civil war. And a new defensive industry has sprung up to police the medium itself: US patent filings that mention deepfakes or synthetic-media detection were essentially nonexistent before 2021 and jumped to 18 in 2025. When you have to invent machines to detect the lies your communication network produces, the network is no longer a straightforward engine of openness.

Kurzweil corrected himself. The most telling evidence is in his own 2024 book. In The Singularity Is Nearer, he retells the fall of the Iron Curtain — but the internet is gone from the story. Now it reads: “the proliferation of communication technology outside the Iron Curtain, from Beatles LPs to color TVs, stirred discontent against the governments that suppressed it.” That is historically correct — the Berlin Wall fell in 1989, two years before the Web existed and a decade before most people got online — but it is a different claim than the one in the 2005 book, which leaned on “the growth of the Internet and related technologies.” He also adds a caveat that would have been unthinkable in his earlier optimism: that the coming decades depend on “minimizing [AI’s] potential to be abused for authoritarian surveillance or to spread disinformation.” The forecaster updated. Fair enough — but the update concedes the point.

The reach bet. Kurzweil predicted that “as the Web reaches deeply into less developed regions, it will drive major strides in the availability of high-quality education and medical knowledge worldwide” (ch. “Promise and Peril of GNR”). Here he looks good on access. The ITU counted 5.5 billion people online in 2024 — 68 percent of humanity, up from essentially nothing in 2005 — with most of the recent growth coming from cheap smartphones in the developing world. Mobile health has quietly delivered: a widely cited review of mobile-health programs in low- and middle-income countries found a positive impact on chronic-disease outcomes, exactly the kind of leapfrogging Kurzweil imagined.

The qualifier “high-quality” is where it frays. A third of humanity — 2.6 billion people — is still offline, and the gap is where the need is greatest: only 27 percent of people in low-income countries are online, and just 38 percent across Africa. And the knowledge that did arrive came bundled with the same misinformation firehose degrading the rich world’s information diet. Access: delivered. Quality and equity: partial.

The moratorium prediction. This is Kurzweil’s best call in the batch, and it is barely remembered as a prediction. He argued that “well-defined and targeted temporary moratoriums, like the 1975 genetics moratorium, may be needed from time to time, though broad relinquishment is counterproductive” (ch. “Promise and Peril of GNR”). That is almost exactly what happened, twice. After a Chinese researcher edited the genomes of two babies in 2018, a group of scientists across seven countries — including two of the original Asilomar organizers — called in Nature for a global moratorium on heritable human genome editing. Then in 2023, more than 30,000 people signed the “Pause Giant AI Experiments” letter, which explicitly invoked the Asilomar precedent and called for a six-month halt on training the largest models. Kurzweil himself attended the 2017 Asilomar conference on beneficial AI and signed its principles.

The catch is that neither brake held. The AI labs did not pause; the arms race accelerated. Germline editing is discouraged but not universally banned. Kurzweil predicted the reflex correctly — society does reach for targeted, temporary, Asilomar-style moratoriums — and also predicted, correctly, that broad relinquishment would not take. What he got right and wrong here is the same thing: these moratoriums get called, and they rarely bind.

The body-modification bet. Kurzweil predicted that body and identity transformations would “occur through many small steps and gain mainstream acceptance rapidly, as happened with in vitro fertilization” (ch. “on Work”). This one is running ahead of his timeline. GLP-1 drugs went from a niche diabetes treatment to something roughly one in eight American adults report taking, with 91 percent awareness, in about three years — a faster mainstreaming of elective body modification than IVF managed in decades. And the sharper edge of the prediction arrived on schedule: in December 2023 the FDA approved Casgevy, the first CRISPR gene-editing therapy, now cleared in eight-plus countries. Editing a person’s own genome to cure a disease is now a medical product, not a thought experiment.

The scorecard

What Kurzweil missed (and what he nailed)

A pattern runs through this batch, and it is the same pattern that shows up across the whole scorecard project. Kurzweil is extraordinary at forecasting capability — how fast a technology gets cheap, small, and universal. He predicted that everyone would carry a supercomputer, and everyone does. He is far weaker at forecasting direction — what people point that capability at once they have it.

The democratization thesis is the cleanest example. He was right that communication technology would become universal and decentralized. He was wrong that universal, decentralized communication necessarily favors freedom. The same network that lets a dissident organize lets a state surveil, lets a troll farm flood, lets a synthetic video erase the line between what happened and what didn’t. Kurzweil mistook a neutral tool for a directional force. His 2024 edits — dropping the internet from the Iron Curtain story, adding the surveillance caveat — are the sound of a careful forecaster meeting that evidence halfway.

Where he was strongest was the least glamorous prediction: that society would keep reaching for narrow, temporary brakes rather than blanket bans. That is a claim about human institutional behavior, not about a technology curve, and it has aged better than any of his network optimism. The lesson for anyone forecasting technology is uncomfortable: the exponential curves are the easy part. The hard part is guessing what a species will do once it can do anything.

Method note

Trend claims here — the rise of disinformation research, the appearance of deepfake-detection filings — come from year-by-year counts across a corpus of roughly 9 million US patents and several hundred million scientific papers, verified against the underlying records. Democracy figures are from the V-Dem Institute’s 2026 report; connectivity figures from the International Telecommunication Union’s 2024 statistics; adoption and approval figures from public surveys and regulators. Kurzweil’s predictions are quoted from The Singularity Is Near (2005), his restatements from The Singularity Is Nearer (2024).