Kurzweil Scorecard: Work, Cities, and the Virtual Century

Twenty years ago Ray Kurzweil sketched a social order organized around virtual presence. Centralized offices and cities would fade. Personal services — health, fitness, education — would “largely move” into simulated environments. Intellectual property law would strain under products made of information. By mid-century, a Nonbiological Persons Act would grant legal standing to minds that had never been biological. Batch 64 of our prediction index groups ten of those social claims together. Some have aged well. A few have aged strangely. One is being decided in a Manhattan courtroom.

The predictions

Kurzweil made most of these in the work and learning chapters of The Singularity Is Near (2005), with a separate cluster about nonbiological personhood drawn from his chapter on the human brain. Treat them as a single social package: what happens to employment, workplaces, services, property, and personhood when the cost of simulated experience collapses.

The ten predictions in this batch, with their stated timeframes:

• Farm and factory employment under 3 percent each in the US (circa 2005)
• Office buildings and cities rendered obsolete by virtual presence (2040s)
• Personal services — health, fitness, education — largely move into VR (2030s)
• Open-source and proprietary versions of every product/service coexist (2020s)
• IP protection becomes critical as product value migrates into information (2020s)
• Virtual prostitution legalized (2033)
• A Nonbiological Persons Act passed (2052)
• MIT’s OpenCourseWare reaches 900 courses, planned full coverage by 2007 (a 2005 factual claim)
• Virtual laboratories and interactive historical personas populate classrooms (2010s)
• The transition to nonbiological experience arrives as a gradual slippery slope rather than a one-time brain upload (long-term)

Where we actually are

On farm and factory employment: Kurzweil wrote that “both farm employment and factory employment in the United States had fallen from about 30 percent each a century earlier to under 3 percent each” (ch. “on the Human Body”). In The Singularity Is Nearer, he restated the farm figure: “farmwork has gone from constituting 80 percent of all labor in the United States in 1810 to 40 percent in 1900 to less than 1.4 percent today.” The claim held on farms. Factories are stickier — manufacturing now runs near 8.5 percent of US workers (about 12.7 million people as of late 2025), up from a 2020 low thanks to CHIPS Act and IRA spending. Direction right, factory magnitude wrong.

On offices and cities: the closest thing in the batch to a bullseye. Kurzweil wrote that “the ability to do nearly anything with anyone from anywhere in virtual reality will make centralized technologies such as office buildings and cities obsolete” (ch. “on Work”). In Nearer he tracked what happened: “At the peak during the pandemic, up to 42 percent of Americans were working from home… the old model of nine-to-five sitting at a desk in a company office has been obsolete for years, but inertia and familiarity made it hard for society to change until the pandemic forced us to.” US office vacancy hit a record 20.7 percent in Q2 2025. San Francisco sits at 27.7 percent, up from 8.6 percent pre-pandemic. Downtown Seattle hit 35.6 percent in Q4 2025. Cities aren’t obsolete, but their economic function has been partly disintermediated. The trigger was a pandemic, not VR. A win with a footnote on mechanism.

On personal services moving into VR: this is the cleanest miss so far. Kurzweil predicted that health, fitness, and education would “largely move to virtual-reality environments” in the 2030s. The XR market is pointing the wrong way. Apple sold roughly 45,000 Vision Pro units in Q4 2025, with Vision Pro digital ad spend down more than 95 percent year-over-year in every major market. Meta shipped 1.7 million Quests in the first three quarters of 2025, and the global VR headset market shrank 14 percent year-over-year. Patent filings on VR collaboration rose from 3 in 2018 to 22 in 2025 — a trickle. The literature is more interesting: papers on VR in medical training climbed from 24 in 2020 to 169 in 2025. A 2022 Frontiers in Medicine scoping review on virtual simulation in undergraduate medical education has 134 citations; its companion thread on digital twins in personalized medicine has 133. VR is colonizing a professional niche — surgical training, rehabilitation — rather than swallowing the service economy. Largely move? No. Move into discrete high-value corners? Yes.

On open-source and proprietary coexistence: this prediction deserves a trophy. Kurzweil restated it in Nearer: “the coexistence of an open-source market… with the proprietary market will become the defining nature of the economy in an increasing number of areas.” Written in 2005 about music, books, and 3D-printed designs. In 2025, a McKinsey survey found 63 percent of organizations use open-source AI models. Meta’s Llama series has made open weights the default substrate underneath proprietary wrappers. This is exactly the pattern he described, now playing out one level up the stack — not just products but the underlying cognitive engines. Call this ahead of schedule on the mechanism he actually anticipated.

On IP protection becoming critical: also a win, visible in the court docket. The New York Times v. OpenAI and Microsoft survived a motion to dismiss in April 2025; the Times is now demanding 20 million private ChatGPT conversations to trace paywall breaches. Getty Images v. Stability AI produced its first major judgment in England in November 2025 — a mostly-for-Stability ruling that nevertheless landed a trademark finding. Over 50 AI copyright cases are pending. US 12,547,683, granted February 2026, claims a zero-knowledge proof system for asserting ownership of a neural network without revealing the underlying watermark key. US 12,548,107, granted the same month, trains an adversarial model that embeds imperceptible watermarks into video frames and survives downstream distortions. Kurzweil’s 2005 anxiety about IP has been translated into cryptographic machinery. Ahead of schedule.

On virtual prostitution legalized by 2033: the world moved the opposite way on AI-generated likenesses. The TAKE IT DOWN Act, signed into federal law in May 2025, criminalizes non-consensual intimate imagery including deepfakes. The NO FAKES Act, introduced April 2025, would make unauthorized AI replicas of a person’s voice or likeness actionable. Simultaneously, AI companion apps have become mainstream, and a literature on parasocial bonds with chatbots is forming (a 2022 Journal of Consumer Behaviour study on chatbot continuance intentions has 85 citations). Kurzweil anticipated the market but not the regulatory counter-move. Wrong mechanism, right instinct.

On a Nonbiological Persons Act (2052): this one is live in the wrong direction. In 2025, Idaho and Utah passed laws explicitly declaring AI not a legal person. The UK Law Commission floated the question and stopped short of recommending personhood. The policy drift is toward treating AI as a tool attributable to its operator, not a rights-bearing entity. 2052 is still 26 years off, so this is too early to call — but the early trendline points away from Kurzweil’s prediction.

On MIT OCW’s course count: in 2005 it offered 900 courses with full coverage planned by 2007. As of 2025, OCW covers more than 2,500 courses across undergraduate and graduate curricula. Trivially ahead of schedule — the world moved past the specific benchmark.

On virtual laboratories and historical personas in the 2010s: VR labs in chemistry and physics exist but remain supplementary, not default. “Talking to Thomas Jefferson” as an educational exercise is now trivially possible with any large language model — but via text, not VR. Wrong mechanism, with the payload arriving a decade late via a different delivery system.

On gradual nonbiological transition: untestable on a short horizon, but the slope is accumulating: copilots, agents, memory-augmented chat. The claim that it arrives as a slippery slope rather than a single upload event looks right in shape so far.

The scorecard

What Kurzweil nailed and what he missed

A pattern jumps out of this batch. Kurzweil was consistently right about direction and wrong about delivery. Offices did empty — through a virus, not VR. Historical figures can be interrogated by students — through language models, not immersive simulations. IP law is buckling under information-as-product — through generative AI training data, not 3D-printed designer shoes. Open weights and proprietary wrappers now coexist — in AI, not 3D printing. He kept betting on VR as the universal solvent, and it kept not being VR.

The second pattern is more subtle. On the two predictions where Kurzweil was most specific about dates — virtual prostitution in 2033, Nonbiological Persons Act in 2052 — society is moving to defensively block the outcomes he anticipated rather than drift into them. The TAKE IT DOWN Act and state-level AI personhood denials are both, in effect, legislative speed bumps placed on his roadmap. It turns out the interesting question isn’t whether a technology becomes possible, but whether a society decides to permit it. That is a dimension of his forecasting framework that the 2005 book underweights.

Method note

Evidence came from three streams: a US patent corpus of 9.3 million grants and pre-grants (searched for VR collaboration, watermarking, and neural network ownership; recent grants read in full), a literature corpus of 357 million scholarly works (filtered by citation count for VR-in-medicine and chatbot-companion threads), and contemporary reporting on employment, office vacancy, XR headset shipments, AI copyright litigation, and AI personhood legislation.