Kurzweil Scorecard: Regulation, Stem Cells, and the Stones-in-a-Stream Hypothesis

The most durably correct thing Ray Kurzweil wrote in 2005 was probably the smallest bet on the page. Buried in a rebuttal chapter — sandwiched between thought experiments about nanobots and totalitarian states — he predicted that the Bush-era embryonic-stem-cell restrictions would not significantly slow cell-therapy research, and that scientists would simply route around the rules. A year after the book went to print, a Kyoto lab did exactly that. The paper now has 26,162 citations.

Batch 98 is four linked claims about whether regulation can slow technology. One is vindicated with almost unsettling precision. One is broadly right. And two are getting a quiet challenge from 2025 that Kurzweil didn’t see coming: the question isn’t whether regulation can stop a technology, it’s whether regulation can move it — to a different country, a different mechanism, or a different company.

The predictions

Kurzweil made these claims in The Singularity Is Near (2005) during a rebuttal of critics like Leon Kass and Leon Fuerth, who argued that government could and should slow emerging biotech. His position, condensed:

• “Despite stem-cell restrictions, cell-therapy research and the broader biotechnology field have not been affected to a significant degree” (ch. “The Criticism from the Likelihood of Government Regulation”).
• “Restrictions on embryonic stem-cell research have accelerated alternative approaches such as transdifferentiation, converting one cell type such as skin into other cell types” (ibid.).
• “Government regulation has had little measurable effect on the accelerating technological trends” (ch. “A Panoply of Criticisms”).
• “Absent a worldwide totalitarian state, regulation will not stop the acceleration of technology because economic and other forces will route around obstacles” (ibid.).

In The Singularity Is Nearer (2024), he restated the thesis with the benefit of knowing about Yamanaka: “With induced pluripotent stem (iPS) cells, we are gaining the capability to rejuvenate the heart after a heart attack and overcome the ‘low ejection fraction’… In effect, they may be tricking the heart into thinking it is in a fetal environment. This procedure is being used for a broad variety of biological tissues.”

Where we actually are

The iPSC timeline is almost a textbook case of “stones in a stream.”

George W. Bush’s August 2001 policy limited federal funding to a small number of pre-existing embryonic-stem-cell lines. Two workarounds emerged within three years. First, in 2004, Doug Melton and David Scadden founded the Harvard Stem Cell Institute with private Howard Hughes Medical Institute money and a physically separate, Harvard-donated lab designed to keep every beaker clear of federal dollars. Melton used that private funding to derive seventeen new stem-cell lines and distribute them globally — free of charge. Second, in November 2004, California voters passed Proposition 71, authorizing a $3 billion state bond to fund stem-cell research and creating CIRM. Kurzweil named both of these in 2005 as evidence that regulation was being routed around in real time.

Then the bigger thing happened. In August 2006, Shinya Yamanaka and Kazutoshi Takahashi showed that four transcription factors — Oct3/4, Klf4, c-Myc, and Sox2 — could reprogram adult mouse fibroblasts into pluripotent stem cells, no embryo required. The 2006 Cell paper now has 26,162 citations by OpenAlex’s count. The 2008 human-fibroblast follow-up adds another 1,300+. Yamanaka shared the 2012 Nobel.

The patent record shows the commercial stampede. Kyoto University was granted US 8,058,065 in 2011, with the core claim reading like a field manual: “A method for preparing an induced pluripotent stem cell by nuclear reprogramming of a somatic cell from a mammalian species, comprising: a) introducing into the somatic cell one or more retroviral vectors comprising a gene encoding Oct3/4, a gene encoding Klf4, a gene encoding c-Myc and a gene encoding Sox2…” Four more Kyoto iPSC patents issued within eighteen months. US 8,048,675, assigned to iPierian, Inc., covers integration-free human iPS cells from blood — the same recipe without the viral integration that made regulators nervous.

By the numbers, the iPSC field looks like a normal exponential, not a regulated one:

For comparison, embryonic-stem-cell papers peaked at 74 in 2015 and have fallen to ~22/year. The field didn’t die; it bifurcated into iPSC, which now has ~130x the annual publication rate.

Cell therapy more broadly is going faster, not slower.

If Kurzweil’s 2005 claim was that stem-cell restrictions wouldn’t dent the broader biotech field, the CAR-T trajectory is the cleanest confirmation. Chimeric-antigen-receptor T-cell therapy literature went from a handful of papers in 2005 to 1,321 last year. ClinicalTrials.gov shows CAR-T trial starts going from two in 2010 to 215 in 2025 — and 71 already logged for 2026 despite only four months elapsed. The FDA has approved six CAR-T products. Commercial cell therapy is a real industry with real revenue; stem-cell restrictions never touched it because it didn’t need embryos.

California’s $3 billion bet on Prop 71 (extended by a further $5.5 billion via Prop 14 in 2020) has now funded 113 clinical trials and seeded more than 50 start-ups, generating an estimated $10.7 billion in California economic activity alone. The first CIRM-backed therapy — a gene therapy for severe combined immunodeficiency, developed at UCLA — cured its first patient in 2014.

The general “regulation is irrelevant” claim is where 2025 pushes back.

Two natural experiments since 2005 complicate the stronger version of Kurzweil’s thesis.

The first is CRISPR germline editing. After He Jiankui announced gene-edited twins in 2018, China criminalized clinical heritable genome editing in its 2020 Criminal Law amendment. A 2019 moratorium call led by Emmanuelle Charpentier, Eric Lander, and Feng Zhang was renewed in May 2025 for another ten years. Germline CRISPR in humans has effectively stopped — not for lack of capability, but because every jurisdiction that could permit it has chosen not to. That’s a measurable effect. Somatic CRISPR, meanwhile, continues at full speed; Casgevy was approved in 2023. The regulation sorted the field cleanly along the heritable/non-heritable line.

The second is the EU AI Act, which entered into force on August 1, 2024. Europe hosts none of the world’s ten largest AI companies. In 2024, the US saw nearly three times as many new AI companies founded and attracted roughly $29 billion in AI investment versus $1.5 billion in Europe — a ~20x gap. Compliance-cost estimates for a single AI system run €29,277 annually. Over thirty EU founders and VCs signed an open letter in late 2024 warning that the Act would “leave Europe behind.” As of November 2025, the EU was already moving to weaken the law. The acceleration Kurzweil predicted is happening — but it’s happening in Texas and Hangzhou, not Paris. Regulation didn’t stop the technology. It moved it.

The US chip export controls tell a similar story in the opposite direction. Advance notice of HBM restrictions let Huawei stockpile a year’s supply. TSMC, per a CSIS analysis, produced nearly three million Ascend AI dies for Huawei in 2023–2024 before the gap was closed. ByteDance, Alibaba, and Tencent collectively spent $16 billion stockpiling 1.3–1.6 million Nvidia H20 units before the ban tightened. DeepSeek’s V3 was trained on 2,048 Nvidia H800s that shouldn’t have been available, and the upcoming V4 may run partly on Huawei Ascend silicon. The controls slowed Chinese AI but did not stop it.

The scorecard

What Kurzweil got right (and what he missed)

The iPSC prediction is remarkable. Kurzweil wrote that restrictions would push research toward transdifferentiation — in print, in 2005 — at precisely the moment Yamanaka was running the experiments that would prove him right. Treat that as the upper bound on how prescient technology forecasting can be. If you had read The Singularity Is Near and shorted embryonic-stem-cell companies while going long on whatever reprogramming platforms existed in 2005, you would have looked like a genius.

But the stones-in-a-stream framing has a failure mode Kurzweil didn’t dwell on. Water does route around rocks — and it ends up in a different place. The EU AI Act hasn’t stopped frontier AI; it has exported it. The chip export controls haven’t stopped Chinese AI; they’ve midwifed an indigenous Huawei/DeepSeek stack that is suddenly a credible threat to Nvidia’s monopoly. The 2019–2025 CRISPR germline moratorium didn’t stop genome editing; it channeled the talent and capital into somatic therapies, which are now a commercial industry while the germline variant remains taboo.

Regulation, it turns out, is not a dam. It’s a nudge. It doesn’t reduce the total energy of a technological field by much. What it does is change the gradient — where the energy flows, who captures the rents, and which mechanisms mature first. For an R&D director reading a regulatory proposal in 2026, the right question isn’t “will this stop the technology?” The answer to that is almost always no, just as Kurzweil said. The right question is “which adjacent mechanism does this regulation accelerate, and am I positioned for it?” Yamanaka was the answer in 2006. DeepSeek may be the answer in 2026. The next one is already in someone’s private lab, running on whatever tool the current rules happen not to cover.

Method note

Evidence for this scorecard came from our mirror of ~9.3 million US patents, ~357 million OpenAlex scholarly works, and the full ClinicalTrials.gov dataset of 541,000 studies. Counts are by publication year or trial start year. Patent numbers, claim text, and paper titles were pulled and read directly. Policy context came from current web sources including CIRM’s public reporting, Nature, CSIS, the US Congressional Research Service, PMC-indexed journals, and contemporary news. Every number in this post traces back to a specific query run or URL fetched this session.