Loading
What changed between revision 2 and revision 4, compared over the article's markdown source with its frontmatter — so a change to status, evidence or reversibility shows up as the edit it is.
Older · revision 2
Import
Re-import of content/ray-kurzweil.md — the file changed after the initial import. Not an edit.
9.9 kB
Revision 2
Revision 4
lines 7–13 → 7–136 unchanged lines not shown
categories: ["people", "minds"]categories: ["people", "minds"]tags: ["singularity", "forecasting", "artificial intelligence", "futurism", "life extension"]tags: ["singularity", "forecasting", "artificial intelligence", "futurism", "life extension"]summary: "American inventor and futurist who formulated the law of accelerating returns and forecasts human-level machine intelligence by 2029 and a technological singularity by 2045."summary: "American inventor and futurist who formulated the law of accelerating returns and forecasts human-level machine intelligence by 2029 and a technological singularity by 2045."updated: "2026-08-23"updated: "2026-10-29"------ ```infobox```infoboxlines 25–33 → 25–3311 unchanged lines not shown
}}`````` **Ray Kurzweil** is an American inventor, computer scientist and futurist whose reading of**Ray Kurzweil** is an American inventor, computer scientist and futurist. His reading oftechnological history — that information technologies improve exponentially rather than linearly,technological history holds that information technologies improve exponentially rather thanand that the exponent itself is stable across substrates — underwrites his forecast of alinearly, and that the exponent itself is stable across substrates; on that basis he forecasts a[[technological-singularity]] around 2045. He built a substantial engineering career in pattern[[technological-singularity]] around 2045. He built a substantial engineering career in patternrecognition before becoming the most widely read forecaster of machine intelligence, and he is arecognition before becoming the most widely read forecaster of machine intelligence, and he is aprominent proponent of radical life extension.prominent proponent of radical life extension.lines 35–41 → 35–411 unchanged line not shown
## Career## Career Kurzweil built a pattern-recognition program that matched high-school students to colleges whileKurzweil built a pattern-recognition program that matched high-school students to colleges whilestill in high school, and studied at MIT under Marvin Minsky. His inventions cluster around readingstill in high school, and studied at MIT under [Marvin Minsky](https://aiwiki.ai/wiki/marvin_minsky). His inventions cluster around readingand hearing. In the mid-1970s he combined the first omni-font optical character recognition with aand hearing. In the mid-1970s he combined the first omni-font optical character recognition with aflatbed scanner and a speech synthesizer to produce the Kurzweil Reading Machine, a print-to-speechflatbed scanner and a speech synthesizer to produce the Kurzweil Reading Machine, a print-to-speechdevice for blind readers developed with the National Federation of the Blind. He later foundeddevice for blind readers developed with the National Federation of the Blind. He later foundedlines 56–63 → 56–6314 unchanged lines not shown
exponential improvement itself increases.[^lar2001] He supports this with time-series plots ofexponential improvement itself increases.[^lar2001] He supports this with time-series plots ofcomputing price-performance running back through electromechanical relays, vacuum tubes andcomputing price-performance running back through electromechanical relays, vacuum tubes andtransistors, arguing that Moore's law is the fifth in a sequence of substrate-specific curvestransistors, arguing that Moore's law is the fifth in a sequence of substrate-specific curvesrather than a one-off. The generalized claim is that when one paradigm saturates, another takesrather than a one-off. In the generalized form of the claim, when one paradigm saturates anotherover, so the composite curve keeps its slope. See [[accelerating-change]] for the broader debatetakes over, so the composite curve keeps its slope. See [[accelerating-change]] for the broader debateover this style of argument.over this style of argument. ```keyfacts```keyfactslines 71–78 → 71–787 unchanged lines not shown
From the curve he derives specific dates. Sufficient compute to emulate a human brain arrives atFrom the curve he derives specific dates. Sufficient compute to emulate a human brain arrives atconsumer prices in the late 2020s; artificial intelligence passes a properly administered Turingconsumer prices in the late 2020s; artificial intelligence passes a properly administered Turingtest by 2029; and by 2045 machine intelligence so exceeds the biological kind that projection pasttest by 2029; and by 2045 machine intelligence so exceeds the biological kind that projection pastthat point fails. He also treats the merger of human and machine cognition as the expected outcomethat point fails. He expects human and machine cognition to merge rather than for machines torather than displacement, an argument developed in [[human-ai-merger]].displace people, an argument developed in [[human-ai-merger]]. ## Longevity and the three bridges## Longevity and the three bridges lines 80–92 → 80–901 unchanged line not shown
Kurzweil was diagnosed with type 2 diabetes as a young man. With the physician Terry Grossman heKurzweil was diagnosed with type 2 diabetes as a young man. With the physician Terry Grossman hewrote *Fantastic Voyage* (2004) and *Transcend* (2009), setting out a "three bridges" argument.wrote *Fantastic Voyage* (2004) and *Transcend* (2009), setting out a "three bridges" argument.Bridge one is aggressive present-day management of health through diet, exercise andBridge one is aggressive present-day management of health through diet, exercise and[[dietary-supplements|supplements]] —[[dietary-supplements|supplements]]; Kurzweil has reported taking a very large number of pillsKurzweil has reported taking a very large number of pills daily, a figure he has said fell overdaily, a figure he has said fell over time. Bridge two is biotechnology capable of reprogramming the body's information processes, thetime. Bridge two is biotechnology capable of reprogramming the body's information processes, thedomain of [[gene-therapy-for-aging]] and [[epigenetic-reprogramming]]. Bridge three isdomain of [[gene-therapy-for-aging]] and [[epigenetic-reprogramming]]. Bridge three is[[medical-nanorobots]] operating inside cells. The argument is that a person need only survive on[[medical-nanorobots]] operating inside cells. A person need only survive on each bridge longeach bridge long enough to reach the next, which is the same structure asenough to reach the next, the same structure as [[longevity-escape-velocity]]. He has also said he expects to be cryopreserved if he dies before[[longevity-escape-velocity]]. He has also said he expects to be cryopreserved if he dies beforethe bridges arrive; see [[cryonics]].the bridges arrive; see [[cryonics]]. > [!caution] Bridge one has thin support> [!caution] Bridge one has thin supportlines 99–110 → 97–1086 unchanged lines not shown
Kurzweil has published self-assessments of his own forecasts, scoring the large majority of theKurzweil has published self-assessments of his own forecasts, scoring the large majority of the147 predictions he made in 1999 for 2009 as correct or "essentially correct".[^rk2010] Independent147 predictions he made in 1999 for 2009 as correct or "essentially correct".[^rk2010] Independentreviewers have been harsher, noting thatreviewers have been harsher. They note that many predictions were phrased loosely enough to admitmany predictions were phrased loosely enough to admit generous grading, and that his most confidentgenerous grading, and that his most confident near-term calls (ubiquitous speech interfaces,near-term calls — ubiquitous speech interfaces, self-driving cars, translating telephones — arrivedself-driving cars, translating telephones) arrived years to decades later than stated, or inyears to decades later than stated, or in narrower form.narrower form. The strongest structural criticism is that curve-fitting is not a mechanism. Paul Allen and MarkThe strongest structural criticism is that fitting a curve does not supply a mechanism. Paul Allen and MarkGreaves argued that software understanding of biology and cognition faces a "complexity brake" thatGreaves argued that software understanding of biology and cognition faces a "complexity brake" thatno hardware curve resolves.[^allen2011] Theodore Modis, whose own work Kurzweil cites, hasno hardware curve resolves.[^allen2011] Theodore Modis, whose own work Kurzweil cites, hasdisputed the data selection behind the composite exponential.[^modis2006] Douglas Hofstadter'sdisputed the data selection behind the composite exponential.[^modis2006] Douglas Hofstadter'slines 113–140 → 111–1382 unchanged lines not shown
Large language models have complicated both sides. Systems released since 2022 achievedLarge language models have complicated both sides. Systems released since 2022 achievedconversational fluency and coding ability on roughly the schedule Kurzweil's compute curve implied,conversational fluency and coding ability on roughly the schedule Kurzweil's compute curve implied,which his supporters treat as confirmation. Critics observe that the same systems fail atand his supporters treat this as confirmation. Critics observe that the same systems fail atpersistent memory, reliable reasoning and grounded world models in ways the brain-emulation framingpersistent memory, reliable reasoning and grounded world models in ways the brain-emulation framingdid not anticipate, and that the connection between his compute extrapolation and any specificdid not anticipate, and that the connection between his compute extrapolation and any specificcapability was always asserted rather than derived. Kurzweil's own dates have not moved: *Thecapability was always asserted rather than derived. Kurzweil has not moved his dates: *TheSingularity Is Nearer* restates 2029 and 2045 unchanged.[^tsin2024] SeeSingularity Is Nearer* restates 2029 and 2045 unchanged.[^tsin2024] The dispute also turns on what[[artificial-general-intelligence]] for the definitional fight this exposes.counts as general intelligence; see [[artificial-general-intelligence]]. ## Reception and legacy## Reception and legacy Kurzweil's practical legacy is assistive technology: reading machines and speech systems thatKurzweil's practical legacy is assistive technology. His reading machines and speech systemschanged what blind and print-disabled people could do, decades before the forecasting became hischanged what blind and print-disabled people could do, decades before forecasting became his publicpublic identity. His intellectual legacy is more contested. He gave [[singularitarianism]] itsidentity. His intellectual legacy is more contested. He gave [[singularitarianism]] itspopular form and its dates, and *The Singularity Is Near* did more than any other book to move thepopular form and its dates, and *The Singularity Is Near* did more than any other book to move theidea from science fiction and the [[extropianism|extropian]] mailing lists into mainstreamidea from science fiction and the [[extropianism|extropian]] mailing lists into mainstreamdiscussion. Nearly every subsequent treatment of machine superintelligence, including work bydiscussion. Nearly every subsequent treatment of machine superintelligence, including work by[[nick-bostrom]] that reaches very different conclusions about risk, positions itself relative to[[nick-bostrom]] that reaches very different conclusions about risk, positions itself relative tohim.him. The unresolved question is whether his method has any predictive content beyond the hardware curveIt remains unresolved whether his method has any predictive content beyond the hardware curve itit started from. Compute has behaved close to the way he said it would. Whether that fact licensesstarted from. Compute has behaved close to the way he said it would. His critics deny that thisthe claim about 2045 — or about anything downstream of hardware at all — is exactly what hislicenses the claim about 2045, or any claim about what lies downstream of hardware, and no resultcritics deny, and no result since 2022 has settled it.since 2022 has settled the question. ## See also## See also lines 149–155 → 147–1538 unchanged lines not shown
## References## References [^lar2001]: `statement` Kurzweil, R. "The Law of Accelerating Returns." Essay, 2001. {Kurzweil's own statement of the argument; the historical series in it are his selection and his normalisation, which is precisely what critics dispute.}[^lar2001]: `statement` Kurzweil, R. "The Law of Accelerating Returns." Essay, 2001. {Kurzweil's own statement of the argument; the historical series in it are his selection and his normalisation, which is what critics dispute.}[^allen2011]: `news` Allen, P. and Greaves, M. "The Singularity Isn't Near." *MIT Technology Review*, 2011. {An opinion essay by two software figures in a technology magazine, not a research result; its claim is about software understanding rather than about the hardware curve.}[^allen2011]: `news` Allen, P. and Greaves, M. "The Singularity Isn't Near." *MIT Technology Review*, 2011. {An opinion essay by two software figures in a technology magazine, not a research result; its claim is about software understanding rather than about the hardware curve.}[^modis2006]: `paper` Modis, T. "The Singularity Myth." *Technological Forecasting & Social Change*, 2006.[^modis2006]: `paper` Modis, T. "The Singularity Myth." *Technological Forecasting & Social Change*, 2006.[^rk2010]: `statement` Kurzweil, R. "How My Predictions Are Faring." Essay, October 2010. {Kurzweil's own review of the 147 predictions in The Age of Spiritual Machines; he grades 86% of them entirely or essentially correct, a figure independent reviewers dispute.}[^rk2010]: `statement` Kurzweil, R. "How My Predictions Are Faring." Essay, October 2010. {Kurzweil's own review of the 147 predictions in The Age of Spiritual Machines; he grades 86% of them entirely or essentially correct, a figure independent reviewers dispute.}removed, struck through added, underlinedLine numbers count the serialised markdown of each revision, frontmatter included.
127 lines are the same in both revisions, and 60 of them are not shown. Neither revision here is necessarily the current article — read Ray Kurzweil for the text this wiki stands behind today.