Skip to content

Ray Kurzweil

Abstract

Ray Kurzweil built real machines before he built a philosophy. In the 1970s he combined optical character recognition, a flatbed scanner, and text-to-speech into the Kurzweil Reading Machine, the first device that could read any printed page aloud to a blind person; Stevie Wonder was among the first buyers and later pushed him to build a synthesizer that could reproduce real instruments. Those inventions worked and shipped. Since the late 1990s Kurzweil has been better known for a set of predictions: that technology improves exponentially, that machines will match human intelligence around 2029, and that a Singularity around 2045 will let people merge with computers and stop dying. He grades his own forecasts as roughly 86% accurate; his critics call the record vaguer and the vision closer to religion than science. Both the engineer and the prophet are worth taking seriously, and they are not the same person.

The Machines That Read and Sang

Kurzweil was born in Queens on 12 February 1948 and finished MIT in 1970 with a degree spanning computer science and literature, having studied under Marvin Minsky. His first company solved a problem that had defeated earlier scanners: omni-font OCR, software that could read text set in any typeface rather than one it had been trained on.

By itself that was a pattern-recognition trick. Kurzweil paired it with two other pieces he needed to build, a CCD flatbed scanner to capture the page and a text-to-speech synthesizer to voice it, and in 1976 announced the Kurzweil Reading Machine: put a book face-down on the glass, and the machine read it aloud. For blind readers it was the first device that could handle ordinary printed material, not just specially prepared braille or recordings (see The Accessibility Revolution in Computing). Stevie Wonder bought one of the first units and became a friend.

That friendship produced the second invention. Wonder complained that electronic keyboards could not reproduce the sound of a real piano or orchestra, and Kurzweil Music Systems built the K250, released in 1984, a synthesizer that used recorded samples of real instruments; in listening tests Kurzweil ran, musicians could not reliably tell its piano from a grand piano, though some later reviewers questioned how hard the test was. Wonder was the company’s musical adviser. The K250 became a fixture in studios and connects Kurzweil to the wider history of digital instruments (see Computing and Music Creation).

Speech, and a Buyer That Imploded

Kurzweil’s third company, Kurzweil Applied Intelligence, worked on large-vocabulary speech recognition, aiming at doctors and lawyers who wanted to dictate rather than type (see Speech Recognition). He sold it in the late 1990s to the Belgian firm Lernout & Hauspie, then a rising star of speech technology, which collapsed in 2001 in one of Europe’s largest accounting-fraud scandals. The technology survived the wreckage; the episode is a footnote in Kurzweil’s story but a reminder that his inventions have outlived several of the companies that bought them.

The Law of Accelerating Returns

From the 1990s Kurzweil turned from building machines to forecasting them. His central claim, the Law of Accelerating Returns, extends the logic of Moore’s Law from chips to technology in general: progress is exponential, not linear, because each generation of tools speeds the making of the next. Plotted that way, computing power heads for and then past the capacity of a human brain, and Kurzweil reads specific dates off the curve.

He set out the argument in a run of bestsellers: The Age of Spiritual Machines (1999), The Singularity Is Near (2005), How to Create a Mind (2012), and The Singularity Is Nearer (2024). The predictions are concrete. A computer passes the Turing test by 2029. Around 2045 comes the Singularity, the point at which machine intelligence so exceeds human intelligence, and merges so completely with it through brain-computer interfaces, that the future beyond it cannot be seen from here. People will back up their minds and, in Kurzweil’s telling, effectively stop dying. In 2008 he co-founded Singularity University with Peter Diamandis to teach these ideas to executives, and in December 2012 he joined Google as a director of engineering working on language understanding, which put the field’s most prominent futurist inside one of the companies building the technology (see The LLM Race).

The Prediction Business

Kurzweil keeps score on himself. In 2010 he reviewed 147 predictions from his books and rated 115 “entirely correct” and another 12 “essentially correct,” for a claimed 86% accuracy. Some hits are real: writing in 1990, he foresaw a computer beating the world chess champion by 2000, and IBM’s Deep Blue did it in 1997, along with the growth of the internet and wireless access.

The critics attack the scoring, not just the misses. Predictions graded “essentially” or “partially” correct have no objective standard, and a forecaster who marks his own work can read vague statements generously. The misses were real too: reviewers of his 2009 predictions noted that continuous speech recognition had not become the main way people create text, that self-driving cars guided by highway sensors had not arrived, and that the economy had done the opposite of boom. His 2029 Turing-test claim is the subject of a public $20,000 wager with Lotus founder Mitchell Kapor, recorded on the Long Bets site and still unresolved.

The deeper objection is about what kind of claim the Singularity is. Kapor called it “intelligent design for the IQ 140 people,” driven at bottom by “a religious impulse,” a promise of resurrection and immortality dressed in graphs. The cognitive scientist Douglas Hofstadter, who does not dismiss Kurzweil as a crank, described the mix as “rubbish and good ideas” so intertwined that “it’s very hard to disentangle the two, because these are smart people.” The pattern the critics point to is that the transcendent parts of the forecast, superhuman minds, uploaded consciousness, the end of death, always sit about two decades out, near enough to matter and far enough to stay untested.

Betting on His Own Body

Kurzweil means the immortality literally, and he has organized his life around reaching it. He has written that he takes a large daily regimen of dietary supplements, at its reported peak around 250 pills, on the theory that he needs to keep his body running until technology can extend life indefinitely, a strategy he summarizes as living long enough to live forever. His father Fredric, a conductor and pianist, died of a heart attack in 1970, when Kurzweil was 22, and he has kept boxes of the man’s papers with the stated hope of one day training an AI on them to reconstruct a version of him.

It is easy to read all this as the vanity of a wealthy futurist, and some of it is. It is also the same move he made with the Reading Machine and the synthesizer: identify a limit, refuse to accept it as permanent, and try to engineer past it. The difference is that those machines were finished and testable, and death is not yet either. Kurzweil’s record is genuinely split. He built tools that gave blind people the printed word and musicians a new instrument, and he sells a dated prophecy of digital immortality that its critics find closer to faith than forecast. Judging the inventor by the prophet, or the prophet by the inventor, gets him wrong both ways.

πŸ“š Sources