A Letter from Ray Kurzweil

To the Editor, IEEE Spectrum

Re: "Ray Kurzweil's Slippery Futurism," by John Rennie

While I appreciate some of the things John Rennie has to say, his review of my predictions is filled with inaccuracies, including misquotes of mine, and misunderstandings of the meaning of my words and the reality of today's technology. For starters, he takes note of my point about selection bias, but his entire article suffers from this bias. While he acknowledges that I wrote over 100 predictions for 2009, in a book I wrote in the late 1990s, he only talks about a handful of them. And he persistently gets these wrong. He writes that I predicted "widespread, foolproof, real-time speech translation." We do in fact have real-time speech translation in the form of popular phone apps. But who ever said anything about "foolproof?" Rennie just made that up like a lot of the factoids in this article. Not even human translators are foolproof. Apparently that has now been removed from the online version.

Rennie repeatedly chides me for taking an overly "lawyerly" interpretation of my predictions. If by that he means paying attention to what the words actually mean, that is true (and as I point out below, this is not always to my benefit). For example, there is a difference between predicting that something will exist versus saying that it will be commonly used. He criticizes my prediction that most computers would disappear from being seen, when in fact at least 98 percent of all computers we currently use in and around the home are embedded in everyday products, such as appliances and cars, which typically have 30 to 100 computers in them. But to Rennie, these are not really computers because, well, you can't see them. So by that reasoning, the prediction couldn't possibly ever be true, because as soon as a computer disappears from view, it's no longer a computer (in the way he thinks of them).

He then goes on to say that not only is this prediction wrong, but it was also obvious, because crude harbingers of these devices existed earlier, and he cites failed products such as early tablet computers. That's the first time someone has called one of my predictions both wrong and obvious at the same time. The reality is that forerunners always exist; for example, the Arpanet in the early 1980s enabled me to predict that a vast web of communications would emerge by the late 1990s. The essence of my predictions is that I project these early crude examples into the future by considering the exponential growth of information technology (what I call the "law of accelerating returns"), rather than the linear extrapolation, which represents most people's intuition.

Rennie says that my predictions "border on the unfalsifiable" and that I have "not admitted that [I] was wrong." Neither observation is correct. Getting back to my "lawyerly" interpretations, I predicted that driverless cars would be common (but not for local roads) by 2009. Indeed Google's AI-based driverless cars have already driven 140,000 miles in California, including through cities and towns, without human intervention. Nonetheless, I rated the prediction as wrong because the technology is not yet commonly used. This is ironic, because I've received substantial feedback that my prediction was prescient, since far more progress has been made on this goal than people expected. But given the proper meaning of the words in my original prediction, I nonetheless consider it incorrect.

The most important predictions I have made relate to the remarkably smooth exponential trajectory of the basic measures of information technology, such as MIPS per constant dollar, which goes back to the electromagnetic calculators used in the 1890 American census. I noticed this about 30 years ago, wrote about it in the mid-1980s, had a graph through 1998 in The Age of Spiritual Machines, then updated it through 2002 in The Singularity Is Near and have recently updated it to 2008. This amazingly smooth exponential trend continues unabated. This graph is not only about the past; it is inherently a prediction and has now held up for decades since I first noticed it. With regard to computation, the phenomenon started decades before Gordon Moore was even born and applies to dozens of basic measurements of information technology (for example, genetic sequencing), not just computation.

John Rennie alludes to an essay I wrote evaluating my predictions, including those in The Age of Spiritual Machines, but he never bothered to ask me for it, as I would have been happy to send a draft to him. As it is, he wrote his article in a vacuum. Here it is: www.kurzweilai.net/predictions.php.

Sincerely,

Ray Kurzweil

Editor's comment, 30 December 2010: Mr. Kurzweil's objection to John Rennie's critique begins with, and makes much of, the use of the word "foolproof." In fact, that word was never used by Mr. Rennie. In an editing error, the word "foolproof" was inserted into a sentence in Mr. Rennie's article, after Mr. Rennie had reviewed the editing. IEEE Spectrum regrets this lapse.