stewart brand on ray kurzweil

Ray Kurzweil gave a talk last night in SF, part of the Long Now Foundation's SALT series (Seminars About Long-term Thinking). He's one of the most visible (and credible) of the "singularitarians", adherents to the notion that accelerating technology is leading to machines with intelligence that will exceed human intelligence, perhaps as soon as the year 2030. He also has a new book out this week, The Singularity Is Near : When Humans Transcend Biology. Whether you think Kurzweil is brilliant or crazy, he is certainly an intellectual provocateur of the first order. Unfortunately I couldn't go, but Stewart Brand published excellent notes:

Attempts to think long term, Ray Kurzweil began, keep making the mistake of imagining that the pace of the future is like the pace of the past. Pondering the next ten years, we usually begin by studying the last ten years. He recommends studying the last twenty year for clues about the rate and degree of change coming in the next ten years, because history self-accelerates. That's Kurzweil's Law of accelerating returns: "technology and evolutionary processes progress in an exponential fashion."

Thus, since the rate of progress doubles every ten years or so, we will see changes in the next 90 years equivalent to the last 10,000 years, and in the next 100 years changes equivalent to the last 20,000 years. It is always the later doublings where the ferocious action is. The many skeptics about the Human Genome project being done in 15 years thought they were being proved right at year 10. They were astounded when the project came in on schedule. "People look at short sections of an exponential growth curve and imagine they are straight lines," said Kurzweil.

Noticing that his audience was astute as well as large (650 in the Herbst), the speaker gave a dense, fast-moving talk. He said that as an inventor and entrepreneur he found that "you have to invent for when you finish a project, not when you start--- you need to figure out what enabling factors will be in place when your product comes to market." That was what started him studying trends in technology. In rapid succession he showed on the screen graphs of technological advance in microprocessors per chip (Moore's Law), microprocessor clock speed, cheapness of transistors, cheapness of dynamic random access memory, amount and cheapness of digital storage, bandwidth, processor performance in MIPS, total bits shipped, supercomputer power, Internet hosts and data traffic, and then on into biotech with cheapness of genome sequencing per base pair, growth in Genbank, and further on into nanotech with smallness of working mechanical devices, and nanotech science citations and patents.

They ALL show exponential growth rates, with no slowing in overall progress, since new paradigms always arise to keep up the pace, as transistors replaced vacuum tubes in computers, and 3D molecular computing and nanotubes will replace transistors. "Everything to do with information technology is doubling every 12 to 15 months, and information technology is encompassing everything."

I was impressed that the growth curves ignore apparent shocks. The 1990s dot-com boom and subsequent bust seemed like a big event, but it doesn't even show up as a blip on Kurzweil's exponential growth curve of e-commerce revenues in the US. At dinner with Long Now sponsors after the talk, he proposed that the stringent American regulations on stem cell research will not slow the pace of breakthroughs in that field, because there are so many political (overseas, for example) and technological workarounds. The fate of individual projects is always unknowable, but the aggregate behavior over time of massive and complex arrays of activity is knowable in surprising detail.

Kurzweil expects this century to provide dramatic events early and often. "With the coming of gene therapy, before we see designer babies we'll see designer baby-boomers." By 2010 he expects computers to disappear into our clothing, bodies, and built environment. The World Wide Web will be a World Wide Mesh, where all the linked devices are also servers, so massive supercomputing can be ubiquitous. Images will be project right onto retinas, helping lead toward true immersion virtual reality. Search engines won't wait to be asked to offer information. By 2030 he presumes that nanobots will occupy and enhance our nervous systems. The brain will have been reverse engineered so that we will understand the real structure of intelligence. A thousand dollars of machine computation will exceed human brain capacity by a thousand times. Shortly after that intelligence begins to break completely free of its biological constrictions and carries humanity into suffusing energy and matter toward potentially cosmic scale (IF the restricting barrier of the speed of light can be worked around). Kurzweil noted that among "singularitarians" he is known as somewhat conservative, expecting a "soft takeoff" instead of hard takeoff.

In the Q & A he dealt with the usual "but what about limitations of resources?" questions with predictions that nanotech would increase efficiencies and make materials so fungible that what are seen now as severe limitations will fall away. Only one question made him pause, and a very long pause it was, sort of a stunned silence. I asked him (through Kevin Kelly), "As everything goes faster and faster, is there anything that will or should remain slow?" Finally Kurzweil said, "Well. You know, even meditation will go quicker." Another pause. "But it might SEEM slow," he said politely.

--Stewart Brand