Inventor and futurist Ray Kurzweil remains coolly confident in his prediction that by the 2030s, blood-cell sized computers will integrate with the human brain and dramatically expand its cognitive capacity well beyond the neocortex’s paltry 300 million or so pattern recognizers.
And why shouldn’t Kurzweil be confident? By his count, he’s been right about 86 percent of the time, and that’s not counting near-misses like predicting we’d all be riding in self-driving cars by now.
Advancing technology’s capacity to mimic, and eventually deeply integrate with, the human brain was one of the central topics of a public Q&A Kurzweil participated in last night in Louisville, KY as part of a promotional tour for his new book, How to Create a Mind: The Secret of Human Thought Revealed. The new book draws on his deep knowledge of language and cognitive hierarchy to predict how computers will continue to expand humans’ ability to store and relate information in what we call “intelligence.”
The talk, which was taped for national airing, ranged widely, as does Kurzweil’s influence on technology and almost any discussion of what the world will look like 30 years from now. For the most part, Kurzweil steered clear of the headline-grabbing philosophical and ethical implications of his predictions, most notably that supercomputing will create a macro cyber-intelligence in which human consciousness will live forever. (As I imagine many of you will note in the comments section, that’s a gross oversimplification of Kurzweil’s fascinating work.) Instead, Kurzweil focused on the fundamentals of the science behind How to Create a Mind, and how current technology already has greatly augmented those lame 300 million recognizers nature gave us.
Kurzweil said his latest predictions are built around his theory of Pattern Recognition Theory of Mind (PRTM), which describes our cognitive processes as a series of nested activities. In How to Create a Mind, he cites the example of how most people struggle to recite the alphabet backward, although the components of the information are clearly stored in the neocortex. But the pattern — or more precisely, the potentially thousands of patterns — in which the neocortex connects those data are the essence of intelligence, human or artificial.
This theory is based on Kurzweil’s own ground-breaking work on optical character and speech recognition software, and a similar theory was employed in large part to program Watson, the supercomputer that recently whooped up on human Jeopardy! champions. Kurzweil defends the Watson project from criticism that the supercomputer was simply running statistical analysis against specialized programming. Watson actually “read” 200 million Wikipedia pages to build its knowledge store, and the statistical analysis it ran was patterned after the human brain’s own hierarchical models of data relationships (Kurzweil dubbed them Hierarchical Hidden Markup Models in his speech recognition work). Kurzweil was quick to note that Watson is not as good as an average human when it comes to understanding a single Wikipedia page; it’s the ability to store massive quantities of information and quickly relate it that makes the computer so “smart.” He suggested that soon, technology derived from the Watson project will be able to aid physicians in diagnosing illnesses, since doctors don’t have enough time or pattern recognizers to read and immediately recall tens of millions of pages of medical research.
And he’s confident that his Law of Accelerating Returns will continue to hold true for computational power, even though Intel now predicts Moore’s Law will run its course by 2022 or so. (If you haven’t read this excellent TechRepublic post by Peter Cochrane, do so now.) Kurzweil describes Intel’s 3D structure for transistors as the sixth paradigm of Accelerating Returns (with Moore coming in at number five) that will continue to drive exponential growth in computing power and get us to that technological singularity everyone is so excited (or freaked out) about.
By the 2030s or 2040s, he envisions micro-computers embedded non-invasively in the brain that will act as an interface to a “cloud” of storage and processing power — it will be like having five or 10 neocortexes on demand. And, given that the adult brain often has to overwrite redundant instances of data to “learn” new things, that won’t be so different than our use of external computers to store and process data today. Fondling his own smartphone throughout the hour-long presentation, he repeatedly described such devices as “brain extenders.”
Other geeky stuff on the table:
- Kurzweil really digs Google’s Project Glass, and says that within five years the devices will become commonplace. When you see somebody on the street, your visor will tell you their name and other quick information so your brain won’t have to do all that work.
- He’s also high on Google Cars, but he still counts that as a “miss” in evaluating his own predictions, given that they are not yet broadly available to consumers.
- On the ongoing advance of technology, Kurzweil says: “A kid in Africa with a smartphone has more information than a U.S. President had 15 years ago.” Process that for a second.
- If you’ve never seen it, be sure to check out this video from 1965 of Kurzweil as a 17-year-old on I’ve Got a Secret, demonstrating a music-composing computer of his own device. Dig brainy Miss America Bess Myerson.
Book cover image courtesy of Amazon.com.