4.15.2006

Review: "The Singularity is Near"

Writing about the future is hard. Ray Kurzweil, a successful inventor, author, and military advisor, is much better at it than most, and he makes an interesting, nuanced prediction of where technology is taking us in "The Singularity is Near". In essence, the Singularity is
[A] future period during which the pace of technological change will be so rapid, its impact so deep, that human life will be irreversibly transformed ... The key idea underlying the impending Singularity is that the pace of change of our human-created technology is accelerating and its powers are expanding at an exponential pace.
Kurzweil cautions, "Exponential growth is deceptive ... People intuitively assume that the current rate of progress will continue for future periods" whereas, to quote my old math professor, past a certain point exponential variables "jump up like a scalded ape."

Kurzweil's title arises from an analogy to cosmology: just as when one approaches a physical singularity, or black hole, the smoothly-but-exponentially-increasing gravitational pull doesn't really prepare oneself for the wild things that happen when one actually reaches it, the prevailing view of the future- what Kurzweil calls the "linear view" of change- doesn't prepare us to understand the coming technological Singularity. In other words (as Kurzweil flippantly quotes), "You know, things are going to be really different! ... No, no, I mean really different!" (Mark Miller, 1986). The date Kurzweil sets for the Singularity, which he has described as "technological change so rapid and profound it represents a rupture in the fabric of human history," is 2045.

The relentless march of computing power; the quantity measured here, calculations per second per $1000, has been undergoing at least exponential growth for 60 years or more. Kurzweil claims our knowledge of the brain's many functions and self-organizing structure is also growing exponentially, and that the convergence of powerful computing and a deep understanding of the brain means The Age of Non-biological Intelligence, i.e. computers much smarter than humans, is much closer than most people think.

How quickly we'll be able to reverse-engineer the brain, whether we'll be able to use brute-force computation (of which we'll have ample) to simulate the parts of the brain we can't or haven't reverse-engineered, and what the results of building a successful non-biological intelligence will be- these are the million-dollar questions.


A logarithmic graph of major events or adaptions in the history of humankind. Note how each event occurs much quicker than the event before it. What happens when the line hits the bottom? The Singularity, according to Kurzweil.

So, what does Kurzweil mean when he argues that "human life will be irreversibly transformed" by 2045? He divides future life-altering technology into the following three categories:
  • Genetics. The first wave of change. Changing our genes on-the-fly with gene therapy will become practical; "Genetic Engineering" will start to live up to its name, rather than referring to our current rudimentary practice of splicing genes from one organism into another and seeing what happens. Intelligence and body augmentation (via gene therapy and implants/prosthetics) are common by 2020, and non-augmented people are at a significant competitive disadvantage. Most genetic ailments, lifestyle problems, and age-related decline (Kurzweil's tagline is "Live long enough to live forever") are essentially eliminated for those who can afford it by 2020 to 2025.
  • Nanotechnology. The second wave of change. Within a few decades nano-manufacturing will take off and nearly all physical goods become extremely inexpensive, the majority of the cost being the price of the product's blueprint or intellectual property. One could download a blueprint for a pair of gloves or an ipod from the internet, and have the household fabricator create them out of pennies' worth of simple 'nanofuel'. The nature of wealth is transformed. Tiny nanobots are integrated into our bodies, aiding in memory and cognition, acting as a super immune system, carrying out repair tasks, and enabling realistic virtual reality. Portions of our bodies and brains gradually become non-biological and increasingly follow exponential technology price/performance curves.
  • Robotics/AI: Kurzweil takes the stance, as Wikipedia puts it, "That the functionality of the human brain is quantifiable in terms of technology that we can build in the near future" and that much of the awesome complexity of the brain arises from simpler probabilistic and self-organizational methods spelled out in our genetic code. He predicts that, using the future fruits of current computing trends and brain research, we will design a new, non-biologically-constrained intelligence more capable than ourselves, which presumably will design an intelligence more capable than itself, and things will really turn wild (technically, Kurzweil argues less that we will build such a thing and more, through advances in direct interfacing with computers, that we will become it). 2045 is the projected date for this wild bootstrapping process to fully commence, as at that point $1000 will purchase as much raw processing power as that of the brains of all of humankind combined.

Two key points in Kurzweil's argument are, firstly, that the domain of exponentially-driven trends is increasing. What began with computing power has spread into other contexts such as communications and media and will soon move into biotechnology. Relatedly, once science achieves full synthetic replacements of something- when we develop full-function replacement red blood cells, full-function replacement hands, full-function synthetic intelligences- these synthetic versions are in a position to benefit very directly from advances in science and will quickly (and definitively) surpass their biological cousins.

Once again, it's difficult to speak of the future. People can:
  • Miss huge trends: "I think there is a world market for maybe five computers." - Thomas J. Watson, Chairman of the Board of IBM, 1943
  • Over-estimate current trends: "[By 2004] Air transportation is making the multi-family apartment house obsolete, as each family now needs a private landing strip." Popular Mechanics, 1954
  • Simply be wrong:












(Popular Mechanics, February 1950, via Steve Sailer)

Given that the history of predicting the future is replete with so many mistakes, why should we listen to Kurzweil? I'd offer two reasons: the first is that Kurzweil grounds his predictions with a nuanced model of technology, his well-developed Law of Accelerating Returns, and a well-researched account of state-of-the-art advances in genetics, nanotechnology, and artificial intelligence. The second is that he has a fantastic track record in understanding innovation. He's founded nine businesses, has been awarded the National Medal of Technology, the $500,000 Lemelson-MIT Prize for innovation, and 13 honorary doctorates, was inducted into the National Inventor's Hall of Fame, and is one of five members of the Army Science Advisory Board. Simply put, Kurzweil's prediction of a technological Singularity may be right or wrong, but we would be ill-served to dismiss him as a crank.

Kurzweil's style is less eloquent than logically elegant. He draws on relevant trends from the history of our scientific and evolutionary progression, sets his context with quotations from leading scientists, and spends a great deal of time talking about some of the cutting-edge research, results, and evidence for convergence from computing, genetics, physics, biotechnology, artificial intelligence, reverse engineering of the human brain and nanotechnology, among other fields. Some of the topics he touches on, such as brain structure, are complex, but he does not introduce complexity needlessly. I find the book to be exceptionally well-grounded with many relevant citations and a fairly comprehensive section on responses to critics (see shortcomings section for caveats). With all of Kurzweil's evidence, it's difficult to dismiss that technology increases exponentially, and he argues that the root cause of this is economic competition. For us to stop this progression, and by implication the Singularity (barring horrific disasters), we'd essentially need to repeal capitalism.

Now some shortcomings. Kurzweil doesn't ground his vision in the current socio-political context, nor does he speak to the issue that much of human behavior, such as the manner in which we depart from perfect rationality, deal with change, and find motivation, is informed by our evolution in certain physical and social environments. He assumes much about the egalitarian nature of technology and unequivocally identifies quality of life with technological advancement. Likewise, he trivializes the power of rapid technological change to both cause widespread ideological dispossessment and empower those thus dispossessed. He doesn't speak to the potential of the first nation, lab, group, or individual that reaches a certain level of technological prowess to solidify their lead by doing something drastic. He doesn't go into what may come of people or intelligences being able to directly tinker with their emotions and motivations, nor the consequences of the deep diversification and customization of human experience that biotech and nanotech will enable. He doesn't contextualize his views in terms of other futurists such as Kevin Kelly. He glosses over any discussion of the subtleties and options involved in how information may be owned, while simultaneously predicting we will move almost completely to an information-centric society.

Some of these criticisms are blunted by the existence of www.singularity.com and www.kurzweilai.net - broad forums for papers on future trends, challenges, and situations - but most readers will not avail themselves of these websites.

In plain terms, Kurzweil understands technology. His writing tends to filter out certain subtleties of human behavior and doesn't seem to acknowledge the causes by which and extent to which rationality and enlightened self-interest can be lacking in certain areas. But none of these criticisms have any great impact on his strongly-supported central thesis on the accelerating nature and likely path of technology.

Despite its limitations, "The Singularity is Near" is simply the best jumping-off point for considering the future that I've come across.

4.10.2006

Nature's Bold Study

Nature's recent study comparing the accuracy of Wikipedia vs. Encyclopaedia Britannica in the sciences- the first such comparison carried out by a premier science journal- has generated significant interest in quite a few circles. Wikipedia is many things to many people, and there has been precious little data regarding its accuracy. In light of the recent controversy surrounding subtle vandalism in John Seigenthaler's Wikipedia article, Nature's conclusion doubtlessly surprised many:

However, an expert-led investigation carried out by Nature - the first to use peer review to compare Wikipedia and Britannica's coverage of science - suggests that such high-profile examples are the exception rather than the rule. The exercise revealed numerous errors in both encyclopedias, but among 42 entries tested, the difference in accuracy was not particularly great: the average science entry in Wikipedia contained around four inaccuracies; Britannica's, about three

Aside from the notable results, which many took to vindicate Wikipedia's editorial model, the public has been witness to an unusually fierce struggle between established academic institutions. Britannica, specifically, has taken out large, fiercely-worded ads in the New York Times and the London Times publicly criticizing Nature's comparison methodology. The controversy centers around study transparency, article selection and how article length was normalized, major vs minor error counts, and the conclusion drawn from the 4:3 error ratio. Britannica's take?


"The study was so poorly carried out and its findings so error-laden that it was completely without merit."


There's high-quality commentary on the current controversy at Arstechnica, on what this may mean for Britannica at Content Matters, and on some of the high-profile events involving Wikipedia that set the stage for this showdown at Plastic. As of today, a google search for Wikipedia Britannica Nature study nets over 300,000 results.

Nature's summary of the study includes links to Britannica's response and Nature's counter-response. Most of the public correspondence (and media coverage on the disagreement) between Britannica and Nature has focused on disputing and supporting factual claims about the study. Such details are critically important to work out, but this emphasis has also obscured some of the broader issues surrounding the study. In that spirit, here are the questions I think should be explored more fully in parallel with the current methodological evaluation:

Why do the study? Assessing Wikipedia's accuracy is important, but as I'm sure Nature realizes, this isn't just about Wikipedia. This is about Wikipedia's community-driven content and editorial model, and has implications for what roles such a model is qualified to play in the future of science.

What this study isn't about:
Simply comparing the number of mistakes per article in a specific article subset doesn't touch on many of the central strengths and weaknesses of Wikipedia and Britannica. Importantly (though not exhaustively), Britannica's editorial consistency and tradition of recruiting Nobel Prize winners to write articles are necessarily passed over, as are Wikipedia's impressive breadth, depth, and agility, as are the potential institutional and organizational biases both encyclopedias bring to the table. This study is intriguing but necessarily limited.

Why have Nature do the study? This sort of comparative institutional analysis is new ground for Nature. There's a lot of value to having this study done- but was Nature the right organization to do it? Nature has, in addition to its professional reputation, one of the best scientific peer-review networks in the world, which was easily re-purposed for this comparative analysis. But much of Nature's great professional reputation arises from filling a certain 'gatekeeper' role in the academic community, and though some of that expertise will carry over into designing a methodology to reliably compare vastly different popular reference works, it's unclear just how much will.

With Britannica responding in force (and from a business perspective, they have little choice but to do so), the stakes are high. The world, not to mention Nature, would be worse off if all that came of this situation was a flawed study and a tarnished reputation for a premier science journal. Then again, should Nature come out of this ambitious, unprecedented undertaking relatively unscathed, the journal may be able to fill a larger role (and one more secure from PLoS-style competition) in the future ecology of academic literature.