Ray Kurzweil and his acolytes strongly believe the Singularity is near. A lengthy profile in the Sunday New York Times business section of June 13 describes how Singularity proponents have even established a institution of putatively higher education, Singularity University, offering a ten-week "graduate" course and a nine-day executive program intended to convey all of the ways that new technologies will converge to permit us to "transcend our human limitations." The ten-week course is only $25,000.
What is the Singularity? According to Kurzweil, the Singularity is the coming "union of human and machine, in which the knowledge and skills embedded in our brains will be combined with the vastly greater capacity, speed, and knowledge-sharing ability of our own creations." Furthermore, "our intelligence will become increasingly nonbiological and trillions of times more powerful than it is today--the dawning of a new civilization that will enable us to transcend our biological limitations and amplify our creativity. In this new world, there will be no clear distinction between human and machine, real reality and virtual reality."
Much of this utopian prognostication is based on extrapolations of past trends of such phenomena as increases in maximum computer memory density, increases in maximum computer processing speed, and increases in maximum amounts of DNA sequence generated per dollar. The potential for folly in such thinking, what I would like to refer to henceforth as the Kurzweil Fallacy, would be apparent you might think to someone as sophisticated as Mr. Kurzweil, an inventor and software engineer who has been highly successful in the commercial arena. Then again, Mr. Kurzweil takes on the order of 150 pills per day and talks of living forever, or at least for hundreds of years, which I regard as a highly dubious proposition. I am willing to go out on a metaphorical limb and predict that Mr. Kurzweil will not live forever. In fact, I am willing to bet that he does not even live to one hundred and fifty years old.
It appears that Bill Gates, Larry Page, and Sergey Brin, all support Mr. Kurzweil in his bold pronouncements. Of course, none of these really rich people have any validated expertise in the life sciences. So, in spite of their presumably deep knowledge of computer technology, it is not clear that they can adequately assess the claims of the Singularity enthusiasts where they relate to "transcending human biology." There is also, of course, the possibility that the Singularity-focused enthusiasm of Gates, Page and Brin is caused in part by thinking that their respective corporate creations will greatly benefit from the technological developments associated with the Singularity. In any case, these three individuals may not be the best-positioned to detect the unanticipated negative effects of various elements of technological progress. I can recommend that they all read, Edward Tenner's, "Why Things Bite Back: New Technology and the Revenge Effect," Fourth Estate, 1996 (think Deepwater Horizon or computer viruses).
Ray Kurzweil, Bill Gates, Larry Page, and Sergey Brin and other techno-utopians should know that exponential growth does not generally persist indefinitely. Furthermore, even if the density of computer memory or processing speed continues to increase and the amount of DNA sequencing per unit cost continues to decline as anticipated by Kurzweil et al., the ability to apply computers in useful ways and the ability to solve difficult biomedical problems are not likely to advance at rates nearly as impressive. Maximizing the efficiency of memory with respect to space or DNA sequencing with respect to cost are, respectively, profoundly less challenging problems (as hard as they may be) than, for example, designing software that effectively simulates human cognition or creating medical treatments that dramatically improve outcomes from human diseases.
In the above context, Ashlee Vance in the New York Times article mentioned above quotes William S. Bainbridge of the National Science Foundation: "We are not seeing exponential results from the exponential gains in computing power," he says. "I think we are at a time where progress will be increasingly difficult in many fields." ... "We should not base ideas of the world on simplistic extrapolations of what has happened in the past," he adds.
Since my knowledge of genetics is stronger than my knowledge of computer science, I will briefly note some considerations that testify to the fundamental silliness of equating DNA sequencing cost efficiency with advances in overall understanding of biomedical phenomena on the molecular level or overall effectiveness in treating illness. While the efficiency of determining DNA sequences has, as cited above, advanced at an absolutely amazing rate for the past couple of decades, the ability to predict phenotypes (i.e. traits) from genotypes (i.e. DNA sequences of genes) has progressed only modestly in that time period. The main reason for such slow progress in anticipating phenotypes based on genotypes is that most traits of interest are due to interactions among numerous genes and between genes and environmental factors.
It should not be difficult to appreciate why quantitative predictions regarding such interactions remain extremely challenging. If there are roughly twenty-thousand genes in a typical human genome (the number is subject to modest individual variation), the total number of conceivable gene-gene interactions is enormous. Furthermore, as suggested by studies in fruit flies and other experimental organisms, which genes interact functionally with any particular gene appears to vary with the overall genetic context, an under-appreciated but potentially critical complicating phenomenon. Assessing gene-environment interactions is even more daunting if only because it is not typically even clear what all of these non-genetic factors are or how to measure them in relevant ways.
In the realm of medical therapy, while many new treatments have been devised involving both traditional small molecule drugs and biological products, such as antibodies, cellular growth factors, and cytokines (molecules that deliver messages from cell to cell), they often improve health only incrementally and at enormous financial cost. In those cases where they are highly effective and substantially reduce suffering or expand lifespan in some patients, they also typically elicit significant undesirable side-effects in some patients (e.g. recombinant erythropoietin, used to treat anemia may accelerate some cancers and a recombinant antibody that effectively treats rheumatoid arthritis can increase susceptibility to serious bacterial infections). There is no basis at present for believing that medical interventions based on the postulated but not-yet-realized nanobots, often-invoked by Singularity enthusiasts for the resolution of all medical threats and malfunctions, will perform their duties without trade-offs and side-effects like those associated with every other therapeutic agent ever employed. It is entirely reasonable to expect significant diagnostic and therapeutic progress to continue, but predicting complete conquest of disease is unrealistic in light of both the numerous deficiencies in our understanding of the subtleties of cellular and molecular function that are likely to persist in some measure for many years and the extremely-difficult-to-avoid trade-offs that afflict most medical interventions. Indefinite human lifespan remains wishful thinking well beyond the realm plausibility.
The opinions expressed above do not reflect official views of the institutions with which I am affiliated.