Monday, September 18, 2017

Reassessing Kurzweil's 2009 Predictions

Raymond Kurzweil's 1999 book The Age of Spiritual Machines remains a touchstone of futurologists and their critics not only because of Kurzweil's continuing presence within the information technology and futurology worlds, but because the book also offered multiple, lengthy, even comprehensive lists of forecasts. Many of his forecasts were about subjects far outside his realm of expertise (macroeconomics, geopolitics, military science, etc.), and unsurprisingly were commonplaces, but a good many of them were precise technological forecasts lending themselves to testing against the day's technological state of the art--and thereby checking the progress of the technologies with which the book was concerned, and perhaps, also our progress in the direction he claimed we were going in, toward a technological Singularity.

Naturally a good many observers (myself included) reviewed Kurzweil's 1999 predictions for 2009 when that year arrived. Different writers, depending on which forecasts they chose to focus on, and how they judged them, came to different conclusions about his accuracy. I emphasized those precise, easily tested technological forecasts, and saw that many had not come to pass. I noted, too, that while not accounting for each and every error, there was a recognizable pattern in a good many of these, namely that Kurzweil assumed advances in neural networks enabling the kind of "bottom-up" machine learning that made for better and improving pattern recognition, as with the speech recognition supposed to make for the Language User Interfaces we never really got.

Looking back on those predictions from almost a decade on, however, it seems worth remarking that the improvement in neural nets was indeed disappointing in the first decade of the twenty-first century--but got a good deal faster in the second decade. And right along with it there has been improvement in most of the areas he seemed to have been overoptimistic about, like translation technology. Indeed, as one of those who looked at his 1999 predictions for 2009 and was underwhelmed, I find myself increasingly suspecting that Kurzweil was accurate enough in guessing what would happen, and how, but, due to overoptimism about the rate of improvement in the underlying technology, was off the mark in regard to when by a rough decade.

Moreover, this does not seem to have been the only area where this was the case. The same may go for the use of carbon nanotubes, another area where after earlier disappointments this decade technologists achieved successes that may plausibly open the way to the new designs of 3-D chips he described, permitting vast improvements in computer performance. (As it happens, Kurzweil guessed that by 2019 3-D chips based on a nanomaterial substrate would be the norm. If IBM is right, he will have been off by rather less than a decade in this case.) It may also go for virtual reality--which likewise seems to be running a rough decade behind his guesses.

The advance in all these areas has, along with a burst of progress in other areas in which Kurzweil took much less interest (like self-driving cars of a different type than he conceived, and electricity production from renewable sources), contributed to a greater bullishness about these technologies--a techno-optimism such as I do not think we have had since the heady years of the late 1990s when access to personal computing, the Internet and mobile telephony began the proliferation and convergence that has people watching movies off Netflix on their "smart" phones while sitting on the bus.1

Of course, it should still be remembered that much of what has been talked about here is not yet an accomplished fact. The advances in AI have yet to make for proven improvements in consumer goods, while those 3-D nanotube-based chips have yet to enter production. At best, the first true self-driving car will not hit the market until 2020, while renewable energy (hydroelectric excepted) is just beginning to play a major role in the world's energy mix. Moreover, that the recent past has seen things pick up does not mean that the new pace of progress will remain the new norm. We might, in line with Kurzweil's expectation of accelerating returns, see it get faster still--or slow down once more.

1. Kurzweil's self-driving cars were based on "intelligent roads," whereas the cars getting the attention today are autonomous.

No comments:

Subscribe Now: Feed Icon