IEEE Spectrum tries to hold Ray Kurzweil to a high prediction standard but does not apply that standard to themselves. Every year IEEE Spectrum tries to pick Technology winners and losers.
IEEE Spectrum says Rays clearest and most successful predictions often lack originality or profundity. And most of his predictions come with so many loopholes that they border on the unfalsifiable.
So how clear and profound are IEEE Spectrum in its picking of winners and losers ? Not clear. They do not define what means for a Technology to win. What is the standard for success. IEEE Spectrum does not specify dates on its picks. IEEE Spectrum claims a high ratio of success, yet they do not review in detail what happened from the point of prediction to the point of declared success.
Spectrum predicted winners
* Google’s Chrome operating system
* Pixel Qi’s dual-mode screen provides both e-paper readability and full-color video.
* Intrinsity’s hot-rodded processor gives cellphones PC smarts.
* IBM helps Russian Railways reinvent the railroad’s data infrastructure.
* NanoGaN’s gallium nitride substrates will help manufacturers make better lasers.
Spectrum predicted losers
* D-Wave Systems’ quantum computers won’t outperform ordinary ones. (I disagree)
* NanoUV’s extreme ultraviolet light source is revolutionary, but that won’t entice chipmakers to use it.
* Cellulosic ethanol— “grassoline”—is an environmental threat rather than a panacea.
* The Chevrolet Volt plug-in hybrid car is imaginative, daring, and superb, but uneconomical.
* Airport security screening will go a lot faster with a new biometric system that reads passengers’ minds.
We took at a look at IEEE Spectrum’s 2009 technology picks There picks in 2009 basically break down to small early stage companies are risky and more prone to failure and big companies with military backing are safer and likely to succeed if they are scaling up deployment with large financial backing this year. In spite of that method, they predicted that Intel would succeed with Larrabee GPGPU chip would be a success and the first generation Larrabee GPGPU was not released as a consumer product.
IEEE Spectrum makes a case of the prediction of internet success. In his 1990 book, The Age of Intelligent Machines, Kurzweil predicted the rise of the Internet as a medium for public communications, commerce, education, and entertainment. IEEE Spectrum makes the case that Minitel existed in 1981 in France. They then notice that others also predicted it as well, which is meaningless.
Predictions are like baseball hitting average. How good a hitting average is depends upon how you compare to other hitters and what is the level of pitching that you are facing. Also, very accurate predictions with a lot of detail is like Babe Ruth calling the location of a home run.
The difficulty of a prediction is also related to making bets in a casino. Correctly picking red in roulette is different from picking one number of the 37 to 38 possible numbers on the wheel.
I have gone through the exercise of making predictions, trying to explain the predictions and reviewing and judging my own predictions.
What I have learned is that you have to be very clear and specific to make a prediction that is clear enough to be judged by a third party. It takes a lot of work to get the background to know what the current situation is and then more to analysis and understanding of competition and other forces to project forward.
Trend extrapolation is usually a safer way to make a prediction, but can be non-trivial if you extend out far enough. Moore’s law holding for one year is trivial but Moore’s law holding for 30 years becomes an interesting and more controversial prediction.
Calling a point and time and change of speed in trends is a more interesting and daring prediction. Predicting transformation in societal behavior can be very difficult and interesting.
Getting insights into the complexities of how change occurs and how technological adoption happens and then having relative success in applying those insights with actually more accurate predictions is useful.
Being able to identify consistent bias that cause inaccuracy and bias in the technological predictions of other predictors is useful to improve the hit rate and accuracy of your own predictions and could eventually educate others on how to make better predictions.
I have identified impartial metrics that can be used to judge the difficulty of technological predictions. It is similar to judging the difficulty of Olympic diving. How many twists and flips go into the prediction ? How high is the jump / How distant is the prediction ? What is the rate of change and number of events that could occur for the thing being predicted ? Certain things like predicting the construction of infrastructure (bridges, skyscrapers, nuclear power plants) have long lead times.
There are also measures of insitefulness and profoundness. What is the ratio of people who agree or disagree at the time the prediction is made ? What is the difference from the mainstream vision of the future ? This would require a steady to establish the mainstream baseline vision of the future.
The current mainstream baseline(s) and majority minority opinions would include some of the following –
Continuation of economic growth trends at 2-5% world GDP growth.
Some resource commodity constraints.
Two mainstream views on climate – no problem at all in the future and the climate problems opinion.
Continued improvement with computer processing speed and memory size.
No technological singularity, no major shifts with super-technology
No molecular nanotechnology
More electrification of cars
No radical life extension
Plus you can add some voting at the time the prediction is made as to how clear it is, how profound it would be if true etc… to go along with the impartial metrics.