Thursday, June 3, 2010

Response to Nassim Nicholas Taleb

I’ve had an interesting quasi-debate with Nassim Nicholas Taleb, with whom I often agree on many subjects. Last Friday, he tweeted:

@nntaleb: The best predictor of whether a technology will be used in 10 years is if it has been in use for 10 years.

I understand his basic point: Tried and true technologies are more likely to survive than fads of the day. But “the best predictor?” Really? As a contrarian and an aspiring futurist myself, I disputed Dr. Taleb’s point. I questioned if he truly believed, say, landline telephones would still be around in ten years. He followed up by posting this in his blog:

I wrote that a technology, book, cultural practice, religion, drug, opera piece, fashion, that is, all nonperishable cultural goods, are likely (in expectation) to last as long as they have been in existence. So when you observe a technology, you can expect it to be in the middle of its life. This note explains the math behind the point.

Note: This is an average across all technologies, a distribution, something probabilistic not deterministic: some fool of randomness wrote to me to wonder whether telephone land lines should be expected to last another century, another idiot tried to use my idea to compare Microsoft to Apple.This is a statistical framework for the dynamics of cultural uses.

Taleb expanded his initial assertion to a more general point that non-perishable goods could be statistically expected to be in the middle of their lifespan. Notice the italicized part (emphasis his). I’m the “fool of randomness” to whom he is referring.

Yet Taleb’s theory fails nearly any empirical test I can think of. There are plenty of examples of technologies which are almost certainly in their twilight (e.g. landline phones) and others which are probably closer to the beginning than the end of their lifespans (e.g. the World Wide Web). Given Taleb’s previous response to my inquiry about landline phones, I don’t think he would dispute this. In fact, he says “This is an average across all technologies, a distribution, something probabilistic not deterministic.”

But if this is the case, what good is the theory? I am racking my brain trying to think of any situation where it might be useful to know the average lifespan of all technologies (or books or companies), as opposed to the lifespan of a particular technology, but I am coming up blank. When we attempt to plan for the future, would it do us any good to know that the average technology can be expected to be around for, say, another 100 years? Of course not. We are only interested in specific technologies (or books or companies), and on that point, Dr. Taleb’s theory is nearly useless.

The things which we usually consider when evaluating the future of a technology are social, economic, and scientific trends: Upcoming technological replacements, the economic niche that the technology fills, how entrenched the technology is into the infrastructure of the modern world, and the advantages and disadvantages of the technology. The length of time it has existed should be, at most, an afterthought to these other criteria, because it is simply not an effective measurement. Taleb presumably does not believe that landline telephones will be around for another century. But the same variables which predict the premature demise of the landline phone are nearly always a better predictor of a technology’s future lifespan than Dr. Taleb's system.

Far from being the best predictor of a technology’s future lifespan, Taleb’s system is almost always far off the mark when it comes to measuring the lifespan of any specific technology. Technologies are not the stock market. The “average across all technologies” is a completely meaningless measurement. Far better measurements for the future lifespan of individual technologies exist.

1 comment:

  1. That confused me as well. It does seem counter intuitive. I would agree with you and wonder if NNT will expand on the concept later on. I think of him as someone who espeically doesn't like theories that aren't useful.