Monday, June 7, 2010

Black Swans and The Fifth Megatrend.

I've just finished reading Nassim Talebs book ' The Black Swan'. It's essential reading if you are thinking about forecasting, and a lucid and entertaining read as well..

The core idea is that many processes thought to be governed by Gaussian distributions are really governed by Power Law Distributions. For example, over a short term, much financial data looks like it clusters around an average value, in the same way as height or weight does. In fact it's a Power Law, where larger scale events are increasingly rare, but overpoweringly large. Over a short or selective term, a set of data like, say, the the size and frequency of bank collapses, drilling accidents or airline disruptions seems like a simple normal distribution, with the probability of severe events tiny and easily quantified. A day later, events make fools of us all.

This idea impacts on recents posts where I set out four key trends which I thought would drive the story or tertiary education through the 21st century: Demographics, Economics, Telepresence and Artificial Intelligence. In each of these trends I somewhat boldly extend a current trend in a broad swish across the century ahead and consider the effects. As Taleb would argue, I am like the Turkey extrapolating his ever increasing meals and weight gain to dream of the mighty bird he will be come spring - a simple inductive fallacy.

It's worth taking the time to consider the rationale behind each of those trends, and how much inertia they have. Are they indeed driven by fairly dull, Gaussian processes, where, for example economic growth generally flits around two or three percent a year, with the occasional wobble up or down a few percent, or are great surges over and back possible - Black Swans, in Talebs language.

Population growth seems to be at fairly low risk of a major shift. It's a consequence of many factors; how many children people want to have, how many they can have, and how long they live. The biological factors are well bounded, at the bottom (you can't have negative children) and at the top (It's really unlikely you'll have 14 children, or live to 200). The variables are driven by things with a lot of inertia - how many children you would like is keyed to economic factors and social expectations which change slowly. Infant mortality is keyed to economic wealth, and how long you live is usually driven by economics too, through the sum of your lifestyle and health from womb to tomb. So population forecasts have a lot of mass behind them. Given that uncertainties coming from basic factors like birth rates and so on amount to over a billion people either way, it would take a particularly vigorous pandemic, or an enthusiastically conducted nuclear war to make much of a difference.

Economic growth is the most interesting one. Predicting 10 fold increase in wealth feels bold in the current climate, but that's just 2.5% a year. Anything under 1% is declared a disaster by the media. Many developing economies scarcely felt the 'Global Financial Crisis' of 2009 and are still growing at rates well in excess of 2.5%. A lot of this growth is probably technologically driven, but given that a very sizeable minority of the worlds population have yet to benefit from the inventions of the 20th century, even if we invent nothing of note from now on, there is plenty of scope. There are limits to economic growth, particularly in the physical 'economy of stuff'. There is only so much physical stuff we can own or consume, and so much physical stuff we can make. The sustainable economics of information, with a marginal cost of zero, is still a subject of debate. It is bold, but not implausible, to suppose that whatever the overall level of growth in the 21st century, the regional inequalities that arose out of the industrial and colonial age will largely even out.

Good quality Telepresence seems like a no brainer - an engineering problem, with an obvious economic payoff for a solution. I can't see how it couldn't come to pass, all else being equal, although some forward advances in the affordability and speed of physical travel might delay it's adoption. Any kind of adverse system shock, like a high mortality pandemic or war, would hit transport nets more and promote telepresence.

Artificial Intelligence is probably the rockiest potential megatrend. Others might not even consider it - it will arrive as a Black Swan to many. Moore's law is a classic inductive case. The long established historical trend doesn't prove it will continue for one more day, no more than the Turkey feeding schedule does. That said, some of the worlds biggest R&D budgets are feeding this turkey and making ever more powerful chips. It's also noting that the law refers to the cost of a unit of processing power, not pure component density. Even if you hit physical  wall on component density, as is frequently prophesied, there is still plenty mileage in making chips cheaper, and with better utilisation making the power cheaper at the desktop. Think of the difference between a chip on a desktop PC, spending most of it's life running screensavers and Freecell, and the same chip in a server farm, running at optimum 24/7/365. Granted it might be running Farmville, but you get the idea. If anything, the risk for Moores law is that something like Quantum Computing will come out of leftfield and rapidly accelerate the curve.

Raw processing power doesn't necessarily mean human like artificial intelligence, but it will chip away at it, until the differences seem pedantic. When the machines can answer the phone, read your x-ray, drive your car and write your essay for you, who cares whether it's 'real' or 'fake' AI.

Overall, Technology seems to be the most risky quarter, with the potential to pull out Black Swans like rabbits out of a hat. It seems likely that they would be 'positive' Black Swans - developments in unexpected places, rather than negative Black Swans, when expected things fail to happen. That said a negative black swan, like perhaps an elegantly tailored bioweapon, cannot be dismissed.

Of course, Taleb would no doubt argue that focusing on these four trends is missing the point. The key trend, the main event that will shape tertiary education in the next century isn't any of those - it's something we haven't -can't - imagine yet, some surprise out of left field that will change the narrative, for better or for worse - Fifth Megatrend, the Black Swan.

No comments:

Post a Comment