I think A.I. -- even just really good narrow A.I. -- could completely throw off a wide range of future predictions about the economy, government, world peace, and so on. The problem I see that it poses for future forecasting is that it introduces dependencies among a large number of seemingly independent areas of consideration. Let me explain:
I am no futurist; but if I were to seriously try to predict the future (as opposed to just tossing out random predictions for fun), in such a way as to keep risk to a minimum, as I imagine many serious analysts do, I would:
Choose potential outcomes that can be reliably estimated (or have been in the past), given sufficient research. The best types of outcome with this property are those that are contingent on the gross movement of a large number of independent sub-events -- such as the outcome of an election (the sub-events being individual votes), where accurate polling can be performed.
Choose several of them that are independent; and whereby "success" is measured by the sum total of the number of correct guesses of these independent outcomes.
This is a classic approach to mitigating risk, by using sums over independent events to "smooth it away". The problem, however, is that you can't ever be sure that all the sub-events really are independent; for example, large-scale phenomena like a global debt crisis can lock all or most of the sub-events into a dependency. While you might be able to accurately estimate outcomes most of the time when things stay independent, every once in a while dependencies like this will cause a large fraction of your predictions to suddenly be way off, all at once!
I believe that the arrival of good narrow A.I. could be one of those phenomena that introduces such systemic dependencies. For, in a world in which we have good narrow A.I., whole classes of industries will be transformed in lock-step -- some correlated, and some anti-correlated. In a world experiencing sudden jumps in A.I. capability, and after the effects of globalization have reached saturation, I would expect to see large, sudden moves in labor markets, correlated across large parts of the globe, as whole classes of jobs become automatable. One day, truck drivers are everywhere; and just a few years later, they are much fewer in number, practically everywhere in the developed world. Shifts like this happened in the past -- e.g. when the refrigerator replaced the milkman -- but now we will see this across many job classes all at once; and the shifts will occur relentlessly. They could, in turn, cause greater political polarization across large parts of the world -- the shifting job markets could put the world on edge; and in such situations middle-ground political stances become less popular.
As a result of this, I think that accurate future predictions must take into account the systemic effects of A.I. on job markets, economies, and governments. Failure to do so could catch governments and large fund managers completely by surprise.
Submitted February 26, 2015 at 07:23AM by starspawn0 http://ift.tt/1A8kQvp thisisthewayitwillbe
No comments:
Post a Comment