Chapter 133 22.3 Positive myopia

With those lessons in mind, Farmer formed a new company with five other physicists, one of whom was a former member of the Chaos Society.This time, what they want to crack is the dream of all gamblers: Wall Street.And, this time, they will use high-performance computers.They'll load those computers with experimental nonlinear dynamics and other tricks that rocket scientists can't keep secret.They will think from the sidelines and allow this technology to take on as much responsibility as possible without their control.They're going to create something, if you will, an organism, that can play multimillion-dollar gambles on its own.They're going to make this organism... (well, beat the drums, please)...predict the future.With a bit of bravado, these seasoned guys put up a new sign: The Forecasting Company.

Those in the forecasting firm figured out that to make big money in the financial markets, all it takes is being able to see what's going to happen a few days in advance.Indeed, recent research at the Santa Fe Institute, where Farmer and colleagues have stayed, explains that "seeing far does not mean seeing well."When you are immersed in the complexity of the real world, there are few clearly defined choices, and incomplete information blinds all judgments. At this time, judging too distant choices will not achieve the expected purpose.Although this conclusion seems to fit human intuition, it is unclear why it should also fit the world of computers and models.The human mind is easily distracted.But suppose, say, that you already have unlimited computing power and are dedicated to the task of prediction.So why isn't it better to see deeper and further?

The simple answer to this question is: When extremely small errors (due to limited information) persist into the very distant future, they will converge into extremely large errors.Even if computation itself is free (and it never is), the cost of dealing with these exponentially growing, error-tainted possibilities is enormous and simply not worth paying.Researchers at the Santa Fe Institute, Yale economist John Genakopoulos, and Minnesota professor Larry Gray once used a computer program for a chess match as a test bed for their forecasting work. (The best computer chess programs, such as the top Deep Thought program, can beat all but a few top grandmasters.)

The results turned out to be completely contrary to computer scientists' expectations. Neither the "deep thinking" program nor the human chess master actually needed to look too far to play very good chess.This limited foresight is what is called "positive myopia."Generally speaking, these masters will first survey the situation on the board, and only make a prediction on the next move of each chess piece.Next, they pick out the most likely move or two and consider the consequences of those moves in more depth.Although every extra step forward, the possible moves will explode exponentially, but in each round, those great human masters will only focus on a limited number of the most likely responses On the move.When confronted with a familiar environment they have experienced in the past and are well aware of the trade-offs in it, they will occasionally take a few steps forward.But, in general, masters (and now add "deep thinking") lay out games empirically.For example: favor moves that increase options; avoid moves that work well but require sacrifices; play from positions that are adjacent to multiple positions.Strike a balance between being forward-looking about the situation and really looking holistically at what's going on.

We encounter similar tradeoffs every day.Whether in business, politics, technology, or life, we must anticipate what lurks around the corners.However, we are never given enough information to make fully informed decisions.We operate in the dark.To compensate, we can only rely on experience, or rough guidelines.The empirical rules in chess are pretty good rules of life that you can rely on. (Attention here, my daughters: Prefer moves that increase your options; avoid moves that go well but require sacrifices; start with vantage points that are adjacent to multiple vantage points. There is a balance between being forward-looking and really looking holistically at the current state of affairs.)

Common sense crystallizes this "positive myopia."Rather than spend years creating a company employee handbook that predicts everything that could happen - out of date as soon as it goes to press - adopt the kind of positive myopia and don't think that far, obviously much better.In other words, devise some general guidelines to deal with things that seem to be bound to happen in the "next step" first, and then deal with those extreme cases when they do happen.If you're in a new city and you want to travel during rush hour, you can plan a detailed route across the city on a map - think farther away - or, try something like, "Go west, get to When you go along the river road, turn left again."Usually, we do a little of both.We try not to think too far ahead, but we do pay attention to what is happening right in front of us.We would meander west, uphill or downhill, and, wherever we were, we would pull out a map to see the next intersection we were about to reach.The method we use is actually a limited look-ahead guided by empirical rules.

The predictive mechanic works just as well without looking like a Prophet.Just about any pattern will do, as long as it can spot limited patterns behind random and complex camouflage.