Chapter 131 22.1 The brain that catches the ball

"Tell me about the future," I begged. I'm sitting on the couch in my tutor's office right now.I had made the trek to this alpine outpost at one of the Earth's energy points, the Los Alamos National Laboratory in New Mexico.The tutor's office is covered with colorful posters of past high-tech conferences, outlining his almost legendary resume: when he was still an unconventional physics student, he recruited a group of hippie hackers to set up an underground organization, using a wearable computer to win the house's money in Las Vegas; he, by studying dripping faucets, became a leading figure in a gang of renegade scientists who invented the burgeoning science of chaos; He, the father of the artificial life movement; he, who now leads the study of the new science of complexity in a small laboratory diagonally across from the Los Alamos Atomic Weapons Museum.

The instructor, Don Farmer, was a tall, thin man in his thirties who looked a lot like Ichabod Crane in his stud tie.Donne is embarking on his next unusual venture, starting a company that uses computer simulations to predict stock prices and beat Wall Street. "I've been thinking about the future, and I have a question," I said. "You want to know whether IBM's stock is going to go up or down!" Farmer suggested with a wry smile. "No. I wonder why the future is so unpredictable." "Oh, it's easy." The reason I ask about predicting the future is that prediction is a form of control, and one that is particularly well suited to distributed systems.By anticipating the future, living systems can change their posture, adapt to the future in advance, and in this way take charge of their own destiny.John Holland said: "What complex adaptive systems do, is predict".

Farmer's favorite example of dissecting predictive mechanics is: "Come on!" he says, throwing a baseball at you.You catch the ball. "You know how you catch this ball?", he asked. "By predicting." Farmer is a firm believer that you have a mental model of how a baseball flies.You can use Newton's classical mechanics formula f=ma to predict the trajectory of a high-flying object, but your brain itself does not store such basic physics calculations.Rather, it builds a model directly from empirical data.A baseball player observes a bat hitting a baseball a thousand times, raises his gloved hand a thousand times, adjusts his predictions a thousand times with his gloved hand.Somehow, his brain gradually compiled a model of where the baseball landed—a model that was almost as good as f=ma, only not as broadly applicable.This model is based entirely on a series of hand/eye data generated during past catches.In the field of logic, such a process is collectively called induction, which is quite different from the deduction process that leads to f=ma.

In the early days of astronomy, before Newton's f=ma, predictions of celestial events were made based on Ptolemy's model of nested circular orbits—a circle within a circle.Because the central premise on which Ptolemy's theory was based—that all celestial bodies orbit the Earth—was wrong, the model needed to be revised every time new astronomical observations provided more precise data on the motion of a star.Still, the nested complex structure is surprisingly robust enough for endless tinkering.Every time they got better data, people added another layer of rings inside the ring-in-ring-in-ring model, and used this method to adjust the model.Despite all its serious mistakes, this baroque analog works and "learns".This simple-minded system of Ptolemy has served exactly 1400 years for the adjustment of the calendar and the actual prediction of the celestial phenomena!

A baseball outfielder's "theory" of airborne objects based on experience, much like the later stages of the Ptolemaic planetary model.If we parse the outfielder's "theory," it's incoherent, off-the-cuff, complex, and approximate.However, it can also be developed.It's a messy theory, but it not only works, it improves.If we have to wait until everyone can understand the formula f=ma (besides, it is better to understand half of f=ma than not to understand anything) before acting, no one can catch anything at all.Even if you understand this formula now, it is useless. "You can solve a baseball in flight with f=ma, but you can't solve it in real time in the outfield," Farmer said.

"Now, here's this!" Farmer tossed another inflated balloon.The thing was bouncing around the room like a drunk.No one can handle this thing.And this is a classic manifestation of chaos—a system with sensitive dependence on initial conditions.A slight undetectable change in the launch of the balloon can also be magnified into a huge change in the direction of flight.Although the law of f=ma still governs the balloon, other forces, such as the driving force, the push and pull of the air lift, contribute to the unpredictability of the trajectory.The crooked balloons in this dance of chaos reflect the elusive waltz of sunspot cycles, ice age temperatures, epidemics, water flowing down pipes and, more pertinently, Stock market volatility.

However, is the trajectory of the balloon really unpredictable?If you try to solve the equation for the balloon's wobbly flight, you'll find that its path is non-linear, so it's almost unsolvable, and therefore unpredictable.Still, a teenager growing up playing games from Nintendo, a Japanese game company, can learn how to catch a balloon.Although it is not completely accurate, it is much better than pure luck.After taking it dozens of times, the child's brain begins to construct some kind of theory, or some kind of intuition, some kind of induction based on the obtained data.After flying the balloon a thousand times, his brain had constructed some kind of model of the flight of the rubber ball.Although such a model cannot accurately predict where the ball will fall, it can detect the flight intention of the flying object, for example, whether it is flying in the opposite direction from the launch, or circles in a certain pattern.Perhaps, over time, the person has a 10 percent higher chance of catching balloons than if he had caught it purely by luck.What more could you ask for than catching balloons?In some games, not much information is needed to make valid predictions.Like running away from lions, or investing in stocks, even if it's just a little bit more than pure luck, it means a lot.

It can almost be said that "living systems"—lion prides, stock markets, evolving populations, intelligence—are unpredictable.The kind of chaotic, recursive causality they have, the mutual causation of the parts, makes it difficult for any part of the system to extrapolate into the future by conventional linear extrapolation.However, the entire system can act as a distributed device for making approximate predictions about the future. To crack the stock market, Farmer worked hard at deriving financial market movements. "The lovely thing about markets is that you don't need a lot of forecasting to do a lot of things," Farmer said.

On the gray back page of the newspaper, there are charts of the ups and downs of the stock market, showing only two dimensions: time and price.Since the day there was a stock market, investors have been carefully interpreting this black line that swings between two dimensions, hoping to find some kind of pattern that can predict the direction of the stock market.Even vague directional cues can pay off, as long as they're reliable.That's why expensive financial newsletters promoting one way or another to predict the future direction of a chart have become a permanent adjunct to the stock world.People who work in this profession are known as chart analysts.

In the 1970s and 1980s, chartists had some success in forecasting money markets because, according to one theory, the powerful roles of central banks and treasuries in money markets constrained Various variables are considered, so a relatively simple linear formula can be used to describe the performance of the entire market. (In linear calculation, a solution can be represented by a straight line in the graph.) And when more and more chart analysts use this simple linear calculation to successfully find various trends, the profit of the market will also increase. It's getting thinner and thinner.Naturally, forecasters began to look to wilder and messier places, where nothing but nonlinear arithmetic ruled.In a nonlinear system, the output is not proportional to the input.And the vast majority of complex systems in the world—including all markets—are nonlinear.

With the advent of cheap, industrially superior computers, some aspects of nonlinearity have become understood by forecasters.Financial prices can be embodied as a two-dimensional curve, and by analyzing the nonlinear phenomenon behind this two-dimensional curve and extracting a reliable model, you can make money, and it is a lot of money.These forecasters can speculate on the future direction of the graph and then place bets on the prediction.On Wall Street, computer geeks who figure out one way or the other are known as "rocket scientists"—stock market analysts.And these technical geeks in suits and shoes, working in the basements of various trading companies, are actually hackers in the 1990s.Don Farmer, a former mathematical physicist, and his colleagues who used to take his mathematical adventures with him, use as offices four brick houses in Santa Fe, a place in the United States that cannot be farther from Wall Street. He is now the hottest stock market analyst on Wall Street. In reality, there are not a few but thousands of factors that affect the two-dimensional graphic trajectory of a stock.When we plot the thousands of vectors for a stock as a line, they are all hidden and only the price is revealed.The same happens when we graph the activity of sunspots or seasonal changes in temperature.For example, you can represent the sun's trajectory on a plan with a simple thin line that changes over time, but the various factors that affect this line are incredibly complex, intertwined, and repeated. cycle.Behind the surface of a two-dimensional curve lives a chaotic combination of forces that drive the curve.A real chart of stocks, sunspots, or the climate would include an axis for all influences, and thus the chart would also be an indescribable thousand-armed monster. Mathematicians have struggled to find ways to tame these monsters, which they call "higher-dimensional" systems.Any living creature, complex robot, ecosystem, or autonomous world is a high-dimensional system.The library of forms is a high-dimensional system building.Just 100 variables can create an enormous number of possibilities.Because each variable behavior interacts with the other 99 behaviors, you can't examine any single parameter without also examining the interacting population as a whole.Even a simple climate model with just three variables, for example, feeds a kind of chaos that makes any kind of linear prediction impossible by having some kind of weird loop back on itself. (Chaos theory was originally discovered because of failures in weather forecasting.)