Chapter 136 22.6 Diversity of predictions

If silicon chips are good enough to act as crystal balls to guide a super-military war, and if algorithms running fast in small computers are good enough to provide predictive technology to see through the stock market, then why don't we retrofit a supercomputer and use it to What about the rest of the world?If human society is just a large distributed system composed of various people and machines, why not assemble a device that can predict its future? Even a little cursory research into past forecasts shows why.In general, those traditional forecasts of the past are worse than random guesses.Those old classics are like cemeteries, where all kinds of prophecies about the future are buried—prophecies that have never been fulfilled.While some prophecies hit the bull's-eye, there is no way to distinguish in advance the rare correct predictions from the large number of false ones.Because predictions go wrong so often, and because it is so tempting and confusing to believe wrong predictions, some futurists, on principle, shy away from making any predictions at all.To emphasize the hopeless unreliability of trying to predict, these futurists prefer to state their bias with willful exaggeration: "All predictions are wrong."

They also have a point.Long-term forecasts that have proven correct appear to be so few that statistically they are all wrong.And by the same statistical measure, there are so many short-term forecasts that are correct that all short-term forecasts are correct. Nothing can be said with more certainty about a complex system than that it will be exactly the same the next moment as it is this moment.This observation is close to the truth.A system is something that lasts; therefore, it is just a process that repeats itself from one moment to the next.A system—even a living thing—seldom changes.An oak tree, a post office, and my Mac run almost unchanged from one day to the next.I can easily vouch for a short-term prediction of complex systems: they will be similar tomorrow to what they are today.

There's another cliché that's equally true: From one day to the next, things occasionally change a little.However, can these immediate changes be predicted?If so, can we add up this series of predictable short-term changes to outline a possible medium-term trend? Can.Although long-term prediction is basically impossible, short-term prediction is not only possible but necessary for complex systems.And, there are some types of medium-term forecasts that are perfectly feasible and increasingly feasible.Despite the Alice-in-Wonderland-esque feel of making some reliable predictions about current behavior, there is a steady increase in human predictive power in all sorts of social, economic, and technological aspects.As for why, I will say below.

We now have the technology to predict many social phenomena, if we can catch them at the right moment.I follow Theodore Modis's 1992 book for an accurate summary of the state of forecasting utility and reliability.Modis proposes three types of ordering in larger networks of human interaction.Each constitutes a range of predictability at a particular time.He applies this research to the fields of economics, social infrastructure, and technology, and I believe his findings apply to organic systems as well.The three scopes of Modus are: Invariants, Growth Curves, and Cyclic Waves. invariant.The natural, unconscious tendency of all organisms to optimize their behavior gradually imbues it with "invariants" that change very little over time.Humans, in particular, are the most qualified optimizers.Twenty-four hours in a day is an absolute invariant, so generally speaking, although the intervals and completed careers are not the same for decades of life, it is obvious that human beings tend to spend a certain amount of time on Do the chores: cooking, traveling, cleaning.If you incorporate new behaviors (e.g., taking flight 0201483408 instead of walking) into the basic dimensions (e.g., how long does it take to commute each day), you will see that the pattern of this new behavior continues to exhibit the same pattern of the original behavior. A pattern can likewise predict (or predict) its future.In other words, you used to walk half an hour to work every day, but now you drive half an hour to work.And in the future, maybe you'll fly half an hour to work.The market's pressure for efficiency is so relentless, so relentless, that it necessarily drives man-made systems in a single (predictable) direction of optimization.Tracking down the optimization point of an invariant often alerts us to the predictability bounds of a rule.For example, mechanical efficiency increases very slowly.Until now, no mechanical system has achieved more than 50% efficiency.It is possible to design a system that operates at 45% efficiency, but it is impossible to design a system that operates at 55% efficiency.Therefore, we can make a reliable short-term prediction of fuel efficiency.

growth curve.The larger, more layered, and more decentralized a system is, the more progress it has made in terms of organic growth.All growing things share several common characteristics.One of them is the life cycle in the shape of an S-curve: slow birth, rapid growth, and slow decay.The annual production of cars worldwide, or the symphonies composed in Mozart's lifetime, would fit this S-curve fairly precisely. "The predictive power of the S-curve is neither magical nor useless," Modis wrote. "Behind the graceful shape of the S-curve hides the fact that the natural process of growth obeys a strict law." This law states that the shape of the end is symmetrical to the shape of the beginning.This law is based on thousands of empirical observations over the history of biology, and the history of life that has formed institutions.This law is also closely related to the natural distribution of complex things represented by bell curves.Growth is extremely sensitive to initial conditions; however initial data points on a growth curve are almost meaningless.However, once a phenomenon has formed an unstoppable trend on the curve, a digital snapshot of its history is formed and can play a subversive role in predicting the ultimate limit and demise of the phenomenon.From this curve, one can extract an intersection point between it and competing systems, or an "upper limit", and the data that this upper limit must be pulled horizontally.Not every system's life cycle presents a smooth S-shaped curve; however, the types and numbers of systems that fit this curve are considerable.Mordis argues that there are more things that obey this law of growth than we think.If we examine such growing systems at the right time (in the middle of their growth process), the emergence of this locally ordered state, as outlined by the law of the S-shaped curve, provides us with an additional range of predictability.

cycle wave.The apparently complex behavior of a system partly reflects the complex structure of the system's environment, which Herbert Simon pointed out some thirty years ago.At that time, he used the trajectory of an ant on the ground as an example.An ant's weaving path across the ground reflects not the complex movements of the ant itself, but the complex structure of its environment.According to Modis, cyclical phenomena in nature can instill a cyclical bias into the systems in which they operate.Modis was fascinated by the 56-year business cycle discovered by the economist Kondrakiev.Moreover, in addition to this economic wave discovered by Kondrakiev, Modis also added two similar cycles, one is the 56-year cycle in scientific development proposed by himself, and the other is Arnov Gruber Study the 56-year cycle of infrastructure replacement.Various hypotheses have been proposed by other authors to account for these apparent fluctuations, with some suggesting that it comes from the 56-year lunar cycle, or the fifth 11-year sunspot cycle, and others have even attributed it to human Intergenerational cycles - as each 28-year generation cohort deviates from the work of its parents.Modis argues that the primary environmental cycle triggers many subsequent internal cycles of secondary and regenerative processes.Once researchers spot any fragments of these loops, they can be used to predict ranges of behavior.

The three predictive patterns described above suggest that at certain moments when the system increases its visibility, invisible patterns of order become clear to followers.It's like the next drum beat, and you can almost hear in advance what it's going to sound like.After a while, disturbances muddy it and the pattern disappears.The predictability range also has big surprises.However, the local predictability does point to some methods that can be improved, deepened, and extended into something bigger. Although the odds of successful large-scale forecasting are slim, amateur and professional financial chartists alike are not discouraged from trying to extract long-wave patterns from past stock market prices.For chartists, any kind of outward cyclical behavior is easy prey: the length of a skirt, the age of a president, the price of eggs.Chartists are forever chasing the mythical "leading indicators" that predict stock price trends and use them as numbers to bet on.For years, chart analysts have been ridiculed for adopting this dubious approach to number logic.In recent years, however, academics such as Richard Sweeney and Blake LeBaron have demonstrated that the chartist's approach often works.A chartist's technical rule can be astonishingly simple: "If the market has been trending up for a while, bet it will keep going up. If it's in a downtrend, bet it will keep going down. "Such a criterion reduces the high dimensions of a complex market to the low dimensions of such a simple two-part rule.In general, this approach to pattern finding works well.This "goes up when it goes up, and goes down when it goes down" model works better than random luck, so it is much stronger than the hype of ordinary investors.Since the most predictable thing about a system is its stagnation, the emergence of this orderly pattern is not surprising—although it is.

In contrast to chartism, other financial forecasters rely on the "fundamentals" of the market to predict the market.These people, known as fundamental analysts, try to understand the driving forces, underlying dynamics, and fundamental conditions in complex phenomena.In short, what they are looking for is a theory: f=ma. Chartists, on the other hand, look for patterns in data and don't care whether they understand why the pattern exists.If there is indeed order in the universe, then the future paths of all complex order are (at least temporarily) revealed somewhere and somehow.One just needs to understand what signals can be ignored as noise.The chartist organizes in the Don Farmer way.By Farmer's own admission, he and his colleagues at forecasting firms are "statistically strict chartists."

In another fifty years, computerized induction, algorithm-based graph analysis, and predictability horizons will be respectable human endeavors.Forecasting the stock market is still an oddity, because the stock market is based more on expectations than any other system.In a game of anticipation, accurate predictions offer no chance of making money if everyone shares the prediction.Forecasting that all a company can really have is a lead in time.As soon as Farmer's team made a lot of money developing a certain predictive range, everyone else would rush in, somewhat blurring the pattern and, in most cases, leveling the chances of making money.In a stock market, success inspires a strong, self-cancelling flow of feedback.In other systems, say a growing network, or a company that is expanding, predictive feedback does not cancel itself.Typically, feedback is self-managed.