Skip to main content
Reading Time: 6 minutes
How to Predict the Future
How to Predict the Future

Predicting the future is usually associated with psychics and seers like Nostradamus or their less paranormal but often equally wrong cousins, the futurists, but in fact we predict all the time. We know that day will follow night and so we can predict that the sun will rise. When we drive, we can predict when the light will turn red or, based on their turn signal, whether the other car will turn right or go straight. We barely think about these predictions because they are mundane and done almost automatically.

When we venture into more complex predictions, greater skill and data is required than simply observing a car’s turn signal. For example, predicting the weather requires complex and precise observations of air pressure, wind direction, moisture, and so one. We use satellites and complex computer algorithms to try to peer into the future but even then meteorolgists are frequently wrong.

Thanks to Chaos Theory and the sensitive dependence on initial conditions, predictions become harder the more complex the system is and the farther out in time one wishes to predict. It is never possible to quantify all variables precisely enough and so there is always a risk of error. Just like the weather, scientists and forecasters have trouble predicting earthquakes, complex human behavior, and economic trends. The answer has been to try to gather more and more data, to analyze it more and more deeply with AI, with data mining, and complex algorithms. Even so, our ability prognosticate has been limited.

I believe, however, that there is another way of predicting the future which has been largely overlooked and which embraces chaos and chance rather than trying to fight it with more and more data.

The root of this method of predicting the future, which I will describe, lies in the Million Monkey Theorem. This thought experiment posits that if you had a million monkeys typing randomly long enough they would eventually produce all of the works of Shakespeare by pure chance. Of course, it might take a near infinite length of time if we relied on the truly random key pounding by agitated monkeys.

The Million Monkey Theorem: An Unexpected Way to Predict the Future

However there is a better way. The original Million Monkey Theorem was developed in the age of typewriters when the only random text that could be created would by single letters. However, we have progressed beyond that. Computers can now produce random text and numbers but we can give them grammatical rules and dictionaries so that the randomly generated text is more likely to make some sense. For example, instead of waiting for a random string of letters to randomly come together to create a real word, we can program the computer to generate only words fond in the dictionary and even string them together subject to some rules, such as requiring at least one word to be a noun and the other to be a verb. In this way the output is more likely make some sense. Less work for the monkeys and less waiting by us.

Given enough time, or enough computing power, you could generate an infinite amount of text. But for our purposes, infinity is not required.

The Million Monkey Theorem: The Key to Predicting the Future
The Million Monkey Theorem: The Key to Predicting the Future

The results created by our Million Computers would vary from incomprehensible gibberish to profound philosophical utterances. In time, they would even re-create the works of Shakespeare. In fact, you could even speed up the process by giving the computer certain grammatical and stylistic rules to create Shakespearean-type plays. The result would be many Shakespearean forgeries, some good and most awful. But among the chaff of millions of pages of random text would be the real thing, which we could find by using programs such as ones that currently check for plagiarism and comparing the output text with the real Shakespeare.

The method above is well known and in fact there are many online generators out there that build poems and even Shakespearean plays. The most effective ones are programs that use AI or neural networks to approximate what Shakespare might have written.

What then does this have with predicting the future? Everything.

We do not need to limit our Million Computer Monkeys to re-creating the works of a dead playwright. We can set them lose on creating useful things. Imagine, for example, if we asked the Million Computer Monkeys to write a biography about you, documenting the the story of your life from beginning to end. In among millions of erroneous biographies and histories that do not describe you at all, would be one which describes every important detail of your history as well as everything that has yet to happen to you including the date and time of your death.

Of course, we could never tell which predictive biography was true until after the fact. Imagine for example, we ask the Million Computer Monkeys to write a text using random output stating when and how you will die. Morbid, I know. The output will generate many fictions, and one truth, but we will not be able to now which “prediction” was accurate until after the fact. This makes the generation of random biographies useless as a tool for predicting the future because the accuracy cannot be tested except retrospectively.

But there is another way that randomly generated text can be used to predict the future, which I think has been overlooked: by generating things that can be tested here and now.

The Method Described

The future as predicted by an AI art Generator
The Future According to an AI Art Generator: Apparently Lots of Abandoned Cars and Enormous Flying Grapes

This is the heart of the method I propose: assume that the Million Computer Monkeys have been programmed to generate random science manuals, or physics equations, or instructions for creating miracle drugs or substances. You can limit their output to this kind of parameter, just like present day neural networks can be programmed to produce blog posts, news articles and even scholarly articles. The key difference here is that you are asking the program to create text about something that can be tested here and now, but which is not currently known, and about which nothing has been published.

For example, imagine a Million Computer Monkeys being asked to generate chemistry manuals. Most will be useless, containing wildly wrong formulas and methods. Others will randomly reinvent the wheel, even recreating now obsolete texts of alchemy or from the Victorian era. But in among the huge haystack will be monkey gold, a glimpse into a future discovery not yet achieved, such as instructions on how to create advanced superconductors, quantum computers beyond our comprehension, and so on. In essence, the random text output would serve as a tool to predict the future, bringing to our present, knowledge that is beyond our time.

The real challenge would be to discover what amounts to a prediction of the future and what is just error-strewn gibberish. But unlike biographies and future history, text that describes a concrete, testable formula or process, can be verified now rather than waiting for the future to happen. Assuming that a program can be developed to simultaneously identify and test promising formulas or inventions at high speed and volume, we could data mine this chaotic output to find the hidden gems, essentially using the Million Monkeys to predict the future.

In fact something like this already exists. A random generator called This Formula Does Not Exist spews out random graphics of non-existent chemical compounds. Most of these compounds could not exist, or are useless, but somewhere among its random output may be the formula for a compound that cures cancer, or can get that gravy stain out of your favorite shirt. The problem is how to find the useful information, to essentially reverse engineer what is being given to us.


Art Produced by the Night Cafe AI in Response to the Prompt: “a beautiful tomorrow” (colourful beautiful watercolor)

A Side Note

All of the images in this article were created using the AI art generator found at NightCafe.studio which makes artwork from text prompts. Although the images are often surreal or impressionistic, sometimes they are uncannily realistic and may in and of themselves suggest another method of predicting the future, by creating actual images of what lies ahead. But that discussion is for another article.