AI and Prediction: When Technology Redefines the Future

From Tea Leaves to AI: A New Era of Prediction

In a classroom filled with executives, the use of AI was discussed with familiar examples: healthcare professionals using it to analyze medical images, managers employing it to draft emails, and a retail company that attempted to use it for note-taking during meetings before realizing that AI was confabulating and lacked contextual understanding. However, one intervention captured everyone's attention: a middle-aged Asian woman, dressed in a beige cardigan and white sneakers, claimed to use chatbots like oracles. She revealed that she had built a billion-dollar empire and that AI had recently accurately predicted a 2% rise in the stock market, causing a nervous murmur in the room.

Modern seers are no longer astrologers or economists, but computer scientists, data analysts, and engineers. Algorithms have become the new tea leaves, animal entrails, and stars through which we hope to glimpse the future. While we often associate predictions with knowledge, they are often closer to the realm of power. Prophecies become arenas where battles for the future unfold, and our expectations tend to bend the social world toward these predictions. When someone forecasts that the world will be a certain way, they prompt others to realize that world. Despite the significance of prediction in our lives, it is remarkable to note that while thousands of books have been written on how to predict, very few address the ethics of prediction.

The Flourishing Prediction Industry

Prediction has become a major industry. Platforms like Polymarket aggregate public expectations regarding future events, collecting vast amounts of data and creating influence. If 58% of users believe the Oklahoma City Thunder will win the NBA championship, why bet against the majority? But betting on these platforms goes far beyond sports or even reality TV. They have transformed political instability, natural disasters, and human suffering into a spectacle, dehumanizing the real victims and gamifying life.

Today, predictions have evolved into weapons of power that justify value-laden decisions under the pretense of facts, but predictions are never facts. Facts belong to the present and the past. A statement about the future can be many things: an estimate, a desire, a warning, but never a fact. What makes the future is that it has not yet occurred. What has not happened does not exist, and there are no facts about what does not exist. Yet, we use prediction more than ever with AI, prediction markets, and experts speaking about the future.

The Fantasy of Conquering Uncertainty

Pierre-Simon Laplace had a dream, often referred to as Laplace's demon. It occurred to him that with enough data and computing power, it would be possible to achieve complete knowledge. If you knew the exact location and momentum of every particle in the universe, as well as all the laws of nature, then you could predict the future with perfect accuracy. Uncertainty would finally be vanquished. As Laplace said:

"Given for an instant an intelligence that could comprehend all the forces by which nature is animated and the respective situation of the beings that compose it -- an intelligence vast enough to submit these data to analysis -- it would embrace in the same formula the movements of the greatest bodies of the universe and those of the lightest atom; for it, nothing would be uncertain and the future, like the past, would be present to its eyes."

AI proponents may not phrase it this way, but what they seem to suggest when they enthuse about the power of machine learning combined with vast amounts of data is that these technologies bring us closer to realizing Laplace's demon. If we can collect every data point, the idea is that we can build enough computing power to analyze this data, allowing us to predict what was previously unpredictable. This predictive power promises to revolutionize all fields of knowledge, from medicine to climate change and politics.

Driven by this fantasy, quantifiers follow your every move; recording, tabulating, and exhaustively analyzing your pleasures and vices; torturing your data until it screams in confession. You are tracked while you drive, search online, exercise, have sex, drink alcohol, consume drugs, travel, sleep, talk with friends and family, spend time on social media, go to the doctor, play online games, read, watch television, and breathe.

We manage and discuss our fears in quantified terms: the probability of contracting cancer, being robbed, that earthquakes will occur, or that another pandemic will arise, that climate change will render our world unlivable, or that another world war will break out. The unbridled optimism to conquer uncertainty through AI is understandable. Computers, data, and statistics have brought incredible breakthroughs. The Bombe computer broke the Nazis' Enigma code. In medicine, regression analysis has been crucial in identifying risk factors for diseases. Mainframe computers have provided new insights into business; centralized data processing has enabled real-time transaction processing and scalability. Manufacturing companies have gained the ability to monitor production efficiency across entire supply chains, identifying bottlenecks and improving resource allocation.

Personal computers emerged in the 1980s. The 1990s and 2000s saw the rise of the Internet and cloud computing, further increasing data availability and processing power. The 2010s marked a turning point with the practical application of deep learning, fueled by large amounts of data and improved hardware like GPUs. Advances in algorithms paved the way for machine learning -- predictive machines.

AI and Prediction: A Power Game

With prediction comes all the patterns of prophecy and power that line our history books. The difference is that AI is prediction on steroids, and we use it not only on the battlefield and in the medical cabinet but everywhere, from the office to the classroom, the courtroom, our roads, our love lives, and beyond.

Machine learning algorithms are predictive machines. That’s all they do, whether they engage in regression, classification, or language. When a machine learning system translates text, it predicts the most likely translation based on millions of examples of previous translations. When it recognizes wolves in photos, it does so by predicting the probability that a given image contains a wolf, based on patterns it has learned from thousands of labeled wolf and non-wolf images. When a large language model answers a question, it predicts what a human would say in its place, based on statistical analysis of books, online forums, social media, and more.

It is not surprising that an "oracle" is a technical term in the context of machine learning. An oracle represents the best possible performance that could be achieved; it is an idealized function that always provides perfect predictions. The triumph of machine learning is a corporate victory far more than a scientific victory. Idealists might find it anticlimactic, even depressing. Someone might bluntly say that we simply threw money at the problem.

What is most remarkable about the success of machine learning is how it has occurred in an unremarkable way. "What is disappointing," said Michael Wooldridge, an AI professor at Oxford, to a group of my MBA students, "is that it did not happen as a result of a scientific breakthrough." He looked around to ensure the weight of his words was felt.

From the 1960s to the early 2000s, the results of neural networks were not very impressive. The symbolic AI group was winning the race and the grants -- until it was no longer the case. Something changed: we obtained more data and more computing power, and machine learning took off. In just a few years, machine translation, for example, went from an unusable state to understandable, then good enough to help lost tourists find their way without knowledge of the local language. It is now good enough that I admit I sometimes preferred a machine translation to the suggestions of a professional translator who had a weakness for verbosity.

The incredible things that machine learning can do did not happen due to a better understanding. It did not require genius. The picture is darker than a lack of inspiration. The means by which such brute force in data and computation was acquired involved theft, exploitation of vulnerable people, fierce use of natural resources, and the construction of a mass surveillance architecture, to name just a few. We may be centuries away from the oracles and astrologers who preceded algorithms, but prediction still primarily concerns power. Power is what enables you to obtain predictive algorithms, and more power is what they grant you in return.