Clairvoyance And Control

“Prophecy: Prediction, Power and the Fight For the Future, From Ancient Oracles to AI

 

By Carissa Veliz

 

Doubleday, 384 pages, $35

 

Oil and gas prices have been so erratic lately that the time-honored roller-coaster metaphor now looks sedate. Yet none of the price shifts have been in response to actual supply. Instead, the market has been making bets on what it thinks the consequences of the Middle East war will be.

 

Traders rely on prediction in the most unpredictable of circumstances. Energy markets are far from unique, as Carissa Veliz, a professor at Oxford University's Institute for Ethics in AI, shows in "Prophecy." Her sweeping account of prediction across history demonstrates why we would do well to approach most forecasts with the skepticism we now show to prophets.

 

The early historical chapters on prophecy and the growth of statistical and scientific prediction give more back story than perhaps we need, fun though the tales are of seers being executed for being wrong -- or even for being right. The moral of these chapters is that prophecy has always been about power. Those who are thought to be able to see the future are granted more control in the present, regardless of whether they in fact can tell what will be.

 

We see that same pattern playing out now, argues Ms. Veliz, taking aim at big tech, the target of her first book, "Privacy is Power" (2020). Governments are so in awe of the tales weaved by tech bros about the coming artificial-intelligence future, she says, that they have given them a free pass to do pretty much whatever they want. The result has been an astonishing transfer of power away from the state and into the hands of a small group of wealthy individuals.

 

Ms. Veliz's declaration that "predictions are not facts" might sound like a statement of the obvious, but it is truth often forgotten. Take the outputs of large language models. These are nothing but statistical guesses about word sequences. Yet too many users treat them as sources of information. Already we are seeing grave consequences of people taking AI replies to health queries as though they were truths.

 

Sometimes, however, we can actually do things with words. As J.L. Austin, an English philosopher of language, observed, words often convey more than information. Predictions, for instance, are "speech acts" that shape the future. Provide someone with a professional reference, predicting that they will be a perfect fit for a job, and they may get that job. Predict a downturn in demand and you may trigger a decrease in supply as people change what they produce in anticipation. Speech-act theory explains why predictions like these can become self-fulfilling prophecies.

 

"Predictions are commands disguised as descriptions," Ms. Veliz warns, "and the more we allow companies and governments to use prediction across society, the more our future is being decided by them." When big tech predicts that everyone will need AI systems to future-proof their businesses, it is effectively telling people to invest in AI, or else. The author makes a strong case that since prophecy shapes the future, we need to take the ethics of prediction seriously.

 

She wants us to see most prediction for what it is: an educated guess, which is no more factual when it is based on numbers, analysis or experience. She argues that too often we use expert prediction as an excuse not to think for ourselves, deferring to impressive-looking charts and graphs as a form of anxiety relief.

 

"What we deem inevitable often never comes to pass, while what we thought impossible awaits around the corner," Ms. Veliz writes. She gives a forensic account of why the data and the facts we base our predictions on are always incomplete and very often of poor quality. What's more, the world is more complicated than any model can capture, so no future scenario is exactly like the ones from the past that we draw upon for guidance.

 

Ms. Veliz acknowledges that we cannot live without prediction. According to the most credible theory in neuroscience today, the brain is itself a prediction machine; without its ability to project onto our perceptions what is has learned to expect, we couldn't function. But the author offers little advice on how to use prediction better. She explains how, with healthcare at full stretch, doctors' predictions often determine who will receive treatment and who won't, with life or death consequences. But resources will always be finite, and it's not clear how doctors could make such agonizing decisions better. The book's epilogue, "Ten Lessons in Prediction," focuses on the merits of avoiding prediction when possible, rather than how to use it better.

 

Ms. Veliz warns that her book straddles genres and defies classification. It is also a little disjointed. Chapters are full of breaks, sometimes connected by clunky transitions but more often with little attempt to stitch the sections together. Her final chapter, on how to thrive amid uncertainty, is sensible enough, but is more of a short treatise on the good life than an analysis of the appropriate role of prediction.

 

These flaws can easily be overlooked because, once the book gets going, its insights, provocations and vivid examples are presented with both passion and clarity of thought. Whether or not it changes how you think about prediction, I cannot say, but I am left convinced that it should.

 

---

 

Mr. Baggini is the author of "How the World Eats: A Global Food Philosophy."” [1]

 

1. Clairvoyance And Control. Baggini, Julian.  Wall Street Journal, Eastern edition; New York, N.Y.. 22 Apr 2026: A15.