One of my favorite quotes – attributable to both baseball Hall-of-Famer Yogi Berra and Nobel-Prize-winning physicist Niels Bohr – has become only more relevant as new technologies get wider adoption: “It’s hard to make predictions. Especially about the future.”
We want to know the future, but by definition, there’s uncertainty about what may happen. That’s why the Urim and Thummim, two of the items that appear in this week’s portion, Tetzaveh, are so striking:
Inside the breastpiece of decision you shall place the Urim and Thummim, so that they are over Aaron’s heart when he comes before God. Thus Aaron shall carry the instrument of decision for the Israelites over his heart before God at all times. (Exodus 28:30)
In the Torah (and in later Biblical texts), the Urim and Thummim have a fairly narrow definition and role – primarily, they were used by kings to guide military actions and in moments of clear national need. But the Talmud, in Yoma 73b, expands the predictive power of these instruments, building on their definitions: “Urim from ‘Or,’ light – because they illuminate their words. Thummim from ‘Tam,’ complete – because they fulfill their words, which always come true.” They provide clear and unambiguous guidance, telling us not just what to do, but what will actually happen.
For most of us, and (if we’re being honest) for most of the political pundit class, predictions are fairly vague. We are often told that a new update of artificial intelligence, or a special Congressional election, or an international incident, will have “long-lasting and far-reaching implications.” But how long? How far? Those predictions are actually quite easy to make, since we can so effortlessly redefine the parameters of what we’re looking for.
In contrast with the imprecise predictions we laypeople might make, a company called Metaculus has been running “prediction tournaments,” where people attempt to predict the answers to very specific and precise questions, often with cash prizes for those who get it correct. And while ordinary people may want to develop a gut probabilistic sense on something we may know and care about (such as the likelihood of Gavin Newsom or Alexandria Ocasio-Cortes becoming the Democratic nominee for President in 2028), this tournament instead asks questions that would require a significant amount of knowledge on a variety of diverse topics: How many maritime piracy and armed robbery incidents will the International Maritime Bureau report for Q1 2026? Will the nationwide #1 party in the Netherlands get at least a 15.0% share in the municipal elections? Will the English Wikipedia have at least 7,145,000 articles before March 1, 2026?
Perhaps not surprisingly, A.I. models are now doing better at predicting the future than even elite human forecasters. After all, the models focus on quantitative measures involving vast reams of data. And in a recent piece in The Atlantic, Ross Anderson shares a predication that feels very likely to come true:
If the A.I. takes gold [in the Metaculus Spring Cup for 2026], that might signal a new era. Human beings—predictors of eclipses, theorists of cosmic heat death—may no longer be the best guides to the future. From this point on, for as long as we exist, we might be asking A.I.s what comes next. We won’t always understand how they arrived at their predictions. This crystal ball may be like a black hole with an event horizon, past which the light of its insight cannot escape. We may just have to take it at its word.
Yet, there’s a crucial difference between what an A.I. prediction machine does and how the Urim and Thummim function: the Urim and Thummim sit “over Aaron’s heart.” The decision here isn’t made by the device alone; it’s by Aaron, who carries it. The Urim and Thummim don’t replace his judgment – they inform it. And when the king needs to consult them, the priest mediates. There are humans in the loop at every step.
Not only that, the Rambam, in Hilchot Klei HaMikdash 10:11, is explicit that the Urim and Thummim were not available to just anyone; they could only be consulted by the king, the head of the Sanhedrin (the Jewish Court System of the Second-Temple Era), or someone whose need rose to the level of national consequence. Not everyone got to ask, and not every question was appropriate to pose to it. And the answer, when it came, still had to be interpreted, weighed, and acted upon by people who would live with the consequences.
Ross Anderson’s image of the crystal ball with an event horizon – past which the light of its insight cannot escape – scares us a bit, precisely because it entails the absence of humans. One of the hallmarks of artificial intelligence is that we humans can’t understand how and why it makes its decisions – it’s a black box, and when we can’t see how a prediction was made, we can’t evaluate whether to trust it, and we can’t identify where it might be wrong. We are asked to simply take it at its word.
Jewish tradition has always been suspicious of oracles that remove human moral agency from the equation. That’s part of what distinguishes the Urim and Thummim from the divination practices the Torah explicitly forbids, such as consulting spirits and the dead. Those practices promised unmediated access to hidden knowledge, a shortcut past the hard work of human deliberation.
Which brings us back to the problem of predicting the future. Predictions aren’t just technical problems to be solved with better models and more data. We always act under uncertainty and make choices that matter precisely because we don’t know how they’ll turn out; we are always responsible for what we do with imperfect information. The Urim and Thummim didn’t eliminate that uncertainty. They helped a community navigate it together, with a priest carrying the names of all twelve tribes over his heart so that no single person’s interest could crowd out everyone else’s.
As A.I. forecasting tools become more accurate and more widely used – in medicine, in finance, in policy, and in our own personal decisions – the question we need to ask is not whether to consult them. Of course we will; we have always wanted to know about the future, and that urge isn’t going away. The question is: Who is carrying these tools close to their heart, as Aaron did? Who bears responsibility for the questions we ask and the answers we receive? Most importantly, who gets to decide what we do next?