Tim Palmer turned uncertainty into the standard way the weather is forecast. Now he is staking a theory of quantum physics on the same idea, and a machine will settle the bet.
A forecast that gives a 40 percent chance of rain is making an admission. It does not know whether it will rain, and it says how much it does not know. Numbers like that come from a method called ensemble forecasting. The forecast is run many times, each run starting from a slightly different state within the range the observations allow, and the runs that end in rain are counted. When the runs agree, the forecast can be trusted; when they scatter, the scatter is the honest answer. [Sourced · WMO, "Executive Council honours IMO laureate Timothy Palmer," June 2024]
Tim Palmer led the building of that method into working forecasts. He trained as a cosmologist. His Oxford doctorate, finished in 1977, was in general relativity under Dennis Sciama, who also supervised Stephen Hawking. He then moved to the UK Met Office, and in 1986 to the European Centre for Medium-Range Weather Forecasts, where he led its ensemble system until 2011. He is now Professor of Climate Physics at Oxford. In 2024 the World Meteorological Organization gave him its IMO Prize, which the Royal Meteorological Society called "the equivalent of a Nobel Prize for meteorologists." [Sourced · WMO 2024; Royal Meteorological Society, 1 July 2024]
The reason to doubt a forecast at all goes back to Edward Lorenz. His butterfly is famous: a tiny difference in a starting state grows until it decides the weather. Less well known is a stronger claim Lorenz made in 1969 about fluids that move at every scale at once, from gusts to storm systems. There, a small error does not simply grow. It climbs from small eddies to large ones, and the forecast loses all skill after a fixed time "which cannot be lengthened by reducing the amplitude of the initial error." Palmer calls this the real butterfly effect. Under it, measuring the starting state more precisely buys no extra time. [Sourced · Palmer, "The real butterfly effect and maggoty apples," Physics Today 77(5), May 2024; Palmer, Döring & Seregin, Nonlinearity 27, R123, 2014]
Palmer borrows an image from the physicist Michael Berry to explain why. Biting an apple with a whole maggot in it is worse than biting one with half a maggot, and a tenth of a maggot is better still, but no fraction, however small, is the same as no maggot at all. An arbitrarily small error is a different thing from zero error. Only a perfect start would allow unlimited prediction, and no start is perfect. [Sourced · Physics Today, May 2024]
Whether the equations of fluid motion really behave this way is not proven. In 2024 Palmer wrote that the question is related to one of the Clay Mathematics Institute's million-dollar Millennium Prize Problems, the one The Penumbra covered in Issue 4, when AI agents produced a claimed solution showing that the Navier–Stokes equations can break down from a calm start. Breaking down and losing predictability are different questions. On whether the real butterfly effect is a strict property of the equations, Palmer's own summary was "No one knows." The September result has not changed that. [Sourced · Physics Today, May 2024; The Penumbra, Issue 4]
Invariant set depicting a fractal. Each circle contains 7 smaller circles that spiral forever revealing a pattern in which the large resembles the small but doesn’t repeat.
The present test of Palmer's views is coming from software. When machine-learning weather models arrived in 2023, he called them an "unanticipated shock" and "an interesting new player on the block." Some now rival physics-based ensembles on ordinary days, and how they handle extreme weather is still being measured. His response is to build physics models fine enough to resolve weather a kilometer across, combine them with AI, and stop duplicating effort across countries: "we do not need to maintain multiple models from different centres to represent model diversity. This is holding back progress." [Sourced · WMO, June 2024]
Palmer's second idea runs the other way from the first. Uncertainty is a limit to respect, and it is also a resource to use. His group has argued for adding deliberate randomness to the small-scale parts of weather models that cannot be calculated exactly, and for computing with less numerical precision so the savings can be spent on finer detail. In 2015 he and the neuroscientist Michael O'Shea took the idea into the head. The brain runs on about 20 watts, and its neurons are small enough that heat makes them noisy. Creative thinking, they proposed, comes from "a synergy between low-energy stochastic and energy-intensive deterministic processing": loose, noisy search turns up candidates, and slow, deliberate thought checks them. [Sourced · Palmer, "Build imprecise supercomputers," Nature 526, 32, 2015; Palmer & O'Shea, Frontiers in Computational Neuroscience 9, 2015; ECMWF symposium summary, Feb 2023]
Put the two halves together and Palmer has a view of how anything new comes to be known. An ensemble forecast is many slightly wrong guesses, kept honest by comparing their spread against what actually happens. A creative mind, on his account, works the same way: noise proposes, checking disposes. If that is right, the growth of knowledge depends on keeping some doubt in the system and testing what it produces. Whether it is right is a question about brains that no one has answered, and the colleagues who summarize these ideas present them as speculative. [Sourced · ECMWF symposium summary, Feb 2023]
The third idea is the boldest. A chaotic system left to run settles onto a set of states it keeps returning to, called an attractor. For Lorenz's equations that set is a fractal: a shape with detail inside detail at every scale and gaps everywhere between. Palmer proposes that the possible states of the whole universe lie on such a set, and that states off it are not physically real. Those excluded states include some of the "what if it had been measured differently" cases that the puzzles of quantum physics rely on. In March 2026 he published a sharper version in PNAS, which he calls rational quantum mechanics. It removes the continuum, the infinitely fine gradation of values, from quantum theory: "Nature abhors a continuum." It also makes a prediction. With more than a few hundred error-corrected qubits, the working units of a quantum computer, entangled together, the best-known factoring algorithm should begin to fail. Palmer expects the test to be possible within a few years. [Sourced · Palmer, "Rational quantum mechanics: Testing quantum theory with quantum computers," PNAS, 2026; University of Oxford Department of Physics, 16 March 2026]
Most physicists do not accept the move at the center of this. Bell's theorem, the result behind the quantum puzzles, assumes that what an experimenter chooses to measure is independent of the hidden state of what is measured. Theories that give up that assumption are called superdeterministic, and critics regard them as needing a conspiracy between the experimenter's choices and the particles. In 2021 the physicist Indrajit Sen built an explicit model of Palmer's proposal and argued that it ends up nonlocal anyway, so that "superdeterminism is unlikely to solve the puzzle posed by the Bell correlations." Jonte Hance, Sabine Hossenfelder and Palmer replied that the model misrepresents the theory; Sen replied that they had misread it. The dispute is not settled. The 2026 prediction is the cleaner test, because a quantum computer will either factor past the limit or it will not. [Sourced · Sen, arXiv 2107.04761; Hance, Hossenfelder & Palmer, arXiv 2108.08144; Sen, arXiv 2109.11109]
For a first feel of the man rather than the papers, the best entry is the interview in Sabine Hossenfelder's Existential Physics (2022), titled "Is Math All There Is?" Hossenfelder is his co-author on superdeterminism, so read it as a conversation between allies. Later in the same book she interviews David Deutsch under the title "Is Knowledge Predictable?", which is the question Palmer's work keeps circling. Across the sky, the head and the quantum, his career makes one claim three times: what cannot be known exactly is not a flaw in the picture but part of what the picture is made of. To follow it, search for the real butterfly effect.
---
Sources- WMO, "Executive Council honours IMO laureate Timothy Palmer," June 2024 — https://wmo.int/node/22907- Royal Meteorological Society, IMO Prize news, 1 July 2024 — https://rmets.org/news/former-rmets-president-honoured-international-meteorological-organization-prize-laureate- Palmer, "The real butterfly effect and maggoty apples," Physics Today 77(5), 2024 — https://physicstoday.aip.org/features/the-real-butterfly-effect-and-maggoty-apples- Palmer, Döring & Seregin, "The real butterfly effect," Nonlinearity 27, R123 (2014) — https://ui.adsabs.harvard.edu/abs/2014Nonli..27R.123P- Palmer, "Build imprecise supercomputers," Nature 526, 32 (2015) — https://www.pubmed.gov/26432226/- Palmer & O'Shea, Frontiers in Computational Neuroscience 9 (2015) — https://frontiersin.org/journals/computational-neuroscience/articles/10.3389/fncom.2015.00124/text- ECMWF symposium summary (Mason), Feb 2023 — https://www.ecmwf.int/node/28659- Oxford Physics, "Rational quantum mechanics," 16 March 2026 — https://www.physics.ox.ac.uk/news/rational-quantum-mechanics-new-theory-quantum-physics- Sen, arXiv 2107.04761; reply arXiv 2109.11109 — https://export.arxiv.org/abs/2107.04761 · https://ar5iv.labs.arxiv.org/html/2109.11109- Hossenfelder, Existential Physics (2022), contents — https://existentialphysics.com/content/- Wikipedia, Tim Palmer (physicist) — https://en.wikipedia.org/wiki/Tim_Palmer_(physicist)