Strange Loops
Feb. 24th, 2025 01:54 pm![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
qatsiBook Review: The Primacy of Doubt, by Tim Palmer
Poor Michael Fish. He will be forever remembered as the forecaster who dismissed the possibility of a hurricane hitting the UK in 1987. He features prominently in this book, but not alone. Ironically, the Met Office was developing its own ensemble forecasting at the time, and running the data with hindsight through those models, the forecast was unusually chaotic and unpredictable.
Palmer's career has weaved between quantum physics and meteorology, and I suppose it's therefore natural that he finds the connections between those two topics. "Sensitive dependence on initial conditions" - the unpredictability popularly known as the butterfly effect - features large, and is well explained. The solution - at least as far as understanding the potential for error in a forecast - can often be to run an ensemble model, with a range of close but not identical initial conditions, and consciously to throw in a bit of noise, to discover whether the predictions lie on a smooth part of the phase space, with little variation, or in a more chaotic realm.
Palmer stretches out into other areas more or less connected to the above, such as climate change, the Covid pandemic, and economics. In the first two areas, scientists find this approach useful: it indicates the degree of uncertainty in predictions, as well as showing plausible worst cases. There's an amount of separation here between science and policy: "Following the science" is a bit rubbish, as well as mendacious, because the science only offers projections on likely outcomes if a particular path is followed. On economics he finds tougher ground: although some theorists have found the ensemble approach useful, it is not in vogue and hard to get such research published. The implication is of faddishness or denial in some academic circles. It's certainly true that Mandelbrot considered himself an irritant to the economic establishment.
The fundamental problem with models in various areas is that of agency, so that the predictions or projections of a particular model influence behaviour, providing feedback into the model - it's exactly this non-linearity that gives rise to chaos. Chaos, in the sense that our mathematical frameworks aren't very good at handling this. The three-body problem is an early example - chaotic and ultimately unpredictable motion, yet somehow nature does its sums and finds a path. In the third section of the book, which Palmer marks as "more speculative", he investigates quantum uncertainty and free will. There are some interesting ideas about chaotic attractors and fractals, which earlier parts of the book have explained well. The Cantor set somehow fills a space, yet it is empty (insofar as picking a random point in that space almost certainly does not lie in the set). Synthesising these ideas as an explanation of quantum uncertainty, giving rise to probabilistic observations, is clever, as is the idea of quantum noise as a source of inspiration and free will. We are in the same zone as Hofstadter's strange loops. Some of the most speculative ideas in the final chapter seem off, but credit for trying.
Poor Michael Fish. He will be forever remembered as the forecaster who dismissed the possibility of a hurricane hitting the UK in 1987. He features prominently in this book, but not alone. Ironically, the Met Office was developing its own ensemble forecasting at the time, and running the data with hindsight through those models, the forecast was unusually chaotic and unpredictable.
Palmer's career has weaved between quantum physics and meteorology, and I suppose it's therefore natural that he finds the connections between those two topics. "Sensitive dependence on initial conditions" - the unpredictability popularly known as the butterfly effect - features large, and is well explained. The solution - at least as far as understanding the potential for error in a forecast - can often be to run an ensemble model, with a range of close but not identical initial conditions, and consciously to throw in a bit of noise, to discover whether the predictions lie on a smooth part of the phase space, with little variation, or in a more chaotic realm.
Palmer stretches out into other areas more or less connected to the above, such as climate change, the Covid pandemic, and economics. In the first two areas, scientists find this approach useful: it indicates the degree of uncertainty in predictions, as well as showing plausible worst cases. There's an amount of separation here between science and policy: "Following the science" is a bit rubbish, as well as mendacious, because the science only offers projections on likely outcomes if a particular path is followed. On economics he finds tougher ground: although some theorists have found the ensemble approach useful, it is not in vogue and hard to get such research published. The implication is of faddishness or denial in some academic circles. It's certainly true that Mandelbrot considered himself an irritant to the economic establishment.
The fundamental problem with models in various areas is that of agency, so that the predictions or projections of a particular model influence behaviour, providing feedback into the model - it's exactly this non-linearity that gives rise to chaos. Chaos, in the sense that our mathematical frameworks aren't very good at handling this. The three-body problem is an early example - chaotic and ultimately unpredictable motion, yet somehow nature does its sums and finds a path. In the third section of the book, which Palmer marks as "more speculative", he investigates quantum uncertainty and free will. There are some interesting ideas about chaotic attractors and fractals, which earlier parts of the book have explained well. The Cantor set somehow fills a space, yet it is empty (insofar as picking a random point in that space almost certainly does not lie in the set). Synthesising these ideas as an explanation of quantum uncertainty, giving rise to probabilistic observations, is clever, as is the idea of quantum noise as a source of inspiration and free will. We are in the same zone as Hofstadter's strange loops. Some of the most speculative ideas in the final chapter seem off, but credit for trying.