The advent of supercomputers, artificial intelligence and and real stress tested algorithms gives us opportunities to understand the world in ways that our grandparents could never have dreamed of. Already they are proving their worth in things like weather forecasting and epidemiology, turning in great results in influenza and Ebola mapping. With Covid-19 the situation is problematic, partly because there’s so much emotional and economic baggage riding on every decision.
Today we are going to offer two pieces for your perusal. Both are critiques of the models used to understand the pandemic. We would like to stress that both the writers and the scientists who created the models are intelligent honest individuals of the highest integrity, driven by facts and a search for the truth. Unlike some of the sneering critics, who seem to have very different agendas. They’re good articles, but for us the real point is different.
Nature* points out the enormous potential for crossover, by using techniques developed in meteorology to help epidemiologists. Adam Kucharski in the Guardian* points out that bringing several different models together has enormous benefits, as the work of one group complements each of the others. In other words if you put effort into developing some good honest research, or a clever piece of technology, someone somewhere will also get the benefit. It’s facile to state how Kepler benefitted from the solid research of Tycho Brahe, or Watson and Crick from the advances in crystallography made by others. A society that values science for its own sake will eventually grow rich. A society that does none at all, because of tax cuts, is going in a different way.
Nature: Forecasters can learn from climate models
Epidemiologists predicting the spread of COVID-19 should adopt climate-modelling methods to make forecasts more reliable, say computer scientists. The researchers have spent months using a powerful supercomputer and techniques that are used to stress-test climate models to audit CovidSim, one of the most influential models of the pandemic, which helped convince British and US politicians to introduce lockdowns to prevent projected deaths. The analysis shows that, because researchers didn’t appreciate how sensitive CovidSim was to small changes in its inputs, their results overestimated the extent to which a lockdown was likely to reduce deaths. But the model correctly showed that “doing nothing at all would have disastrous consequences”, says chemist and computer scientist Peter Coveney.
Nature | 6 min read
Reference: Research Square preprint
The guardian:here’s why we need covid models,evenif they’re controversial
#covidsim #neilferguson epidemiology #covid-19 #computermodels