Can’t resist this seasonal one from the brilliant folks at Joy of Tech:

There are apparently some very capable folks working at Google’s DeepMind.  First the brilliant AlphaFold and now GenCast  — their New AI model that they claim (see their presentation in Nature here) “advances the prediction of weather uncertainties and risks, delivering faster, more accurate forecasts up to 15 days ahead.”  The Nature paper claims that GenCast has “greater skill and speed than the top operational medium-range weather forecast in the world, ENS, the ensemble forecast of the European Centre for Medium-Range Weather Forecasts…GenCast is trained on 40 years of best-estimate analysis from 1979 to 2018, taken from the publicly available ERA5 (fifth generation ECMWF reanalysis) reanalysis dataset. Reanalysis provides a reconstruction of past weather by computing analysis for historical dates and times.”  Though the comparison is with the 2019 ENS modeling (and the latter has improved since then), this seems pretty impressive.

David Wallace-Wells, writing in the NY Times, does an excellent job of exposing the false notions inherent in some of the new Covid mythologies being created.  Excerpts:

“…the federal pandemic response was actually supervised by Trump and many of those whom he appointed, the first time around. Americans often tell the story of Covid now as though our pandemic response was run by safetyist liberals in an unreasonable panic. But while Trump was remarkably indifferent to Covid in 2020, he was also, for the entire period we now remember erroneously as “lockdown,” in charge. (Americans often remember that period as stretching for multiple years; in fact, all but one state withdrew its stay-at-home orders within three months.)”

On Bhattacharya: “according to his analysis, the fatality rate for Covid was several orders of magnitude lower than was being reported at the time, and that even if it were to infect every single American, the result might be only about 33,000 deaths. Since the first outbreak, more than 1.2 million Americans have died, officially, from Covid; excess mortality figures are even higher. Recently, Bhattacharya has tried to emphasize that his opinion piece merely floated one estimate for fatality rate. But for every death his estimate implied, there were, in the end, more than 35.”

“But on the most basic and essential questions about the pandemic, the public health establishment was also, actually, right: Covid was extremely bad, ultimately killing upward of one million Americans and producing a death toll much more in line with worst-case predictions than “just the flu” reassurances; the vaccines were really good, ultimately saving millions of lives globally and, in countries where uptake was more universal, single-handedly drawing the worst of the emergency to a close; and the best way to minimize the ultimate death toll for Americans was to limit the number of infections that happened before those vaccines were distributed, whether that was achieved by trying to “flatten the curve” or Operation Warp Speed or both.”

Zeynep Tufekci, also in the NYT, has more to add re. Bhattacharya:

“In the early days of the pandemic, Bhattacharya repeatedly predicted that the virus would likely kill about 20,000 to 40,000 Americans. (The death toll turned out to be about 1.2 million.)...In early 2021, with no evidence, Bhattacharya declared that a “majority of Indians have natural immunity” to Covid-19, claimed “vaccinating the whole population can cause great harm” and predicted his preferred approach would “reduce death rates from Covid infection to nearly zero.” Shortly afterward, India suffered a deadly wave that killed millions of people in just a few months— among the highest, fastest death rates of any country.”