Michael White’s presentation at the USU Watts Spring seminar series is titled ‘The Spectacular but Disturbing Rise of Artificial Intelligence in Weather Prediction.’
“But really, I thought a lot about this and I want to go with an alternate title, which instead is going to be death or salvation in weather forecasting, an unfinished opera in four acts.”
Those ‘acts’ are actually a reference to the history of weather forecasting. Starting with the Navier–Stokes equations of the 1800’s and progressing to the first ‘entirely inaccurate’ prediction of 1922. Afterwards, incremental advances slowly progressed in the field until Lorenz’s famous ‘chaos paper’ of 1963.
“That just blew this entire field up. This is the paper that kind of launched the idea, chaos, in weather prediction, and many fields of science. And what they showed was an incredible sensitivity of the future to minute variations in the current state. So tiny differences in the state of the land surface winds pressures could propagate to enormous differences in the weather forecast. This was an absolute bombshell.”
From there, progress picks up again, though now in the light of uncertainty, and continues to the present day.
“And it's the part where I don't know what the answer is, I have some idea about how we got to where we are, but where we're going to end up is largely going to be up to you.”
As a senior editor at Nature, White has a front row seat to a slew of new and exciting developments in weather forecasting, like neural network-based approaches, known as AI, that can execute predictions more than 10,000 times faster than modern techniques. However, White is unsure of the future of such remarkable advances. Will they simply be used to filter ever larger pools of data, or can they help us probe deeper into the theory and practice of weather forecasting on our planet and others?
“Let's take them to Jupiter. Let's take them to Mars. Let's interact with the exoplanet community, who are really interested in being able to understand atmospheric circulations, habitability on exoplanets. So, these can be probed in a very wide-ranging sense”
But how we do that is up to us.
