AI and meteorology have a vexing problem

Current data is key to weather models, but that takes effort and money.

The Minnesota Star Tribune
May 10, 2024 at 12:23AM

No sign of my flying car but yesterday my Amazon “Amish butterscotch peanut butter” order arrived via FedEx. I’m so confused. But well-fed.

All this worry about artificial intelligence because natural intelligence doesn’t seem to be working. A tool - or will it replace me? Yes. AI may help us issue a 1-2 week tornado outlook or give a day or two notice for severe flooding in years to come, but there’s a vexing problem.

To predict the weather we need to “initialize” weather models with current data. Weather balloons, ocean buoys, airport data, planes, etc. And that’s expensive. AI can’t substitute for the fuel we put into the weather model engine. But, yes, I may soon be getting my weather from a friendly, 24-7 weather avatar that never takes a vacation or asks for a raise. Uh oh.

A stray shower is possible later today, a few T- storms may sprout late Sunday. Saturday looks like the sunnier, drier day. No blockbuster storms, as the pattern dries slightly.

Next week 70s should be the rule. Wait, let me ask Alexa.

about the writer

about the writer

Paul Douglas

Columnist

Paul Douglas is a nationally-respected meteorologist, with 40 years of broadcast television and radio experience. He provides daily print and online weather services for the Star Tribune.

See Moreicon

More from No Section

See More
FILE -- A rent deposit slot at an apartment complex in Tucker, Ga., on July 21, 2020. As an eviction crisis has seemed increasingly likely this summer, everyone in the housing market has made the same plea to Washington: Send money — lots of it — that would keep renters in their homes and landlords afloat. (Melissa Golden/The New York Times) ORG XMIT: XNYT58
Melissa Golden/The New York Times

It’s too soon to tell how much the immigration crackdown is to blame.