Using IPCC climate models to inform planning and policy is tough. Despite billions of dollars spent over multiple decades, these models can’t make accurate enough predictions to be used for anything other than scaring people into forking over more money to support climate science.

A breakthrough paper titled A Non-Intrusive Machine Learning Framework for Debiasing Long-Time Coarse Resolution Climate Simulations and Quantifying Rare Events Statistics published by MIT scientists makes no attempt to fix core climate models. That would be both intrusive and blasphemous.

Plus, the authors admit, it wouldn’t work.

“People have tried to dissect into climate model codes that have been developed over the last 20 to 30 years, which is a nightmare, because you can lose a lot of stability in your simulation,” MIT Professor Sapsis explains. “What we’re doing is a completely different approach, in that we’re not trying to correct the equations but instead correct the model’s output.”

Neural network AI trained on recorded regional weather data is used to “nudge” the output of the broken models toward giving better answers. Sure enough, this kludge does a better job predicting things that have already happened.

The new techniques haven’t been used to make predictions about things that haven’t happened yet, like generating warnings about when and where the next big hurricane is going to strike. That will require more billions of dollars over additional decades, assuring the sustainability of climate science. But you can be certain that whenever and wherever a big hurricane hits this will be proof that climate change is an existential threat to the planet, and only Settled Science™ can save us.