Marohasy’s Machine Learning Paper Gets Schooled

This is a guest post by ClimateDenierRoundup 

The big news in denierworld this week is obviously the latest (and peer-reviewed) #ExxonKnew study.

But before Exxon stole the spotlight on Tuesday, the newest paper of note in the Denivory Tower was one published earlier this month in an obscure and soon-to-be discontinued journal. The paper claims current warming is driven by natural forces, and is not uniquely human-caused.

Despite being based entirely on usually denier-derided computer modeling, the study was immediately–and lazily–championed by the climate denial fake news apparatus. Co-author Jennifer Marohasy wrote about the study in an op-ed for the Spectator and at her own blog; her description of her findings was uncritically copy-and-pasted into a Michael Bastasch story at the Daily Caller, reposted at WUWT and TallBloke, praised by James Delingpole at Breitbart and briefly linked to at Drudge (complete with an irrelevant photo of Al Gore).  

Yet a quick fact check on science Twitter shows the study’s conclusions are “based on inaccurately cited data that's incorrectly scaled & incorrectly aligned in time.” What’s more, even if the authors had used the data correctly, it’s still “essentially just a complicated curve-fitting exercise,” as And Then There’s Physics points out in a short and simple debunking of this complicated exercise. Marohasy et al’s paper tries to use machine learning to find natural patterns in warming unrelated to carbon pollution, and admits that it doesn’t take actual real world physics into account. As And Then There’s Physics concludes, the study uses a “naive approach – that completely ignores [the] physics” that govern the climate system.

Given that the study itself admits as much, it’s a bit bizarre that the authors couldn’t have seen these many intelligent criticisms about their artificial intelligence paper coming. Unless, of course, they did, and the allure of fawning fake news coverage and a convenient talking point about the denier-beloved Medieval Warming Period was just too great to resist.

There’s also the added bonus that the study was funded by a denier and that the authors work for an Australian industry-funded think tank. Doesn’t take any intelligence, artificial or otherwise, to deduce what’s going on here.

You can lead AI to denial, but you can’t make it think.