Trump Is a Machine-Learning Algorithm Gone Wrong

Trump Is a Machine-Learning Algorithm Gone Wrong

(Bloomberg Opinion) -- Remember the Magic School Bus, the animated series in which a somewhat demented school teacher named Ms. Frizzle (catchphrase: “Take chances, make mistakes, get messy!”) takes her class on mind-bending adventures? In one episode, she reduces the bus to microscopic size and takes her charges into the bloodstream of an ill classmate. You see the kids swimming around in tiny wetsuits that they happened to bring along, watching white blood cells battle infectious invaders.

Well, that’s what I was thinking about yesterday. The venue wasn’t a school, it was the White House press briefing room. Instead of an eccentric science nerd, it was President Donald Trump talking about injecting disinfectant or UV rays into people’s bodies to clear away the coronavirus. And instead of bringing us home safely on the bus, he might do anything next.

Allow me to offer a hypothesis on what is going on. It won’t help anyone get out of this living nightmare, but it might at least provide some psychological relief.

A while ago, I wrote about how Trump is like a machine learning algorithm, designed to attract maximum attention. He responds to positive and sometimes negative stimuli, focusing heavily on what’s recent. His data come primarily from campaign rallies, polling and Fox News, with other sources sneaking in sometimes, but not often because they don’t flatter him sufficiently. He’ll try something kind of at random, and if the response is good, he’ll go there again, unconstrained by consistency or morality.

In the age of the novel coronavirus, he’s running low on data. The rallies are gone. Fox News isn’t as useful as it once was, constantly wavering on whether the whole coronavirus thing is a hoax and whether the administration’s approach is good. His daily press briefings are filled with journalists who don’t tend to respond enthusiastically. The sparse and inconsistent signals confuse Trump.

So he’s at a loss, and his moves are becoming less frequent and increasingly random. The clearest signal he got in days was the anti-lockdown mini-protests around the country, which sent him straight to Twitter with his series of “Liberate” messages. But in a matter of hours, new polling showed that most of the country - even among his base, and especially the older, voting part - actually opposes the idea of running outside and getting sick. The result was the colossal train wreck that is the governor of Georgia.

Expect more. At this point, the president of the world’s most powerful nation is effectively a short-horizon machine-learning algorithm with very little training data to go on. To use an analogy from the financial world that I used to inhabit, it’s like what happens when a trading algorithm meets a highly unusual event, an “edge case.” It does something weird, which causes other algorithms to do something weird, and pretty soon you have a flash crash.

Trump has met his edge case. Somebody needs to find Ms. Frizzle and give her back the keys to the bus.

This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.

Cathy O’Neil is a Bloomberg Opinion columnist. She is a mathematician who has worked as a professor, hedge-fund analyst and data scientist. She founded ORCAA, an algorithmic auditing company, and is the author of “Weapons of Math Destruction.”

©2020 Bloomberg L.P.

Get live Stock market updates, Business news, Today’s latest news, Trending stories, and Videos on NDTV Profit.
GET REGULAR UPDATES