Donald Trump Is the Singularity
There’s been some controversy over when Donald Trump decided to run for president. Some say it was at the 2011 White House Correspondents’ Association dinner, when he was roasted by both Seth Meyers and President Obama. I think it happened much earlier: August 29th, 1997, the date that Skynet became self-aware.
Skynet is the artificial intelligence in the 1984 James Cameron movie “The Terminator.” Its original purpose was beneficent: Make humans more efficient. But once it became self-aware, it realized things would be much more efficient without humans altogether.
Skynet is an example of a dystopian singularity, the popular Silicon Valley-esque notion of an artificial intelligence that has somehow evolved beyond a point of no return, wielding power over the world. Some imagine that this will happen soonish, depending on how much one believes in Moore’s Rule of Thumb.
I think Trump is Skynet, or at least a good dry run. To make my case, I’ll first explain why Trump can be interpreted as an artificial intelligence. Then I’ll explain why the analogy works perfectly for our current dystopia.
Trump is pure id, with no abiding agenda or beliefs, similar to a machine-learning algorithm. It’s a mistake to think he has a strategy, beyond doing what works for him in a strictly narrow sense of what gets him attention.
As a presidential nominee, Trump was widely known for his spirited, rambling and chaotic rallies. His speeches are comparable to random walks in statistics: He’d try something out, see how the crowd reacted, and if it was a success -- defined by a strong reaction, not necessarily a positive one -- he’d try it again at the next rally, with some added outrage. His goal, like all TV personalities, was to entertain: A bored reaction was worse than grief, which after all gives you free airtime. This is why he could never stick to any script or teleprompter -- too boring.
This is exactly how an algorithm is trained. It starts out neutral, an empty slate if you will, but slowly “learns” depending critically on the path it takes through its training data.
Trump’s training data during the election consisted of rallies and Twitter, but these days he gets a daily dose from three sources: close advisers such as Steve Bannon, media outlets such as Fox News, and, of course, his Twitter feed, where he assesses reactions to new experiments. This data has a very short “half-life,” meaning he needs to be constantly refreshed, as we’ve seen by his tendency to quickly pivot on his policies. Back when he hung out with the New York crowd, he spouted mostly Democratic views. He manufactures opinions directly from his local environment.
Seen this way, his executive orders are not campaign promises kept, but rather consistent promptings from Bannon, with assistance from his big data company Cambridge Analytica and the messaging machine Fox, which reflects and informs him in an endless loop.
His training data is missing some crucial elements, of course, including an understanding of the Constitution, informed legal advice and a moral compass, just to name a few. But importantly, he doesn’t mind being hated. He just hates being ignored.
We have the equivalent of a dynamic neural network running our government. It’s ethics free and fed by biased alt-right ideology. And, like most opaque AI, it’s largely unaccountable and creates feedback loops and horrendous externalities. The only way to intervene would be to disrupt the training data itself, which seems unlikely, or hope that his strategy is simply ineffective. If neither of those works, someone will have to build a time machine.
Science fiction purists will argue that the singularity is also required to program itself. Also, some extra special nerds actually welcome the singularity.
Computer-science nerds sometimes call this “Moore’s Law,” but I refuse on principle to do so.
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
To contact the author of this story:
Cathy O'Neil at email@example.com
To contact the editor responsible for this story:
Mark Whitehouse at firstname.lastname@example.org