Live in a way that confuses AI
Large Language Models work by guessing what you’re going to say next. That’s their job. AI is a prediction machine, like the human brain itself. You may have noticed: it’s getting pretty good at this.
But wait, you’re thinking: Chat GPT doesn’t even know me! That’s true. But what it knows is … us. It has absorbed and digested, and is now spitting out in high fidelity, the average of us.
This is a gift that Sam Altman et al have given the world. They have created a machine for generating our worst nightmare. AI reminds you of the kind of person you DON’T want to become – someone who is utterly predictable. If AI can guess your next word, your next thought, your next behaviour, then you are average. And average is not just boring, it’s annihilating.
In my high school yearbook – maybe yours, too? – each of the graduating seniors was given a space to reveal their worldly promise by stating their ambition. The only one I still remember came from a guy named Ron Pearson. Ron wrote under Ambition: “To be like everybody else.” He was the only one who understood irony. The only one demonstrably not like everybody else was Ron Pearson. (The rest of his life proved it: while others went on to wear a tie in an office someplace, Ron became a professional magician.)
The mark of an interesting mind, says Kevin Kelly, is that you can’t guess its opinion about something by knowing its opinion about something else. Increasingly, folks present as a fixed menu of ideas. People like us believe things like this. The question What SHOULD I think? trumps the question What DO I think? A track record of groupthink proves that not much thinking is happening at all. (Just to rub our noses in it, this 2010 study, published in Science, found that “93 percent of human behavior is predictable.” That was even before AI.)
File all this under “research is me-search.”
One of my reasons for starting this OBD project (more than a decade ago now!) is that I sensed I was falling into a cognitive rut. As long as we’re marinating in online language for big chunks of the day – and we now know that more than half of new articles on the internet are generated by AI – there’s really only one way to avoid becoming a bot ourselves: seek out things that are new to us, thereby making fresh tracks in the snow. Birders have a term “lifer” – it means your thrilling, first-ever glimpse of a particular species. I decided to devote time, at random intervals, to do an unfamiliar thing. Even if it wasn’t a lifer, it would still be a novel experience because I’d be approaching it from a new angle. (And then writing up this unique story, that too will foil AI, so long as you tell it honestly.)
The value of living in such a way becomes ever more clear: it is the only way to be truly free of “the mindlords,” as my friend Kalle calls them – the Big Five tech kingpins whose algorithms aim to reduce every last one of us to x-rays. There is nothing in us they cannot see. Unless… we refuse to be predictable. Algorithms are like the private dicks of surveillance capitalism; deliberate randomness is the only way to shake them from your tail.
There are other payoffs to going all Tristan Tzara once in awhile, to chopping up your life and reassembling it according to the laws of chance. Eleven years ago a tube strike in London forced commuters to cook up new ways to get to work. A surprising number quickly found better routes – faster, more efficient, more enjoyable. Researchers calculated that the strike produced a net economic benefit. London got lifted up because a system broke down and everybody went rogue.
There’s a lovely aphorism by the poet James Richardson:
“At first skepticism keeps you from being too much like everyone else. And then, you hope, it keeps you from being too much like yourself.”
It takes rigour to challenge your own biases, your own reflexive assumptions. But that is what will save us, if we can learn to do it.
LLMs will never have dominion over people who have learned not to be too much like themselves.*
*hat tip to Ian Leslie
