AI, IQ, and Other Four-Letter Words I Don't Understand After 10 PM

Feb. 6, 2025

So, Sam Altman, the big cheese over at OpenAI, thinks his silicon children are getting smarter. He’s throwing around “IQ” like it’s a goddamn measure of anything, let alone the ghost in the machine. Says they’re jumping a standard deviation every year. Spiritual answer, he calls it. Probably had a few spirits himself before spouting that gem.

Look, I’ve spent more time staring into the bottom of a glass than I have at lines of code, but even I can smell the bullshit wafting off this one. It’s thicker than the smoke in my apartment after a particularly rough deadline. These tech gurus love their buzzwords, their metrics, their ways of making the incomprehensible sound, well, still incomprehensible, but important.

They’re saying IQ tests are a bad benchmark for AI. Well, no shit, Sherlock. I thought that was obvious, and my brain runs on a steady drip of cheap bourbon and nicotine. It’s like comparing a goddamn racehorse to a rocket ship. Sure, they both move, but one’s going to win the Kentucky Derby, and the other’s going to get you to Mars – assuming it doesn’t explode on the launchpad, which, given my luck with technology, is a 50/50 proposition.

This Oxford researcher, Wachter, she’s got the right idea. “Apples and oranges,” she says. Exactly. One’s a fruit, the other’s a citrus abomination that gives me heartburn. And comparing AI to human intelligence? It’s like saying my beat-up typewriter is smarter than me because it can type the same sentence 500 times without complaining.

The real kicker? These IQ tests, the ones they’re using to “measure” AI, are about as reliable as a weather forecast in April. They’re biased, culturally loaded, and probably designed by some egghead who thought “selective breeding” was a good idea. Eugenics, they called it. Sounds like something you’d order at a fancy restaurant, then regret immediately.

And these AI? They’ve got an unfair advantage. Infinite memory, access to the entire damn internet – which, let’s be honest, is mostly cat videos and arguments about whether pineapple belongs on pizza (it doesn’t, by the way). It’s like giving a kid a cheat sheet the size of the Library of Congress and then calling him a genius when he aces the test.

This Keyes fella, from the University of Washington, he gets it too. Says these tests are “easy to game.” Like shooting fish in a barrel, if the fish were made of data and the barrel was the size of the Pacific Ocean.

And then there’s Cook, from King’s College. He’s talking about crows using tools. Now, I like crows. Smart birds. But I’m not about to send one to Harvard. My brain, on the other hand – when it’s not busy processing the existential dread of another sunrise or wondering where I left my lighter – is juggling a million things at once. Bills, deadlines, the nagging feeling that I should probably eat something other than ramen, the lingering scent of that redhead from the bar last night…

AI doesn’t have to worry about any of that. It doesn’t have to worry about hangovers, or heartbreak, or the crushing weight of human existence. It just…computes.

So, what’s the point of all this? This whole IQ charade? It’s a distraction. A shiny object to keep us from asking the real questions. Like, what are these things for? Are they going to make our lives better, or are they just going to replace us all, leaving us to wander the wasteland, scavenging for scraps of meaning in a world run by algorithms?

They are searching for better AI tests. Khlaaf, from the AI Now Institute, says that. And the nature of computation means that they have abilities that surpass us. This is not news, pals. When I lost my job as a technical writer, I knew that the end was near. The machines are not coming. They are here.

Look, I’m not saying AI is all bad. It’s probably good for something. Maybe it can finally figure out how to fold a fitted sheet. Or write a decent country song. Or mix a perfect Old Fashioned. But let’s not pretend it’s “intelligent” in the same way we are. It’s a tool, a fancy one, but still just a tool.

And as for me? I’m going to keep doing what I do best: observing the world, one blurry, whiskey-soaked moment at a time, and trying to make sense of the beautiful, terrifying mess of it all.

Another round, barkeep. Make it a double. And keep ’em coming.


Source: Why IQ is a poor test for AI | TechCrunch

Tags: ai algorithms machinelearning humanainteraction technologicalunemployment