Why the thing that kills us won’t be a Terminator, but a chatbot that read a weird Reddit thread from 2014
The End Will Be Hallucinated
I haven’t really been reading the news much this year. It’s depressing, and it’s already giving me ulcers in places I didn’t know you could get an ulcer. But today, I flicked through the paper to see what’s up.
I regretted it immediately.
So. Turns out, the Bulletin of the Atomic Scientists recently released their 2026 statement, and if you were hoping for a “Good job, everyone, take five,” you are going to be very fucking disappointed. They have moved the hands of the Doomsday Clock to 85 seconds to midnight..
The Bulletin attributed the new time to “dangerous trends in nuclear risk, climate change, disruptive technologies like AI, and biosecurity”.
❝ Catastrophic risks are on the rise, cooperation is on the decline, and we are running out of time. Change is both necessary and possible, but the global community must demand swift action from their leaders.”
For the uninitiated, the Doomsday Clock is a symbolic timepiece maintained by a group of very smart, very stressed people in Chicago who gather once a year to decide exactly how fucked we are as a species. “Midnight” represents the moment we collectively annihilate ourselves via nuclear war, climate collapse, or a biological error.
We have now lost five seconds from previous years. I don’t know what you can do with five seconds. You can’t even skip a YouTube ad. But apparently, that is the margin of error we’ve managed to shave off our survival.
I know I write about AI like a bit of a Doomer, but I really am curious as to how this is all unfolding. As a cartoonist, I’ve been complaining about AI for years (mostly because it draws hands that look like a bundle of sausages exploding in a microwave.) But I didn’t realise I was actually critiquing the harbinger of our extinction.
I have a specific, professional history with this particular flavour of existential dread. A few years ago, I was commissioned to illustrate “The Essential Sam Harris” series. One of them was: “Making Sense of Artificial Intelligence”. It was a companion guide to his podcast. You should go read it if you’re interested in the topic. It’s well written for anyone who isn’t a scientist in Chicago.
Working on it forced me to confront the realities of AI long before ChatGPT was helping high schoolers cheat on their English Lit essays. Back then, Harris’s warnings felt theoretical. Today, his TED Talk (“Can we build AI without losing control over it?”) is being watched the same way Bill Gates’s TED Talk on pandemics was watched during the COVID-19 lockdowns. We are squarely in the “Oh, that is what you meant!??” phase of history.
While drawing that guide, I learned that the problem isn’t necessarily that robots will hate us. It’s that they might be too good at what we ask them to do.
Can we talk about the 85 seconds thing?
85 seconds is a very specific and irritating number. It seeems more like the microwave timer when you’re trying to melt butter but not explode it. It also creates a terrible sense of cognitive dissonance. The scientists are squealing, “The house is on fire! The roof is collapsing! The AI is learning how to enrich uranium and build breakdancing robots!” And my response is to go spend fifteen minutes at the bodega deciding which flavour of LaCroix represents my vibe this morning. I can’t identify with the threat because it’s just too big to fathom. Or maybe I’m avoidant. Maybe both?
This is the central tension of modern life: We’ve been expected to hold two opposing thoughts in our heads simultaneously.
Civilisation is dangling by a thread over a pit of radioactive crocodiles.
I really need to get my teeth cleaned before my insurance resets.
What is fascinating about this year’s statement (and this pertains to my own little corner of the world) is the heavy emphasis on Artificial Intelligence. Machine Learning. Robot Overlords. The Bulletin cites AI as a “threat multiplier.” Reading this report, there’s this sort of dark hilarity in the idea that the thing that might end us isn’t a muscular Terminator with an Austrian accent, but a Large Language Model that hallucinates a reason to launch a missile because it read a misinterpreted Reddit thread from 2014.
We are building systems we don’t understand to solve problems we created, and the scientists are basically saying, “Hey, maybe don’t give the car keys to a toddler.”
Right now, we are finally talking about “The Continuum of Competency” and realising that we’re no longer the smartest things in the room. We are waking up to the fact that we need to define our wishes very, very carefully before we rub the lamp.
In a way, I actually find the clock kind of optimistic. We’ve been on the brink of destruction for nearly eighty years. We are the species that invented the H-bomb, but we also invented jazz, penicillin, and avocado toast. We are a mess. We are a chaotic, violent, beautiful, stupid mess. And yet, we haven’t hit midnight yet.
85 seconds is terrifyingly close. It is a breath away. But it isn’t zero.
So anyway. Don’t spend those remaining seconds doom-scrolling. Spend them making something. Draw a picture. Write a joke. Pet a dog. If we’re going to go out, we may as well go out doing something fun.
‘til next time
your pal,
PS. Morris doesn’t know anything about the clock, and lives in complete ignorance of all of this shit. I’m jealous. He doesn’t have one ulcer.
I love this: "85 seconds is terrifyingly close. It is a breath away. But it isn’t zero.
So anyway. Don’t spend those remaining seconds doom-scrolling. Spend them making something. Draw a picture. Write a joke. Pet a dog. If we’re going to go out, we may as well go out doing something fun." And I, too, am jealous of Morris. Our biggest mistake is thinking we're the smartest guys in the room.
To be discussed!