This is the second in a series of posts I hope will help readers navigate the perils and promise of AI. In part one, I put on my happy hat and tried to imagine the best plausible future shaped by artificial intelligence.

Now let’s veer to the other side of the Cone of Plausibility and explore the darkest plausible future shaped by AI. The AI-pocalypse scenario paints a picture of the worst-case outcomes of the current tech boom. Think of it as a cautionary tale that can spur us to move things in a different direction, using our resources and influence as a field.

Elizabeth Merritt, VP Strategic Foresight and Founding Director, Center for the Future of Museums, American Alliance of Museums

---

Dear Mom,

Thanks for the care package. I especially appreciate the cookies, because it is too dang hot in the apartment to run the oven and home-baked chocolate chip is THE BEST.

Eric and I found two more roommates to split the rent. Jet is a bike messenger (a lot of local businesses are sending important documents by pouch rather than risking yet another data hack). Lia is a home health aide, which is a reliable gig, even if the pay is cruddy. Eric got RIFed from his assignment piloting military drones (he’s been going on long rants about how we are going to regret turning over combat decisions to AI), but he got a job with a private security firm catering to bigwigs who want to flaunt some human muscle.

I feel like the weak link in the chain—it’s nearly impossible to find design work. All my old clients are using image generators for their cover art and graphics. Actually, it’s nearly impossible to find ANY work—I’ve sent out over 500 applications this month for all sorts of jobs, and haven’t gotten even one nibble. At least I don’t feel like a unique failure—all of my buddies from art school are in the same pot. I joined a support group that meets at the library every Wednesday to swap tips on gaming the AI systems that everyone posting jobs is using to screen applications.  

The biggest tension in the apartment is about water—each of us can have a three-minute shower a day without exceeding our water ration. But whoever gets up first usually lets the water run longer, and then we don’t have enough left for cooking or laundry. (Or the rest of us skip showers, which is just nasty.) The upside is that we bond over our shared hatred of the data center that’s sucking the local reservoir dry. And making the neighborhood hotter. And crashing the power grid when the temperature goes above 115, at which point they start running their diesel generators, stinking up the neighborhood and making it too noisy to sleep. What idiots approved the permit for that place, anyway? Oh yes, the local officials who bought the company line that it would bring jobs and prosperity to the city.

My biggest source of stress relief is chilling at the local art museum. Literally chilling—they have top notch AC—and spending time with my favorite Kahlos and Kandinskys always makes me feel better. But I miss hanging out with the gallery guards and front desk staff, who were mostly fellow artists. Almost all of them have been replaced by AI systems, which I suppose saves the museum money, but I really miss the human connection. I’ve resolutely refused to download the museum’s AI interpretive chatbot on my phone—my own little act of Luddite rebellion. (As is my sending this missive by physical post. Check out the cool stamp, which I am pretty sure was designed by a real person.)

Mom, I know how worried you are about me, and I appreciate your offer to have me come home and live with you and Dad. (I consider it proof of divine providence that you paid off your mortgage just a week before the market crashed.) Let me give it a few more months and come December, if I haven’t found a job even vaguely related to my MFA, I’ll move back to Chicago and take Aunt Lil up on her offer to apprentice at her plumbing business. I can keep making art on my own time, and maybe I can sell it on Etsy, now that they’ve instituted their “made by humans” guarantee.

With love,

Audie

Is this Scenario Plausible?

Unfortunately, very. Here are some of the data, trends, and projections around which the story is built:

Implosion of the human workforce, with mass layoffs and unemployment.

Supporting projections include:

• 30 percent of US jobs automated by 2030 (Timothy Prestianni, National University Blog)
• 100 million US jobs lost by 2035 (McKinsey Global Institute)
• 35 percent unemployment in the next 2-5 years (Dan Shulman, CEO Verizon) For scale, unemployment in 1933, at the nadir of the Great Depression, was 25 percent.

Many communities rendered effectively uninhabitable due to:

• Water scarcity. Data centers across the US are projected to consume up to 73 billion gallons by 2028 (equivalent to about 730 thousand households) with the majority in areas already suffering from drought.
• Climate change. Cornell researchers project that by 2030 AI data centers could pump up to 44 million metric tons of CO2 into the atmosphere, the equivalent of adding another 10 million cars to US roads.
• Health impact of heat, noise, and air pollution.

Collapse of internet security due to:

AI-powered malware and autonomous hackbots, along with a catastrophic drop in public trust in digital content driven by the ubiquity of fake posts, images, research reports, and books created by AI. Taken together, these two trends may disrupt the central role played by the internet as a repository of information and platform for essential digital services, ranging from health to finance.

Warfare amplified by autonomous weapons

For example, weaponized drones that select targets without the input of a human operator) and AI-driven threat analysis and strategizing that may escalate conflict.

How does it all end?

This story paints a picture of society in the midst of an AI-induced crisis. What’s the final act of this play? The AI-pocalypse may end with the implosion of AI itself due to:

• Model collapse, a documented property of Large Language Models in which LLMs trained on AI-generated (aka synthetic) data eventually fail due to amplification of errors and imperfections. Comprehensive model collapse of systems trained on data harvested from the web is particularly plausible now that 40 percent of music uploads, more than a third of podcasts, and more than half of new English language articles published online are primarily written by AI. Basically (#irony) by displacing human-created content, AI may be sowing the seeds of its own destruction.

And/or

• Financial collapse of the speculative AI economy. In 2025, 60 percent of the growth of the US economy was driven by AI. By that time, the cumulative investment in AI was over $888 billion, while the total software revenues were $175 billion. A recent study by MIT found that efforts to improve profitability by integrating AI into operations have a 95 percent failure rate. Even some of the leaders in AI tech are warning this AI financial bubble could collapse, and the fallout could be worse than the dot.com crash of 2000 that wiped out trillions of dollars in retirement savings, led to widespread tech layoffs, and sparked a national recession.

Where Does this Scenario Lie in the Cone of Plausibility?

Bad as this may sound, it isn’t the darkest scenario people have imagined. In this version of the story, AI doesn’t become sentient, take over all our infrastructure and try to kill humanity. (Or at least, those of us who haven’t uploaded our consciousness into the cloud to effectively become AI ourselves.) While some people buy into these narratives, I think they lie outside the bounds of the Cone of Plausibility, or so far in the future that it could distract us from the actual challenges we face right now.

Instead, this scenario is based on the far end of credible projections from mainstream sources—but experts disagree. What is your gut feeling about the probability of this future? Does it feel unlikely, possible, or inevitable? If you believe it could happen, does that make you feel depressed, discouraged, or intrigued? By the end of this series, I hope you will feel inspired to take action, and have some ideas about what actions to take.

What’s Next

I’m working on the third scenario in this set, “Muddling Along,” forecasting what may be the most probable future. No heroic interventions, no spectacular mishaps, just a spot in the center of the Cone of Plausibility projecting where we will end up if we simply keep going as we are now. Stay tuned.