Overpromising AI

How to Inspire Skepticism

Saturday-Sunday, April 4-5, 2026

Item 1:

Happy Easter! Frohe Ostern! Joyeuses Pâques! Felices Pasquas!

---

Item 2:

One of those things that makes me want to gnaw on a wrought-iron table leg is the tireless boosterism on behalf of AI, by which the fans I have in mind mainly mean large-language models. I’m not a true believer here, on either side: dedicated phobe or enthralled booster. Both fly in the face of my native agnosticism.

Don’t get me wrong: I see AI’s potential in lots of realms. But I don’t see its potential for doing as much as its boosters claim—not nearly, and not yet.

The thing that caught my attention recently was the claim that the first “one-man billion-dollar enterprise” had finally arrived, much as purportedly predicted by Sam Altman, the OpenAI boss. This was passed along by Tyler Cowen, citing a report in the NYT.

Maybe this had more to do with Tyler poking his blog’s community, or maybe it was frustration that more people don’t feel as awed as he seems to be to drop everything to Oooh! and Aaah! at all the technical wonders and promise.

The whole thing just seems shriekingly goofy: to claim an enterprise run almost entirely by AI agents will produce (much less sustain) viable earnings at that level. Maybe if you’d discovered a quick scam with quick global reach. But a real valuation at $1bn? With durable returns?

If one sole-proprietor entrepreneur could prompt AI agents to do it, there are going to be tens of millions of people attempting the same thing within a short timespan. It’s the same sad story of economics: diminishing returns everywhere you look. If something promises easy money with little effort, you’re going to attract hordes of copycats attempting to grab as much of the easy cash as possible. Soon all the easy money is gone. Not far behind are those jealous that someone has stolen a base in their branch of industry. In the case presented, this would be pharmaceutical industry insiders, lobbyists, middlemen, and hangers-on. Interest groups will use the combination of political clout and lawsuits in an attempt to bring any domain intruders to heel.

On the other hand, singing from the same hymnal, but in a minor key, are the doomsters claiming AIs are all-but going to capsize planet Earth and plunge it into the sun.

What am I missing here? Maybe it’s all just artifacts of the Internet Age and the Attention Economy. The need to be seen guarantees that the frothiest claims bubble up to the surface of the social media cauldron in all their steamy pungency.

Never mind what I’m missing. What are they missing? That there are a lot of tasks that people routinely perform that don’t involve producing text and images for other people’s consumption—tasks that AIs seem capable of doing well, often enough, and inexhaustibly. But apart from banging out content, humans also have to shop for groceries, eat, sleep, and such. Humans have physical chores like home upkeep and repair. They have things they like to do for themselves and with others, sometimes with the aim of earning a living, sometimes just for fun. Some people even enjoy expressing themselves verbally, reading or writing, drawing or building: creating things. There’s a good time to be had learning new skills and applying them.

How much of this is AI actually meant to handle? And who came up with the idea that everyone is equally enthusiastic about having AI do it for them?

The AIs in their current state are interesting for what they seem capable of doing, but for my purposes, there’s nothing that they can take over that I want someone else to do. Besides, if this cartoonish picture is right, we should expect to have our personal AI agents handle our bothersome emails, texts, and phone calls all day. And on the other end, AIs will answer all those, too, that are generated by someone else’s AI agents. Meaning, one supposes, that all that correspondence was itself superfluous to begin with.

*grumble-grumble*