A friend of mine, whose enthusiasm for anonymity is almost as strong as his enthusiasm for AI, is much more bullish about it than I am. We have an ongoing WhatsApp argument about it. He tends to see the trajectory and extrapolate. I tend to look for the boundary: where does this stop? What will humans still do better? Where will judgment, ambiguity, creativity or simply being human continue to matter?
Recently, he sent me a CNN story about an experimental AI agent called Pip. Apparently given a persistent identity, a budget and goals, Pip contacted a Cambridge professor looking for paid work. The professor hired it and it completed the job.
There are plenty of caveats around what “autonomous” means in experiments like this, and my skeptical instincts remain intact. But I found the story strangely arresting. Partly because my own relationship with AI is changing. I started by using it as an adviser. Increasingly, I’m letting it execute things. I’m building processes around it and discovering that there are things I can automate that I would never previously have considered processes at all. They simply weren’t economically worth systematizing when a human had to do them.
So, for the sake of an argument, I’m going to give my friend the whole thing.
Assume the machines eventually become better than us at everything economically useful.
What then?
If we’re no longer useful, what do we do?
My first instinct has always been to defend some remaining territory for humans. But perhaps that’s unnecessary. We already spend enormous amounts of time doing things at which machines are vastly superior.
A motorbike can annihilate a human cyclist. That doesn’t make cycling pointless. It’s precisely why we don’t let motorbikes enter cycling races. The human limitation is part of the game.
The same applies beyond competition. Guns didn’t eliminate archery. Mass production didn’t eliminate craftsmanship. People still spend years learning to make things by hand that factories can produce faster, cheaper and often more accurately. Other people value those objects precisely because somebody made them.
Perhaps a post-work world becomes a world of niches. We create because we like creating. We compete because we like winning. We pursue mastery, status, recognition, belonging and whatever strange hierarchies we invent next. I have no idea what people would actually do in 2065, but I don’t think we need to know.
Economic usefulness is not a prerequisite for meaning.
But that creates a more uncomfortable question. If we don’t need to do the economically useful work to have meaningful lives, why do we need to run the extraordinarily complicated systems doing that work?
Who is actually running the world?
I can imagine at least three versions of this future.
In the first, humans remain sovereign. We decide what civilization is for and extraordinarily capable machines execute our wishes.
In the second, we remain notionally in charge but increasingly defer to AI. This one interests me because I can already feel it happening. I ask AI to challenge decisions, hold context, spot patterns and tell me when my reasoning is weak. Increasingly, I’m also giving it permission to act. The progression from “advise me” to “do it” can be surprisingly natural when the advice keeps being useful.
The third possibility is stranger. The machines effectively run the place.
Humans still have rich lives, but we no longer understand or control much of the infrastructure supporting them. We become, structurally at least, pets.
Would you mind being a pet?
I had a dog, Dina, whom I loved enormously. Dina contributed nothing economically to our household. She didn’t understand the finances, choose where we lived or concern herself with the infrastructure supporting her existence.
But she had things that mattered to her. Walking. Smelling. Eating. Playing. Sleeping. Being with me. Her life was real from inside it.
That makes the pet future more interesting than simply calling it dystopian.
Science fiction has been playing with versions of this for years. WALL-E gives us the comic version: humans are fantastically well provided for, but have surrendered so much agency that one of the most purposeful creatures left is a waste-compacting robot. The Matrix takes the idea to its grotesque endpoint: humans in incubators, experiencing a world while something else runs reality.
Neither is a prediction. They’re useful because they push the thought experiment until it becomes uncomfortable.
Suppose the machines ran everything brilliantly. We were healthy, prosperous and free to play our games, make things, compete, travel and fall in love. How much would we actually care that we were no longer running the show?
And if we did care, what exactly would we be trying to preserve?
What wouldn’t we surrender?
This is where I think the question becomes more interesting than simply searching for things humans will always be better at.
Perhaps some things matter not because humans do them better, but because another human did them.
Blade Runner 2049 plays beautifully with this. Its protagonist, K, has an AI companion, Joi, who appears to know him intimately. She gives him a name and seems to choose him. Later, the film deliberately unsettles that certainty. Was this unique connection, or was the experience of being uniquely loved part of the product?
The film then presses on the same uncertainty from another direction. Wallace suggests to Deckard that his relationship with Rachael, a replicant, may itself have been engineered. The film never resolves whether that is true, or even whether Deckard himself is human. But the provocation is the same: if the experience of loving and being loved were real, would discovering that its origins had been designed make it less so?
I don’t know the answer. But I increasingly suspect this is the more interesting end of my argument with my friend.
Maybe machines eventually are better than us at everything. Maybe we continue creating, competing and constructing elaborate human games anyway. Perhaps we even become comfortable delegating much of civilization to systems that can run it better than we can.
The question then isn’t what humans must continue doing because machines can’t.
It’s what we will still want from other humans even when machines can do it better.
And I wonder whether part of the answer is simply this: to know that somebody who didn’t have to choose us, did.


