Take an example. A translator, let's call her Claire.
A few years ago, translating a twenty-page document took her two days, and she billed several hundred euros for it. A normal price for difficult work that required years of study.
Today, anyone can get a decent translation of those twenty pages in thirty seconds, for nothing.
For you, that's excellent news: you can read any document, in any language. For Claire, it's a catastrophe.
That little story sums up a debate that is stirring up a lot of researchers at the moment. And it raises a question that concerns all of us.
The rule, in one sentence
When a machine can do a task, the price of that task falls towards the price of the machine.
That's it. It's a rule as old as economics. And it has two faces.
When you buy that task, you win: what used to cost hundreds of euros now costs almost nothing.
When you sell that task, because it's your job, you lose: what used to earn you hundreds of euros now earns you almost nothing.
And most of us are both at once. We buy things, and we sell our work so we can afford to buy them.
The real question
If everything gets cheaper, including your work, are you richer or poorer?
It all depends on one thing: what falls fastest. The price of what you buy, or the price of what you know how to do.
If your expenses halve and your salary drops by 10%, you come out ahead. If it's the other way round, you lose, even while living in a world where everything is cheaper.
And that's where the argument becomes worrying. AI hits office work first: writing, translating, analysing, summarising. So it drives down the price of what many people sell well before it drives down the price of what they actually buy. Rent, a house, a nursery place or healthcare don't get cheaper because a model can write.
In the early nineteenth century, weaving machines drove down the price of clothing. In the long run, everyone gained: today a t-shirt costs almost nothing. But the weavers of the time saw their incomes collapse, and many families endured poverty for a whole generation. Some ended up smashing machines. History proved them wrong about the future, and right about their own lives. Both are true at the same time.
What won't become cheap
Not everything automates at the same pace. We saw this yesterday with the predictions made in 2017: what is done with words moves fast, what is done with hands moves slowly.
And then there's everything that rests on trust, presence, responsibility. A doctor delivering bad news. A craftsman who guarantees his work. Someone who signs, and who answers for what they sign. A machine can produce the same content. Not the same guarantee.
This is not easy consolation: these jobs will not be able to absorb everyone whose work is losing its value. But it shows where value is shifting.
Where does the money saved go?
This is perhaps the most important question, and it is rarely asked.
When Claire's translation goes from several hundred euros to almost nothing, that money does not disappear. It goes somewhere. A little into your pocket, if you were the client. A lot into the pocket of the company that no longer hires a translator. And some of it to the company providing the AI.
This distribution is not decided by technology. It is decided by choices: prices, wages, taxes, rules. The job cuts of recent months already show which way the scales tip when no one chooses.
What to take away
"Everything will cost less" is only good news for those who buy more than they sell.
Most of us sell our labour. The real question is therefore not whether prices will fall: they will fall. It is who will benefit from the fall.
And that, unlike the speed of machines, is not yet decided.