These conversations have unsurprisingly sent many workers into a panic (and possibly contributing to support efforts altogether). stop construction Some of the data centers spiked last week). This panic is not being helped by lawmakers, none of these A coherent plan has been prepared for what will happen next.
Even economists who have caution AI has not cut jobs yet and will not cause any crisis in future coming around With the idea that it could have a unique and unprecedented impact on the way we work.
Alex Imas, who works at the University of Chicago, is one of those economists. When we spoke Friday morning he shared two things with me: a frank assessment that our tools are very poor at predicting what this will look like, and a “call to arms” for economists to start collecting the kind of data that could make planning possible to address AI in the workforce.
On our frustrating tools: Consider the fact that any job is made up of individual actions. For example, part of a real estate agent’s job is to ask clients what type of property they want to buy. The US government transcribed thousands of these works into one huge list First launched in 1998 and updated regularly since then. This was the data that OpenAI researchers used in December to assess “how”exposed“One job is AI (for example, they found a real estate agent was 28% exposed). Then in February, Anthropic used this data in analyzing millions of cloud conversations which tasks People are actually using its AI to accomplish this and that’s where the two lists overlap.
But knowing the AI exposure of tasks can lead to a misleading sense of how much of a risk a given job is, says Imas. “Exposure alone is a completely meaningless tool for predicting displacement,” he told me.
Of course, this is the most frustrating case example – for a job in which literally every task Can be done by AI without any human instructions. If it costs less for the AI model to do all those tasks than you get paid – which is not a given, as reasoning models and agentic AI can scale enough bill-And it could make them, well, likely out of a job, Imas says. This is an oft-cited case of elevator operators decades ago; Perhaps today’s parallel is a customer service agent doing thorough phone call triage.
But for most jobs the matter is not that simple. And specifics matter, too: Some jobs are likely to have bad days, but knowing How And When? This would be hard to answer when just looking at exposure.
Take writing code, for example. Let’s say, someone who builds premium dating apps can use AI coding tools to create in a day what previously took three days to create. This means the worker is more productive. The worker’s employer, by spending the same amount of money, can now achieve greater output. Then will the employer want more employees or less?
