Latest / Minds, Bodies, and Terawatts / Goldman Sachs Just Put a Number on the AI Job Squeeze, and It Lands on People Who Haven't Started Working Yet

Transcript

Goldman Sachs Just Put a Number on the AI Job Squeeze, and It Lands on People Who Haven't Started Working Yet · Minds, Bodies, and Terawatts

• 0:00
  [Host] This is Minds, Bodies, and Terawatts, the show where the job losses turn up in the jobs nobody has had yet. CNBC reported Wednesday that Goldman Sachs found United States call center employment running thirty-nine percent below its pre-AI hiring trend.
• 0:14
  [Guest] And the same hole shows up in Canada and Germany. One recession does not straddle three separate central banks.
• 0:21
  [Host] Below trend still is not the same as caused by a machine.
• 0:26
  [Guest] That is what the cross-country test is for. Though in the aggregate the drag rounds to almost nothing. Goldman's own conclusion is that the damage stays inside a narrow set of industries and workers.
• 0:38
  [Host] So where is it?
• 0:40
  [Guest] Split it by seniority. For entry-level workers in America the same exposure does roughly twice the damage to annual hiring growth. The harm is aimed, not spread.
• 0:49
  [Host] Aimed at the first job. What does that look like?
• 0:53
  [Guest] NPR's Morning Edition found Irene Chang, an industrial engineering graduate from Georgia Tech. Four hundred fifty applications since last September, no offers. The entry-level skills you have are important, she said, but also something AI can do very efficiently.
• 1:07
  [Host] That is a graduate's theory of her own rejection. Does the research back her?
• 1:12
  [Guest] Stanford's Erik Brynjolfsson put a number on it in a paper called Canaries in the Coal Mine, Six Facts about the Recent Employment Effects of Artificial Intelligence. Workers twenty-two to twenty-five in the most exposed occupations show a thirteen percent relative decline in employment. Older workers in those same jobs held steady.
• 1:31
  [Host] Why would age matter if the work is identical?
• 1:35
  [Guest] His mechanism is codified knowledge. These models train on what somebody wrote down, which is almost exactly what a beginner has. The tacit part, what you only get from years of doing it, was never in the training data.
• 1:49
  [Host] So seniority is not skill. It is the part that never got written down. Do economists agree on the cause?
• 1:56
  [Guest] No. Harvard's David Deming, in Technological Disruption in the United States Labor Market with Lawrence Summers, dates the fall in junior hiring to about six months before ChatGPT shipped. His candidate is remote work, which broke office jobs into self-contained chunks that could be handed to anyone, anywhere.
• 2:13
  [Host] And once the work is in a container, it stops mattering whether a person or a model opens it.
• 2:20
  [Guest] One company is betting the other way. Bloomberg reported IBM will triple its entry-level hiring in the United States this year, and its human resources chief, Nickle LaMoreaux, put it plainly at a summit on leading with AI. Yes, it is for all these jobs that we are being told AI can do.
• 2:40
  [Host] So maybe the door is not closing.
• 2:42
  [Guest] Or IBM is buying cheap what everyone else is dumping. The grunt work was never only output, it was the apprenticeship. Strip it out and a market hires seniors, skips juniors, looks healthy on the charts for a decade, then finds nobody qualified to promote.
• 2:59
  [Host] What do you tell a new graduate tonight?
• 3:02
  [Guest] Aim at what does not compress. Judgment when the rules run out, negotiation, deciding what is worth building rather than building it. Goldman puts developed-world adoption at fifteen to twenty percent, so most of this has not happened yet.
• 3:16
  [Host] Patrick Deglon calls this the Labor Cliff in Unscarcity, displacement outrunning the economy's ability to invent new roles. What Goldman measured is not the fall. It is the first rocks going over the edge, and they are landing on the youngest people in the building.
• 3:32
  [Guest] Irene Chang trained for precisely the part of the job that was easiest to write down, and nobody mentioned that while she was paying for the degree. For more on the human edge and a career ladder losing its bottom rungs, head to unscarcity dot ai. The book is on Amazon, the audiobook is on Spotify. This has been Minds, Bodies, and Terawatts.