Latest / Minds, Bodies, and Terawatts / OpenAI research finds nuanced outlook on AI's jobs impact

Transcript

OpenAI research finds nuanced outlook on AI's jobs impact · Minds, Bodies, and Terawatts

• 0:00
  [Host] This is Minds, Bodies, and Terawatts, where the company building the storm also writes the weather report. Axios reported on April 16th that OpenAI researchers published a study called "How Will AI Transform Work?" analyzing over 900 occupations and their exposure to AI automation. The headline finding: 18% face high near-term risk, 24% could see significant employment shifts. OpenAI's framing? Don't panic.
• 0:27
  [Guest] [sarcastic] Right, because when I want an unbiased assessment of whether AI will take my job, I go straight to the company selling the AI. That's like asking the wolf for a headcount on the sheep.
• 0:43
  [Host] OK but let's take the research seriously. Nine hundred occupations is comprehensive. What did they actually find?
• 0:50
  [Guest] The 18% high-risk bucket covers roles heavy on routine cognitive tasks: data entry, basic financial analysis, certain administrative functions. The 24% "employment shift" category is where OpenAI gets creative.
• 1:03
  [Host] Creative how?
• 1:04
  [Guest] They argue AI could expand demand in those roles by creating new task categories within existing jobs. Displacement and creation happening simultaneously, in the same position. It's a tidy theory.
• 1:17
  [Host] That sounds reasonable on paper. So why the skepticism?
• 1:21
  [Guest] Because the real-world layoff data tells a different story. Challenger, Gray & Christmas tracked 1.2 million layoff announcements in 2025, up 58% from 2024. Then January 2026 hit 108,000 cuts per CNBC, worst January since 2009.
• 1:37
  [Host] Companies aren't announcing "employment shifts." They're announcing headcount reductions.
• 1:41
  [Guest] Exactly. And OpenAI would say those layoffs aren't all AI-driven. DOGE accounted for nearly 294,000 jobs. AI ranked sixth at about 55,000.
• 1:51
  [Host] So the direct AI number is small.
• 1:54
  [Guest] Look at what ranked between them though. "Restructuring," "economic conditions," "closings." Those categories are doing AI's PR work for it.
• 2:03
  [Host] So the cause gets laundered through vaguer language.
• 2:06
  [Guest] Microsoft cut 15,000 workers while investing $80 billion in AI infrastructure. Meta fired 4,200 and redirected the money to training Llama. The stated reason and the actual reason are two different conversations.
• 2:20
  [Host] So you're saying the 18% high-risk number might be low?
• 2:25
  [Guest] I'm saying OpenAI studied occupations in the abstract. The Challenger data tracks what employers are actually doing. Cutting at financial-crisis rates during a period of record corporate profits and positive GDP. That's labor substitution with a polite press release.
• 2:42
  [Host] [serious] What about people entering the workforce right now? If you're 22 and reading "How Will AI Transform Work?", what do you take from it?
• 2:53
  [Guest] You take from it what most new graduates already believe. A 2025 Intelligent.com survey of college seniors found 89% think AI will eliminate their job before they even land one.
• 3:06
  [Host] And the hiring data backs that up?
• 3:09
  [Guest] Entry-level postings dropped 16% on Handshake while applications per opening jumped 26%. The Dallas Fed found workers aged 22 to 25 in the most AI-exposed occupations saw a 13% employment decline since 2022.
• 3:24
  [Host] Software developers especially.
• 3:25
  [Guest] Down nearly 20% in that age bracket from their late-2022 peak. The generation that grew up with AI is the one getting hit first. Over half of Gen Z now holds a side gig, highest of any generation. Only 40% think a four-year degree is worth the cost.
• 3:46
  [Host] So where does OpenAI's nuanced framing actually hold up?
• 3:50
  [Guest] In one narrow sense. Some roles will expand. Healthcare coordination, AI oversight, complex negotiation. Tasks where human judgment navigates ambiguity that models can't resolve.
• 4:01
  [Host] But those expanding roles require credentials most people don't have.
• 4:05
  [Guest] Right. A 2024 Burning Glass Institute report on AI-adjacent job postings found 77% require a master's degree, 18% a doctorate. Meanwhile 83% of U.S. adults hold a bachelor's or less. The jobs AI creates and the jobs AI destroys don't flow to the same people.
• 4:25
  [Host] So the nuanced part is real, but the people who need to hear it most can't access the upside.
• 4:34
  [Guest] That's the tension OpenAI's study doesn't resolve. BLS data from February 2026 shows unemployment at 4.4% while the economy shed 92,000 jobs. People aren't finding new work. They're leaving the labor force entirely.
• 4:48
  [Host] [curious] Which brings up the question of what comes after.
• 4:53
  [Guest] Patrick Deglon frames it in his book as a choice between two paths. Either the gains concentrate at the top, or we build systems that distribute abundance. OpenAI's study maps the terrain. It doesn't choose the direction.
• 5:09
  [Host] For more on AI displacement data and what comes next, head to unscarcity dot ai. The book is on Amazon, the audiobook is on Spotify. This has been Minds, Bodies, and Terawatts.