GovAI and Brookings: many of the people most at risk if AI transforms work are also the best placed to find new jobs, as researchers disagree over AI's impact
GovAI and Brookings add a transition lens to that debate: workers whose roles face the most change may not be the workers least able to move into new work. That distinction matters because exposure estimates and displacement outcomes are not interchangeable.
First-order effects
The study challenges a simple reading of AI exposure as a proxy for workers’ reemployment risk, especially for the people it identifies as both highly exposed and well positioned to find new roles.
GovAI and Brookings put worker mobility—not only the likelihood of task transformation—at the center of near-term assessments of AI’s employment effects.
Second-order effects
Employers and policymakers evaluating AI workforce risk will need to separate job-change exposure from workers’ capacity to transition, rather than treating high exposure as synonymous with unemployment.
The finding strengthens the case for tracking observed labor-market outcomes alongside competing forecasts, a dispute highlighted by contrasting research on US employment trends.
Third-order effects
If the pattern persists, AI labor policy could focus more on transition pathways, training, and matching than on occupation-wide job-loss estimates alone.
The durable question for the industry may shift from how many jobs AI affects to whether affected workers can move into new work quickly enough; the current research disagreement means that remains unsettled.
The trend: AI’s labor-market debate is moving from headline job-displacement forecasts toward measurement of task change, worker mobility, and realized transitions.
“If your whole job appears on a screen you're cooked” = PMC and menial service class Armageddon. Basically, the entire “knowledge worker” or “information worker” is the first to get wiped. Just like mechanical engineers in Detroit in the early 50's who set up all the
I'd bet $1000 that from now to 2030, most “susceptible” jobs see *increased* share of labor. In the models these types of charts are based on, it is explicitly not “AI can substitute” but “AI is related”. AI is a complement too! Who *doesn't* want to code right now, for instance?
I read the methodology. The scores are generated by.........asking Gemini flash to make one up?!? Am I missing something? I really like karpathy but this doesn't seem useful at all
Karpathy scored all 342 US occupations for AI exposure. Everyone's sharing the treemap. The actual insight is one line in the scoring rubric: If your entire job happens on a screen, you're exposed. Software devs: 8-9. Roofers: 0-1. Medical transcriptionists: 10/10. 342
Andrej Karpathy has launched an analysis of AI on jobs. Karpathy is by no means interested in hype or exaggeration. Using AI, he concluded that out of 143 million working people in the US, approximately 57 million are at high to very high risk of their jobs being negatively [imag…
5 minutes ago, @karpathy just dropped karpathy/jobs! he scraped every job in the US economy (342 occupations from BLS), scored each one's AI exposure 0-10 using an LLM, and visualized it as a treemap. if your whole job happens on a screen you're cooked. average score across [imag…
New: AI job exposure is important, but there's more to the picture. Here's what the research says on adaptability. — Find your job here 🎁 wapo.st/4cP2ZHM [image]
The feared AI jobs apocalypse may be fine for relatively affluent workers like programmers. — It may not be for clerical and administrative workers. — Great visuals by @kevinschaul.bsky.social with a reminder: We are historically awful at predicting how tech changes work. — …