Anthropic's 2030 AI economy model ties rapid growth to weaker knowledge-worker jobs
- Anthropic's Economics team built a US model that projects how AI capability and adoption could affect jobs, GDP growth, unemployment, and wages through 2030.
- The model treats each occupation as a bundle of O*NET tasks, then classifies AI's effect on each task as unchanged, augmented, automated, or a newly created task.
- In scenarios where AI raises growth to about twice the normal rate, unemployment remains within its historical range and wages stay flat or rise depending on the industry.
- In scenarios with growth faster than any prior economic period, the model projects worse wages and job prospects for knowledge workers, despite much higher aggregate wealth.
- Anthropic's interactive scenario explorer lets users enter assumptions about AI capability and economy-wide adoption, then compare the resulting 2030 projection with other users' predictions.
Hacker News opinions
I found the scroll-driven presentation completely unreadable. If an AI company cannot make a page usable without destroying scrolling, that is a bad look.
Anthropic's message has swung from Amodei's doom warnings to "it will not be so bad." I see vague assumptions and equations that few economists will inspect, packaged for a reassuring policy message.
The honest answer is that nobody knows what the economy will look like in 2030. Predicting the future is a fool's errand.
I do not see the scenario where Anthropic is wrong about AI's value in kind, including a net-negative economic effect. Leaving that out is a major blind spot for labs whose salaries depend on AI succeeding.
I expect AI to change society, but I doubt most companies spending trillions on it will get positive ROI. A cheap open-weight Chinese model that trails by six months could wipe out the economics of many AI vendors.
These look like marketing materials dressed up as studies, not genuine independent research.
If "wrong" means AI has less impact than claimed, business as usual is already the baseline. I can see why the paper focuses on unusual futures, though a severe AI-investment recession is still plausible.
Current AI is already affecting jobs. Ask junior graphic designers, web developers, educational book writers, and developer advocates.
Four years in, I have not seen anything meaningful enough to justify these forecasts.
My job market has clearly worsened, and AI has changed work for many people. Saying nothing meaningful has happened ignores that.
The nurse example feels economically naive. If AI lets one nurse do the work of two, a cost-driven employer will usually hire fewer nurses rather than give each patient twice as much attention.
I do not expect the social order to redistribute productivity gains on its own. Without worker organizing and action by people with power, most of the gains will go to wealth holders.
In ordinary markets, fewer labor hours should mean lower prices, more access to services, or both. Healthcare is unusual because policy and regulation distort those supply-and-demand effects.
I do not trust Anthropic to describe AI's economic effects neutrally because it has an incentive to call the outcome net positive.
A nurse whose role becomes obeying AI recommendations may lose agency and autonomy, even if task output rises.
I do not think AI can be a healthcare decision maker when a wrong decision can kill someone. "The AI did it" is no excuse if a nurse has 15 patients and misses a dangerous treatment change.