The people building the AI that might replace you think it's coming faster and harder than you do.

The Summary

The Signal

This is what belief gap looks like at scale. Most Anthropic employees think AI will absorb at least 80% of their own jobs by 2030. Only 5% of regular US workers see the same fate for themselves. That's not a difference of opinion. That's two different movies playing on the same screen.

The implications cut deeper when you look at what Anthropic staffers do. These aren't factory workers watching robotic arms on an assembly line. The median employee expects 95% of the work of building new AI models to be automated by 2030. They're predicting their own obsolescence while still cashing paychecks to build the thing that makes them obsolete.

"Anthropic runs roughly a year and a half ahead of the economy on AI adoption, so what happens at Anthropic could be a leading indicator."

Anthropic's updated research adds robotics to the equation, which changes the threat surface. Jobs that felt safe because they required physical presence, gardening among them, now rank higher on the disruption scale. The theory is sound: AI provides the intelligence, robots provide the embodiment, and suddenly weeding a flowerbed becomes automatable.

But here's where theory meets dirt. The author notes the reality gap: identifying the right plant, kneeling in awkward spots, pulling weeds by the roots without damaging surrounding plants. These aren't chess problems. They're messy, variable, context-dependent tasks that break robots in ways that look trivial on paper but brutal in practice.

Key tensions in the data:

  • Anthropic employees are more optimistic about AI progress and economic growth than the general population
  • They simultaneously predict unemployment topping 12% by 2030
  • The robotics addition to the research dramatically shifts which jobs appear vulnerable

The expectation gap matters because it reveals information asymmetry. Anthropic employees see the training runs, the capability jumps, the internal roadmaps. They know what's in the pipeline. US workers see ChatGPT write emails and assume that's the ceiling. One group is building the future. The other group is reading about it in headlines.

The Implication

If you're betting your career on physical tasks keeping you safe, the AI+robotics combination changes the odds. But don't panic about gardening robots just yet. The distance between "can be automated in theory" and "is worth automating in practice" is measured in decades of messy implementation, not years of exponential curves.

More importantly, pay attention to belief gaps like this one. When the people building the technology have radically different expectations than the people using it, someone's going to get surprised. The smart move is to run the same mental exercise Anthropic employees are running: what percentage of your current job could AI handle by 2030? If your answer is below 20% and you're not a gardener, you might be the one who's wrong.

Sources

Business Insider Tech | Business Insider Tech