Andrej Karpathy, former co-founder of OpenAI and ex-head of AI at Tesla, has released a comprehensive interactive map that scores every major occupation in the United States economy by its exposure to AI replacement. The tool, available at karpathy.ai/jobs, evaluates 342 occupations using Bureau of Labor Statistics data and assigns each a score from 0 (no AI risk) to 10 (fully replaceable).
Key Findings
- The average AI exposure score across all U.S. jobs is 5.3 out of 10
- 42% of all jobs (roughly 59.9 million positions) score 7 or higher
- $3.7 trillion in annual wages sit in occupations scored 7 or above
- Jobs paying over $100K average an exposure score of 6.6, nearly double the 2.9 average for jobs under $35K
Most Exposed Occupations
Medical transcriptionists are the only profession to receive a perfect score of 10 out of 10, indicating near-complete replaceability by AI. Software developers scored between 8 and 9, with the Bureau of Labor Statistics already projecting a 6% decline in programming positions.
Other high-risk categories include:
- Customer service representatives (2.8 million workers, score 9)
- General office clerks (2.6 million workers, score 9)
- Bookkeepers and accounting clerks (1.6 million workers, score 9)
- Data analysts and financial analysts (high scores across the board)
Safest Professions
The pattern is clear: if a job requires physical presence and hands-on interaction, AI poses little threat. Roofers, painters, ironworkers, janitors, plumbers, and electricians all scored between 0 and 1. Nurses and home caregivers also remain well-protected.
The fastest-growing and least exposed jobs are wind turbine technicians (+50% projected growth, exposure score 2) and solar panel installers (+42% growth, exposure score 2).
The Income Paradox
One of the most striking revelations is the inverse relationship between income and job security. College graduates face an average exposure score of 6.7, while workers without a degree average just 4.1. Higher-paying, screen-based knowledge work faces the greatest disruption risk.
A Dual-Track Future
The analysis paints a picture of a workforce splitting into two tracks: rapid AI integration in knowledge work versus resilience in physical trades. However, experts note that prompt engineering skills and the ability to effectively direct AI tools can substantially reduce personal replacement risk, even within high-exposure job titles.
Karpathy's map builds on similar research from organizations like Anthropic and the World Economic Forum, which has projected that while approximately 85 million jobs could face displacement by AI, around 97 million new positions may emerge in the process.
What This Means
For workers, the message is nuanced: screen-based jobs face real transformation, but adaptation through AI literacy and upskilling can turn exposure into opportunity. For businesses, the map highlights sectors ripe for AI copilot tools, workflow automation, and training products targeting high-exposure roles.
The interactive visualization is open source and available for anyone to explore at karpathy.ai/jobs.
Source: Andrej Karpathy