An OpenAI cofounder ‘vibe coded’ an analysis of the U.S. labor market’s exposure to AI, and the highest-paying jobs have the worst scores
In a recent analysis, Andrej Karpathy, an OpenAI cofounder and former director of AI at Tesla, explored the vulnerability of various professions in the U.S. labor market to artificial intelligence (AI) and automation. This analysis has raised significant concerns regarding the future of jobs as AI technology continues to evolve.
Understanding the Analysis
Karpathy utilized data from the Bureau of Labor Statistics (BLS) to evaluate the exposure of different occupations to AI. He created a scoring system ranging from 0 to 10, where 10 indicates the highest level of exposure to AI and automation. The overall weighted exposure score for the labor market was calculated to be 4.9.
Key Findings
One of the most striking revelations from Karpathy’s analysis was that high-paying jobs, specifically those earning over $100,000 annually, had an average exposure score of 6.7. In contrast, professions with salaries below $35,000 had a much lower average exposure score of 3.4. This trend suggests that as salaries increase, so does the potential for job displacement due to AI.
High-Exposure Professions
According to Karpathy’s findings, several high-paying professions received exposure scores of 9, indicating a significant risk of automation. These professions include:
- Software Developers
- Computer Programmers
- Database Administrators
- Data Scientists
- Mathematicians
- Financial Analysts
- Paralegals
- Writers and Editors
- Graphic Designers
- Market Researchers
The high exposure scores for these roles are largely attributed to the increasing sophistication of AI tools, which can perform complex tasks that previously required significant human effort.
Low-Exposure Professions
On the other end of the spectrum, professions such as construction laborers, roofers, painters, janitors, and grounds maintenance workers received much lower exposure scores, typically around 1 or 2. For example:
- Construction Laborers: Score of 1
- Home Healthcare Aides: Score of 2
- Nursing Assistants: Score of 2
- Massage Therapists: Score of 2
- Veterinary Assistants: Score of 2
- Barbers and Bartenders: Score of 2
This data suggests that jobs requiring physical presence and hands-on skills are less likely to be replaced by AI in the near future.
Reactions to the Analysis
Karpathy’s graphic quickly gained traction online, sparking discussions and predictions about a potential “jobs apocalypse” for white-collar workers. However, he later removed the data, explaining that it was a “vibe coded” project inspired by a book he was reading. He expressed concern that the data had been misinterpreted and hoped to provide a tool for others to explore the BLS dataset visually.
Broader Context of AI in the Labor Market
While Karpathy’s analysis raised alarms, it also aligns with broader discussions regarding the impact of AI on the labor market. A report by AI startup Anthropic titled “Labor Market Impacts of AI: A New Measure and Early Evidence” indicated that actual AI adoption is still a fraction of its potential capabilities. The report highlighted that older, highly educated, and well-paid workers are at a greater risk of displacement.
Contradictory Evidence
Despite the fears surrounding AI-induced job losses, some analysts argue that the economy is not on the brink of collapse. A report by Citadel Securities countered the doomsday narrative presented by Citrini Research, which predicted catastrophic economic impacts due to AI. Citadel pointed out that job postings for software engineers had increased by 11% year-over-year in 2026, indicating a growing demand for tech professionals.
Moreover, the report noted that the daily use of generative AI for work remains stable, with little evidence suggesting imminent displacement risks. Instead of a declining economy, the U.S. is experiencing a surge in new business formations, and the construction of AI data centers is driving a localized boom in construction hiring.
Economic Boundaries of AI Adoption
Citadel also emphasized that if automation were to expand rapidly, the demand for computational resources would increase, potentially raising costs. This could create a natural economic boundary, where the marginal cost of compute exceeds that of human labor for certain tasks, preventing widespread substitution.
Conclusion
As the landscape of the labor market continues to evolve with the advancement of AI technologies, the findings from Andrej Karpathy’s analysis provide valuable insights into the potential vulnerabilities of various professions. While high-paying jobs may face greater risks, the overall impact of AI on the economy remains a complex and debated topic. The conversation surrounding AI’s role in the labor market is ongoing, and it will be crucial for workers, employers, and policymakers to navigate these changes thoughtfully.
Note: The information presented in this article is based on various sources and analyses as of October 2023 and may be subject to change as new data emerges.
