By Jayujyoti Mullick
A recent analysis by AI researcher Andrej Karpathy showed that artificial intelligence could actually reshape and reconstruct the global job market. Several high-paying white-collar professions might face a growing risk of automation. The OpenAI co-founder and former Tesla AI chief analysed and ranked hundreds of professions according to their exposure to the artificial intelligence systems, triggering widespread debate among technology experts and workers alike.
Karpathy’s analysis report has evaluated 342 occupations, according to the U.S. Bureau of Labor Statistics by assigning each role an exposure score ranging from zero to ten. This scorecard indicated how easily AI tools might be able to perform the tasks involved for the mentioned job.
The scorecard suggested that several digital and high-skilled professional jobs, including software developers, financial analysts, writers, editors and graphic designers, could face significant disarrangements due to ongoing advancement of artificial intelligence capabilities.
READ: White House doubles down on firing of Labor Statistics head (
The study by Karpathy gained even more attention following Tesla CEO Elon Musk’s weightage on the broader implications of AI’s rapid development. Elon Musk suggested that automation could transform the concept of employment altogether in the long run.
Karpathy’s scorecard highlighted the effect of artificial intelligence on professionals that primarily involve digital work, data analysis or written communication. Roles producing digital outputs and relying heavily on computer-based tasks are probably among those most vulnerable in the analysis. Computer programmers, database administrators, mathematicians and other technology professionals were also listed among professionals with high exposure to AI advancements.
However, jobs dependent on physical labor or direct human interaction appear to be far less likely affected by AI. Professions such as construction workers, janitors, roofers and ironworkers ranked among the least exposed in the analysis. Also, service roles such as bartenders, barbers, nursing assistants and home health aides also received low threat scores because they require physical existence and personal interaction with consumers that current AI models cannot easily replicate on a large scale.
However, the report quickly went viral online, sparking strong debate on the existence of artificial intelligence as a replacement of large segments of the modern workforce. With the ongoing conflict about the research, Karpathy later clarified that the project was a quick experimental exercise rather than a definitive prediction about the future of employment.
After the analysis spread widely on social media, he said the project had been misunderstood and described it as an exploratory attempt to visualize occupational data.
READ: The American way of firing: Sudden layoffs are reshaping India’s white-collar work culture (
“This was a saturday morning 2 hour vibe coded project inspired by a book I’m reading. I thought the code/data might be helpful to others to explore the BLS dataset visually, or color it in different ways or with different prompts or add their own visualizations. It’s been wildly…”
For decades, the technological disruption was widely feared to affect manual labor first, while professional and knowledge-based jobs have been considered relatively secure. However, with the advancements in generative AI, this has been challenged. Artificial Intelligence demonstrates the ability to write text, generate code and analyze large datasets at a remarkable speed.
The viral analysis report also led to an unexpected development. Karpathy later removed many parts of the project, including the public code repository that hosted the data behind the AI exposure rankings. Reports indicate that while the interactive website showing the visualization is available, the GitHub repository containing the raw data and code have been deleted after the analysis began spreading conflict rapidly across various social media platforms.
The removal of the repository added another layer of debate to the ongoing investigation surrounding the study. The evaluation had analyzed 342 occupations representing roughly 143 million U.S. jobs and assigned each role to an AI exposure scorecard using a large language model.
According to estimates associated with the project, about 42 percent of jobs received exposure scores of seven or higher, representing nearly 60 million workers and around $3.7 trillion in annual wages across the U.S. economy.