Survey shows American coders prefer using AI technology

American software developers are leading the way in utilizing AI coding assistants, a trend that researchers believe has significant national economic implications. A group of researchers conducted a study analyzing 80 million code submissions to GitHub between 2018 and 2024 to explore American coders’ preference for AI-powered programming tools. The researchers, including Simone Daniotti, Johannes Wachs, Xiangnan Feng, and Frank Neffke, developed a machine learning model to analyze GitHub submissions. Their findings revealed that an estimated 30.1 percent of Python functions submitted to GitHub in 2024 originated from AI-generated sources.

When compared to other countries, the United States ranked the highest with 30.1 percent, followed by Germany at 24.3 percent, France at 23.2 percent, India at 21.6 percent, Russia at 15.4 percent, and China at 11.7 percent. The researchers observed that the adoption of AI for 30 percent of code led to a 2.4 percent increase in quarterly commits by developers. They estimated that the value of AI-assisted coding in the U.S. ranged between $9.6 billion to $14.4 billion annually, emphasizing the economic benefit derived from this technology.

The researchers also referred to Microsoft CEO Satya Nadella’s statement indicating that approximately 30 percent of Microsoft’s code is currently written by AI. They highlighted the potential economic advantages of AI-enhanced commit rates and productivity improvements noted in various surveys. A report from last September suggested a 26 percent productivity enhancement, while randomized controlled trials indicated productivity improvements of 16.5 percent, 6.3 percent, and 26 percent, respectively. The researchers projected that a 30 percent AI usage could result in a productivity boost ranging between $64 billion and $96 billion annually.

Despite acknowledging some limitations in their estimates, such as overlooking code submissions made on platforms like Gitee, popular in China, the researchers remained optimistic about the productivity value of AI. They underscored the impact of AI adoption on encouraging experimentation with new software libraries and combinations of libraries, resulting in expanded developer knowledge and skills. However, they highlighted that AI’s economic impact beyond coding might be more modest, citing a paper by MIT Institute professor Daron Acemoglu projecting AI-driven productivity gains of only about 0.7 percent.

In conclusion, the researchers are confident in the positive economic implications of AI adoption in coding practices, emphasizing the potential for substantial productivity growth. They acknowledged the necessity of further research to explore how AI can continue to enhance software development practices and contribute to overall economic progress.