- 30.1% US Python Code is written by AI coding assistants
- More recent developers are even more likely to use AI
- Technological companies also use more code generated by AI
A new research document entitled “which uses AI to code? The overall distribution and the impact of the generating AI” have found that American software developers are the most intensive users of AI coding assistants in the world.
In December 2024, artificial intelligence would have generated almost one in three Python function (30.1%) by American developers on Github.
This puts American developers in front of their global counterparts in terms of AI use, with countries like German (24.3%), France (23.2%), India (21.6%), Russia (15.4%) and China (11.7%) on the trained.
American developers use the most AI coding assistants
Researchers have also noted that more experienced developers are less likely to use AI (28%) compared to new users of Github (41%) which could be more receptive to the latest additions of the platform.
Despite its enormous productivity promises, AI does not seem to have had such a great impact.
The 30% move of code generated by AI is correlated only with an increase of 2.4% of quarterly commits. The researchers grant the economic value of the AI assisted coding in the United States to $ 9.6 billion and $ 96 billion a year, according to realistic productivity gains observed.
However, Daniotti et al noted for the use of AI could be linked to greater experimentation, with a 2.2% increase in new libraries and a 3.5% increase in new combinations of libraries observed, suggesting that technology could help developers develop in new programming areas.
The trend is correlated with large technological companies such as Google, Meta and Microsoft, which now admit that a large proportion (up to approximately a third) of their code, depending on the project and the use case, is generated by AI.
However, in the case of this study, the researchers admitted that the analysis focuses exclusively on open-source Python projects on Github, so the model actually assumes that the rates of use of AI in Python are observed in other languages.
However, they hope that quantified research could help AI skeptics to make better informed decisions about how they are used and its effects on the labor market.
