In a recent study conducted by the GitClear analytical company, the use of generative artificial intelligence in software development has been found to have various consequences. The study analyzed 153 million lines of changed code, comparing data from 2023 to previous years when AI did not play a significant role in code generation. The results are significant in shedding light on the impact of AI in this field.
The study highlights several key findings:
- There is an increase in “Code Churn,” which refers to the percentage of code lines discarded within two weeks of writing. It is expected that this indicator will double by 2024, which consequently raises the risk of errors when implementing the code in production.
- The growth of copypasting in code surpasses the rate of code renewal, removal, or movement. GitClear suggests that the composition of AI-generated code resembles the work of a short-term developer who does not integrate their work carefully into the larger project.
GitClear emphasizes that AI tools excel in code addition but can result in the accumulation of “technical debt” caused by AI. Additionally, the study raises concerns about the wages of engineers. GitClear suggests that if managers evaluate engineer salaries based on the number of changed code lines, combining this approach with AI can lead to the creation of poor-quality code.
However, it is difficult to determine whether AI tools will have an overall positive impact on software development. GitClear acknowledges the benefits of using AI to obtain personalized programming answers but also highlights that reading bad code is one of the most challenging tasks for developers.