Research Raises Concerns Over AI Impact on Code Quality

Share post:

Recent findings from GitClear, a developer analytics firm, indicate that the increasing reliance on AI assistance in software development may be adversely affecting code quality. The study, which analyzed a staggering 150 million changed lines of code from both private and open-source projects, has brought to light several concerning trends, including a rise in code churn and a higher incidence of repetitive code.

GitHub’s Copilot, a leading AI coding assistant used by over a million developers, has been reported to increase task completion speed by 55%, with 46% of code in enabled files being generated by the tool. However, GitClear’s research shifts the focus from the quantity to the quality of code produced with AI assistance. The study points out that AI tools tend to suggest additions to the code but seldom recommend updates, relocations, or deletions. This pattern raises questions about the conciseness and readability of the resulting code.

A significant finding of the research is the increase in code churn, which has escalated to 7.1% from 3.3% in 2020. Additionally, there has been a noted decrease in the movement of code, suggesting a reduction in code refactoring activities. The researchers have expressed concerns over the growing tendency to use copied and pasted code, citing it as a major challenge for the long-term maintainability of software.

While the exact reasons behind these trends are subject to speculation, the researchers associate them with the burgeoning use of AI in coding. They advise engineering leaders to closely monitor these trends and consider their potential impact on future product maintenance.

The study concludes that AI coding assistants, in their current state, are unlikely to replace human developers. The AI tools are deemed too error-prone and not yet capable of securely modifying existing code. This insight may offer some reassurance to developers who fear being supplanted by AI, highlighting the current limitations and challenges of AI in coding scenarios.

Sources Include: The Register

SUBSCRIBE NOW

Related articles

Nvidia CEO Warns U.S. Risks Falling Behind China in AI Talent Race

While demand for Nvidia’s new AI chips surges, CEO Jensen Huang says the greater challenge is America’s shortage...

Amazon’s Project Kuiper Takes Aim at Starlink as Satellite Internet Race Intensifies

Starlink’s dominance in satellite internet is facing its first real test. Amazon’s Project Kuiper has moved from theory...

Judge May Hold Apple In Contempt For Defying Court Order, Opens Door for Fortnite’s Return to iOS

A federal judge has ruled that Apple violated a 2021 injunction by continuing to charge fees on external...

ASUS Tackles GPU Sag with Built-In Gyroscopes in ROG Strix Cards

ASUS is taking a high-tech approach to a common PC hardware problem: graphics card sag. The company will...

Become a member

New, Relevant Tech Stories. Our article selection is done by industry professionals. Our writers summarize them to give you the key takeaways