Recent research by Model Evaluation & Threat Research (METR) reveals that AI coding tools may actually slow down software developers, contrary to their optimism about such technology. In a controlled trial with 16 experienced developers, it was found that AI tools increased completion time by 19%. While developers anticipated a 24% speed-up, they reported only a 20% perceived improvement despite the actual slowdown. The study highlighted that developers accepted less than 44% of AI-generated code, often requiring significant revisions, resulting in approximately 9% of their time spent on code cleanup. Factors contributing to the slowdown include over-optimism regarding AI’s utility and challenges in complex coding environments. Although current AI tools may hinder productivity, the METR study acknowledges the potential for future improvements in AI capabilities. Such contrasting findings emphasize the need for a more nuanced understanding of AI’s impact on software development, particularly in real-world scenarios. Follow ITPro for updates on technology trends.
Source link
