A recent study has found that artificial intelligence (AI) can actually hinder software developers' productivity, contrary to the expectation that it would increase efficiency. The study involved 16 software developers with an average of five years of experience, who were tasked with completing 246 tasks with and without the use of AI tools. Surprisingly, the results showed that AI led to a decrease in productivity, with task times being higher than without the technology.
The study revealed that developers lost time writing prompts for chatbots, waiting for AI to generate results, and cleaning up AI-generated code. Experienced developers were particularly affected, as they had to transform their agendas and problem-solving strategies into outputs that AI could understand. While AI can generate code, it often requires manual review and editing to ensure quality and relevance.
Economist Daron Acemoglu believes that markets overestimate the productivity gains from AI, suggesting that only 4.6% of tasks in the US economy will become more efficient with AI. Anders Humlum, assistant professor of economics, emphasizes the need for critical thinking about when to implement AI tools and notes that getting productivity gains from technology requires organizational adjustments, supplementary investments, and improving worker skills through training.
The study's findings highlight the importance of understanding the limitations of AI in software development. While AI-powered tools like GitHub Copilot, Tabnine, and DeepCode can be valuable for developers, they should be used judiciously. As the use of AI in software development continues to grow, it's essential to consider the potential impact on productivity and to develop strategies for effective implementation.
