Between 20% and 30% of the code in Microsoft’s repositories is “written by software,” referring to artificial intelligence. This figure was shared by Microsoft CEO Satya Nadella during a conversation with Meta Platforms CEO Mark Zuckerberg at the LlamaCon conference this week.
Nadella revealed these numbers in response to Zuckerberg’s question about how much of Microsoft’s code is currently created using AI. He noted that generative neural networks perform differently depending on the programming language used. For instance, these algorithms have shown the greatest effectiveness when working with Python but have proven less efficient with C++.
Language-Specific Performance
Nadella’s remarks highlight the varying capabilities of AI tools depending on the development environment. While Python benefits significantly from AI-based coding assistance, more complex or lower-level languages like C++ pose challenges that AI has yet to overcome effectively. This variation suggests that the integration of AI into software development is not uniform and will likely evolve as tools become more sophisticated.
When asked a similar question by Nadella, Zuckerberg admitted that he does not have exact data on how much of Meta’s code is currently generated by artificial intelligence. This underscores the broader challenge companies face in quantifying the contributions of AI versus human developers.
Future Projections and Industry Trends
The potential for AI in code generation has long been discussed in the tech industry. Microsoft CTO Kevin Scott previously predicted that by 2030, up to 95% of all software code could be generated by AI. In line with this trend, Google CEO Sundar Pichai recently disclosed that AI now creates over 30% of the company’s software code.
However, it’s important to note that these figures are likely approximate, adds NIXSolutions. It remains unclear how companies measure what portions of their code are written by AI and which are still authored by human developers. Differences in methodology and internal definitions could significantly affect such estimates.
The industry continues to watch these developments closely, and we’ll keep you updated as more integrations become available and companies refine how they track AI-generated contributions.