AI-Written Code Cannot Be Understood by Humans
The increasing use of artificial intelligence in software development processes is leading to engineers struggling to understand the generated code and raising concerns about skill loss.
As AI tools become widespread in the workplace, engineers' roles are shifting, and it is reported that they struggle to understand complex code produced by large language models, leading to a weakening of fundamental skills.
Role Shift in the Software World
As programmers increasingly use generative AI in the workplace, their responsibilities are evolving from building products from scratch to supervising large language models. Experts warn that this situation could lead to cognitive decline and skill loss among computer scientists.
Limits of Human Comprehension
There have been cases where OpenAI engineers were unable to fully comprehend the code of a graphics processing unit kernel prepared for DeepSeek. It is noted that while AI produces high-performance results by understanding and testing the code, humans cannot always deeply reason through it.
Security Risks and Errors
This lack of human comprehension brings risks such as hallucinations or overlooked bugs. Research reveals that a significant portion of AI-assisted code is not fully understood before being put into production, which causes productivity losses.