The maintainer of the Dillo web browser has proposed a method for filtering out changes made through AI. The Dillo project currently only accepts patches created by individuals, but determining whether a submitted patch was created by a person or AI can be time-consuming. To address this issue, developers submitting changes for the first time are being asked to provide a recording of their development session as proof of their work. When using Vim, developers can record their session using the asciinema utility.
Although the idea is still in the discussion and experimentation stage and has not been made mandatory, it aims to make it easier to distinguish between patches created by humans and those generated by AI. However, there is concern that developers could potentially use AI to falsify session recordings to pass off AI-generated patches as human-written code.
Experiments have shown that the artificiality of such recordings is readily apparent, as they do not capture the human-like actions, mistakes, and thought processes that occur during the coding process. AI models are typically trained on existing code, which does not fully encompass the nuances of human coding practices.