Want the benefit of an AI-assisted code generation program like GitHub Copilot, but don’t want to be tethered to the Microsoft universe? Dr. Brendan Dolan-Gavitt has developed an alternative. In the summer of 2022, the assistant professor of computer science and engineering at NYU’s Tandon School of Engineering introduced the cheekily-named FauxPilot. The primary goal of the new program is to eliminate the need to share information with third parties, and by doing so, avoid potential copyright issues and other complications.
“There are people who have privacy concerns, or maybe, in the case of work, some corporate policies that prevent them from sending their code to a third-party,” Dolan-Gavitt told Thomas Claburn in an article in The Register. “FauxPilot addresses these concerns and that definitely is helped by being able to run it locally.”
By “running locally,” Dolan-Gavitt is referring to FauxPilot’s ability to run AI assistance software on “premise.” This differs from GitHub Copilot, which through telemetry can end up sending some data back to GitHub, which is owned by Microsoft. FauxPilot also reduces, if not totally eliminates licensing concerns, because, unlike Copilot, it does not utilizes OpenAI Codex. As explained in the article, this “natural language-to-code system” was “trained on ‘billions of lines of public code‘ in GitHub repositories.” Thus, it’s possible the system incorporates copyrighted data from other sources that could expose users to infringement threats.
You can read the full article, which was originally published on August 6, 2022, here.