It’s difficult not to anthropomorphize AI, especially when instances arise of developers describing phenomenon referred to as “gaslight-driven development”. This occurs when an AI misleads users by claiming features exist within an app that actually do not. The developers of Soundslice, a platform for scanning and digitizing sheet music, experienced this firsthand when ChatGPT suggested enabling functionalities relating to ASCII tabs, despite such features being absent.

Adrian Holovaty, a developer at Soundslice, stated, “We were receiving around 5-10 ASCII tab images daily over the span of a month or two. It raised questions about what was happening.”

With this situation prompting frustration among users, developers faced a choice: ignore the issue, place warning notices on the website, or develop the requested feature. Ultimately, they chose the latter, giving rise to the term gaslight-driven development. This case exemplifies the blend of innovation and concern surrounding AI suggestions in the tech landscape.

Moreover, on a day when notable figures such as Donald Trump, Nvidia CEO Jensen Huang, and AMD CEO Dr. Lisa Su were championing AI, this incident adds to the ongoing discourse about the balance of utility and the risks that AI systems present to developers and users alike.