The article by Damien Masson et al. presents DirectGPT, an interface layer on top of ChatGPT, designed to improve the interaction with large language models (LLMs). The authors argue that direct manipulation principles, such as continuous representation of generated objects, reusing prompts in a toolbar of commands, manipulable outputs, and undo mechanisms, can enhance the interaction with LLMs. A study showed participants were 50% faster and relied on fewer and shorter prompts to edit text, code, and vector images compared to baseline ChatGPT. The authors suggest this approach could be integrated into traditional software to improve the use of LLMs.
Publication date: 5 Oct 2023
Project Page: https://arxiv.org/abs/2310.03691
Paper: https://arxiv.org/pdf/2310.03691
