Abstract

This work in progress explores the potential of large language models (LLMs) for rapid prototyping of Internet of Things and smart home systems. The study demonstrates the capabilities of LLMs in generating simulated sensor data and making decisions for such systems. By presenting a proof of concept of an iOS app that uses GPT 4 API for decision-making and device generation, the study proposes that LLMs are helpful not only in mimicking the role of physical sensors but may also prove useful in replacing human actors in projects that employ wizard-of-oz prototyping method. This approach helps practitioners to simulate various scenarios without using any hardware. Furthermore, this study calls for the incorporation of qualitative reasoning within the prototyping framework, emphasizing the significance of including nuanced insights obtained from LLM-generated data based on user prompts. Our technique aims to enrich the prototype process by encouraging reasoning with qualitative data, making it easier to explore multiple scenarios and improving the robustness of IoT and smart home systems.