Home // CENTRIC 2023, The Sixteenth International Conference on Advances in Human-oriented and Personalized Mechanisms, Technologies, and Services // View article

LLM Assisted No-code HMI Development for Safety-Critical Systems

Authors:
Matthias Harter

Keywords: requirements engineering; safety critical embedded software; model-based software design; automatic code generation

Abstract:
This paper represents the outcome of an empirical study conducted with Large Language Models (LLM) on the question whether or not we can expect current and future Artificial Intelligence to assist engineers in the development of embedded software systems for safety-critical applications. Experiments with GPT-4 and other LLMs suggest that current models are capable of assisting developers in part in the design of Human-Machine-Interfaces (HMI) for instruments and displays used in aircrafts without the need to manually write a single line of source code (no-code development) while maintaining the highest level of safety and reliability demanded by authorities and customers. The study does not present generally accepted quantitative measures in answering the question how well suited current language models are for this task, but rather provides a qualitative assessment of the capabilities of state of the art AI. It also sheds a light on the deficiencies of today’s LLMs in fully understanding technical systems in depth. Instead of completely replacing human engineers we should rather strongly rely on human-in-the-loop policies for the most critical phase of the progressively automated development process, even with more sophisticated and powerful LLMs on the horizon. It should be noted that providing objective evidence to support the argumentation and the findings in this short paper will be the subject of future work.

Pages: 8 to 18

Copyright: Copyright (c) IARIA, 2023

Publication date: November 13, 2023

Published in: conference

ISSN: 2308-3492

ISBN: 978-1-68558-100-8

Location: Valencia, Spain

Dates: from November 13, 2023 to November 17, 2023