AI FOR CREATIVE DESIGN PROCESSES: ITS CURRENT STATE AND CHALLENGES
DS 136: Proceedings of the Asia Design and Innovation Conference (ADIC) 2024
Year: 2024
Editor: Yong Se Kim; Yutaka Nomaguchi; Chun-Hsien Chen; Xiangyang Xin; Linna Hu; Meng Wang
Author: Koga, Tsuyoshi
Series: Other endorsed
Institution: Yamaguchi University
Page(s): 454-457
Abstract
This paper addresses a current status and issues of generative AI by applying a trial implementation example of using AI in product design to the new design of an automobile part by die casting. This paper also indicated that it is difficult to eliminate erroneous output from an AI that has been fed unreliable data by using Transformer technology centered on attention to infer and respond to frequently occurring descriptions from data fed into the AI. In order to reduce hallucination, it is possible to make inferences while considering statistical fluctuations and uncertainties that humans are not good at by using a foundation model that has learned only from in-house data that has full knowledge of the background of the data acquisition for inference. On the other hand, the issue of the need to incorporate hardware implementation into the system due to increased power consumption.
Keywords: Generative AI, AI Assisted Design, Foundation Model, In House Data, Inference