Quantum Computational Intelligence with Generative AI Image for Human-Machine Interaction

Chang-Shing Lee, Mei-Hui Wang, Jun-Kui Chiang, Naoyuki Kubota, Eri Sato-Shimokawara, Yusuke Nojima, Giovanni Acampora, Pei-Yu Wu, Szu-Chi Chiu, Sheng-Chi Yang, Chyan Zheng Siow

Abstract: This paper introduces a Quantum Computational Intelligence (QCI) agent equipped with a content attention ontology model, specifically designed to enhance human-machine interaction based on a Generative Artificial Intelligence (GAI) image generation agent for Taiwanese/English learning and experience. Its diverse primary applications include social media analysis on Facebook groups and YouTube learning videos related to the 2023 IEEE CIS Education Portal (EP) Subcommittee, as well as in the areas of Taiwanese/English language learning and dialogue experience with GAI image generation. To establish the knowledge and inference models for the QCI agent, we initially developed a Taiwanese/English learning and experience ontology, including a content attention ontology, and an image attention ontology. The QCI agent utilizes metrics such as the number of views, posts, and comments to predict the fuzzy number of reactions. In addition, the GAI image agent generates Taiwanese speech-based/English text-based images and evaluates the fuzzy similarity score between Taiwanese/English and the attention ontology together with the Sentence BERT (SBERT) agent. This Taiwanese/English fuzzy similarity score is further validated through human assessments, with these evaluations subsequently serving as an additional metric for comparative analysis of Human-Machine Interaction (HMI). Furthermore, the GAI image agent is designed to create images and Chinese/English texts from text/speech translated by the Meta AI Universal Speech Translator (UST) Taiwanese/English agent. A Particle Swarm Optimization (PSO)-based machine learning mechanism is employed to train the QCI model for assessing learners' performance and predicting the performance of others. The National University of Tainan (NUTN) Taiwan-Large Language Model (NUTN.TW-LLM) agent has been further enhanced to support interactive learning experiences for HMI. An SBERT-based assessment agent is used to calculate fuzzy similarities between questions and answers in Taiwanese/English experiences and dialogues. Experimental results demonstrate the feasibility and efficacy of the proposed QCI model, equipped with QCI&AI-FML (Artificial Intelligence-Fuzzy Markup Language) and machine learning capabilities, for social media and language learning applications on HMI. In the future, we will extend the QCI model to various HMI applications for student learning around the world.