On November 8, 2024, Corey Clark, PhD, of the SMU Guildhall and Department of Computer Science, took the stage at the Technology-Enhanced Immersive Learning (TEIL) Seminar Series with a presentation that invited participants to think beyond traditional AI uses. His talk, titled “Bridging the Gap in AI: From Language Models to Intelligent Collaborators,” attracted an audience of students, educators, and researchers. They gathered to explore how AI, specifically language models, can evolve into true collaborators rather than remaining tools of passive use.
The Vision of AI as a Collaborator
Dr. Clark began by discussing the limitations of current AI language models. Although they process enormous data and mimic human responses convincingly, they lack foundational understanding. According to Dr. Clark, this gap prevents language models from performing tasks that require context-based decisions, critical thinking, and reasoning. His lab, the Human and Machine Intelligence (HuMIn) Game Lab, is dedicated to tackling these very challenges, pushing AI beyond simple pattern recognition to true collaboration. This approach, as he highlighted, relies on blending language models with various tools, data sets, and knowledge frameworks, creating “intelligent collaborators” that can support complex tasks across different fields.
AI in Action: The Case for Autonomous Agents
Dr. Clark highlighted his dynamic team, including postdoctoral researchers, graduate students and undergrads, and their work on autonomous agents. Dr. Clark explained that these agents are equipped with tools beyond basic language processing, like knowledge graphs and context-based filtering, allowing them to interact effectively with both structured and unstructured data.
A key moment was the introduction of an AI agent prototype that can solve problems in real-time and assess its own performance. This agent exemplifies “self-learning with a graph chain of thought,” allowing it to answer questions by creating its own contextual understanding of data—enabling a remarkable jump in accuracy from 40% to 96%.
Bringing Personality to AI: A New Frontier
Dr. Clark also discussed his project on “driving agents with personality.” His team used a dataset of 50,000 human personality tests to train language models to mimic specific personality types. The results were impressive: the model’s responses closely matched real human data, showing it could consistently respond within a chosen personality style. This breakthrough paves the way for AI applications in areas needing human-like interaction, such as education, mental health, and customer service.
Looking Forward: Collaborative Agents in Gaming and Beyond
Dr. Clark concluded with an eye on the future. He expressed optimism that, as AI continues to integrate into different sectors, it can transform into a reliable collaborator, not just a programmed responder. The projects underway at the HuMIn Game Lab reflect this aspiration, focusing on AI agents that are explainable, adaptable, and able to say “I don’t know”—a fundamental step toward building trust with human users.
If you are interested in collaborating with peers in technology-enhanced learning, immersive learning, and AI/machine learning spaces, join us at the upcoming TEIL seminars. Dr. Jiun-Yu Wu, Professor in the Department of Teaching and Learning at SMU, will be presenting on December 6. More information on TEIL at SMU is available at smu.edu/teil.