Скачать или смотреть Neural-symbolic artificial intelligence - state-of-the-art, what’s missing, and next steps

THE DEPARTMENT OF COMPUTER SCIENCE GRADUATE SEMINAR 2022
Guest Speaker Series Presented By:

Dr. Bart Gajderowicz
https://bartg.org/
"Neural-symbolic artificial intelligence - state-of-the-art, what’s missing, and next steps"

Friday, September 23, 2022
11:30 am
https://www.lakeheadu.ca/about/news-a...

Abstract:
Neural-symbolic artificial intelligence (or hybrid-AI) is described as the third wave of AI and a requirement for achieving anything resembling artificial general intelligence. It has spurred many new research questions, projects, and funding sources, with a growing number of multi-discipline collaborations across computer science, cognitive science, economics, psychology, sociology, and life sciences. In response, an increasing number of journals, conferences, and workshops are being organized across the world to answer the call. These efforts promise to combine the speed of iterative improvements and robustness of machine learning, mainly deep neural networks, with the conciseness and explainability of symbolic AI, particularly knowledge representation and reasoning and its many subfields, including formal logic, problem definition and problem-solving, common-sense reasoning, ontologies, planning, and scheduling. This talk will cover the current state-of-the-art research towards Neural-Symbolic AI, describe how each branch of AI contributes to this work, and the challenges that remain. As a use case, I will present current research at the Centre for Social Services Engineering at the University of Toronto on developing a neural-symbolic model for natural language understanding in data-poor domains.


Dr. Bart Gajderowicz completed his PhD at the University of Toronto, where he was the Social Service Simulation project director at the Centre for Social Services Engineering research group. His current work focuses on modelling, simulating, and evaluating complex social systems using artificial intelligence, focusing on data-poor domains, with experience working on smart cities and sustainable communities. This work incorporates the development of service-focused ontologies, data standards, AI language models, and knowledge extraction methods. With a focus on data-driven methods, Bart has applied his work in the areas of social service provisioning and impact measurement, cognitive models for simulating social behaviour, information diffusion in social networks, and identification of social factors in market trends.