4/29

12pm

Carnegie 113

Please join the Cognitive Science and Computer Science Departments on Wednesday, April 29th, for a seminar by Dr. Ora Lassila titled "Symbolic AI in the Age of LLMs". Pizza and refreshments will be provided.

While the new "Generative AI” and Large Language Models (LLMs) capture headlines, the untapped potential lies in combining these technologies with decades-proven symbolic AI techniques. In this talk we review the history of AI and discuss symbolic techniques such as ontologies, logic-based reasoning, rule-based systems, constraint satisfaction, and planning. Many questions asked about AI already decades ago are now relevant again to understand the new forms of AI. Older symbolic AI techniques are key ingredients when building autonomous, “agent-based” AI systems. Symbolic techniques, especially ontologies and reasoning, also form the underpinnings of modern knowledge graphs. These graphs have been found to be an essential component when leveraging LLMs and when mitigating the issues and shortcomings of the new technology.

Bio:

Dr. Ora Lassila has studied, researched, and applied AI technologies his entire professional career, while working in organizations such as CMU, MIT, Nokia Research, Pegasystems, State Street, and Amazon Web Services. In the late 1990s he took his experience of symbolic AI and frame-based knowledge representation and started the work on the W3C RDF representation language, arguably the most successful effort in taking knowledge representation research “from the lab” and moving it into the software industry “mainstream”. He was a co-author of the article that laid out the vision for the so-called Semantic Web, and most recently has been co-chairing the W3C working group that is defining the next version of RDF. He received his doctorate in CS and AI from the Helsinki University of Technology.