DIKU Bits: Thomas Troels Hildebrandt
Title
Hybrid/Neuro-Symbolic AI for the People: Combining Sub-symbolic and Symbolic AI to get Accurate and Accountable Chatbots
Abstract
LLMs are easy to use and believed to automate everything - from legal and medical reasoning to solving complex math problems. However, the sub-symbolic neural networks of LLMs may give answers going against established knowledge and legal rules. Symbolic, knowledge based AI on the other hand can be proven to be accurate, but are tedious to build and interact with. The talk will describe how we in the XHAILE research project combine the two technologies to get the best of the two worlds.
Read more about the XHAILe research project
Bio
Thomas Hildebrandt is professor in software engineering and head of the research section for software, data, people & society. With a background in formal process models he has in more than 10 years been leading inter-disciplinary research and innovation projects with focus on methods and technologies for developing reliable and flexible software systems suited for the people who use them, including digitalisation of law, workflows and business processes information systems.
Time and place
March 3th, in Lille UP1 from 12:15-13:00.