UMBC Ebiquity Research Lab presents
Can we ever trust our chatbots?
Towards trustable collaborative assistants
Professor and AAAS Leshner Fellow
Artificial Intelligence Institute, University of South Carolina
3-4pm ET Wed, 16 Nov. 2022, ITE 325B and Webex
AI services are known to have unstable behavior when subjected to changes in data, models or users, especially in sustainable and societal applications. Such behaviors, whether triggered by omission or commission, lead to trust issues when AI works with humans. The current approach of assessing AI services in a black box setting, where the consumer does not have access to the AI's source code or training data, limits the transparency, explainability, safety, and adoption of AI. The consumer has to rely on the AI developer's documentation and trust that the system has been built as stated. Further, if the AI consumer reuses the service to build other services which they sell to their customers, the consumer is at the risk of the service providers (both data and model providers). In this talk, I will cover chatbots (collaborative assistants), the problem of trust in this context and how one may make them more trustable. I will cover software testing, AI robustness, randomized control trial and the idea of rating AI based on their behavior. I will highlight some of our work, present key results and discuss ongoing research
Biplav Srivastava is a Professor of Computer Science at the AI Institute at the University of South Carolina. Previously, he was at IBM for nearly two decades in the roles of a Research Scientist, Distinguished Data Scientist and Master Inventor. Biplav is an ACM Distinguished Scientist, AAAI Senior Member, IEEE Senior Member and AAAS Leshner Fellow for Public Engagement on AI (2020-2021). His focus is on promoting goal-oriented, ethical, human-machine collaboration via natural interfaces using domain and user models, learning and planning. Biplav has been working in AI trust for the last 3 years pursuing ideas in AI testing, rating, randomized control and adversarial learning. He applies these techniques in areas of social as well as commercial relevance with particular attention to issues of developing countries (e.g., transportation, water, health and governance). Biplav's work has led to many science firsts and high-impact commercial innovations ($B+), 190+ papers and 60+ US patents issued, and awards for papers, demos and hacks. He has interacted with commercial customers, universities and governments, been on multilateral bodies, and assisted business leaders on technical issues.
Host: Manas Gaur, manas@umbc.edu