Events Calendar

NL Seminar-Contextualization for Human-AI Interactions

Speaker: Justin Cho, USC/ISI, USC/ISI

Talk Title: Contextualization for Human-AI Interactions

Abstract: Meeting hosts only admit on-line guests that they know to the Zoom meeting. Hence, you’re highly encouraged to use your USC account to sign into Zoom. If you’re an outside visitor, please inform us at (nlg-seminar-host(at)isi.edu) to make us aware of your attendance so we can admit you. Specify if you will attend remotely or in person at least one business day prior to the event. Provide your: full name, job title and professional affiliation and arrive at least 10 minutes before the seminar begins. https://usc.zoom.us/j/94650895633?pwd=FiYw69aOiNZ9KkmIdO0JpaiG59QctB.1 Meeting ID: 946 5089 5633 Passcode: 282370 Recent developments in AI are nothing short of amazing, but goal posts move, and we quickly discover that AI remains insufficient for fulfilling many real world tasks. The shortcoming can be largely attributed to a lack of contextual understanding on the AI’s part. This is not surprising given that the dominant training and evaluation paradigm for AI models  prioritizes scale and rapid progress. As a result, we’ve developed a bias for textual data, instruction data with transactional interactions, aggregated and simple preference data, and evaluation tasks that can be easily verified. In this talk, I present research that explores the opposite side of this bias for enabling more useful and contextualized human-AI interactions. Specifically, I introduce three research directions to demonstrate that utility is a function of context and that teaching an AI model to understand the specific context of its interaction with humans is crucial for successful outcomes. (1) How an interaction takes places: human-AI interactions will expand beyond communicating with a textual interface, such as speech. I present how we can adapt language models for speech-based interactions with literature-guided prompts and speech-based preference data. (2) Who the user is: every user is different and sparingly share their data. I demonstrate how we can align language models to individual users without any fine-tuning and using small amount of per-user data. (3) What is the goal: complex tasks require evaluations that take a more holistic approach that goes beyond the immediate model response. Through a case study of using language models as content moderators, I argue that evaluations for complex tasks need to account for each group of stakeholders as the perceived effectiveness of language models vary significantly among them. 

Biography: Justin Cho is a PhD Candidate at University of Southern California advised by Jonathan May. His research has centered around refining the context in which language models are involved in human-AI interactions, such as enhancing dialogue models with grounding techniques and understanding of the interaction modality, personalizing language model outputs, and applying them for social good. He has previously led USC’s team into the semifinals for the fourth Alexa Prize Socialbot Grand Challenge, co-organized the Conversational AI workshop at ICML 2024, and interned at Meta’s Conversational AI team, Amazon Alexa, and Amazon AGI.  Subscribe here to learn more about upcoming seminars: https://www.isi.edu/events/  For more information on the NL Seminar series and upcoming talks, please visit: https://www.isi.edu/research-groups-nlg/nlg-seminars/

Host: Jonathan May and Katy Felkner

More Info: https://www.isi.edu/events/5488/nl-seminar-contextualization-for-human-ai-interactions/

Webcast: https://www.youtube.com/watch?v=N5CpzeC1nEM

Location: Information Science Institute (ISI) - Virtual Only

WebCast Link: https://www.youtube.com/watch?v=N5CpzeC1nEM

Audiences: Everyone Is Invited

Contact: Pete Zamar

Event Link: https://www.isi.edu/events/5488/nl-seminar-contextualization-for-human-ai-interactions/

This event is open to all eligible individuals. USC Viterbi operates all of its activities consistent with the University's Notice of Non-Discrimination. Eligibility is not determined based on race, sex, ethnicity, sexual orientation, or any other prohibited factor.