When the Ars Electronica festival took a deep dive into examining what “truth” means in today’s world, Johannes Kepler Universität Linz (JKU) invited broad participation from CCC, which included a workshop in LLMs:
Large Language Models (LLMs) and Us
Led by CCC graduate student Belén Saldías, Research Scientist Jad Kabbara, and Professor Deb Roy
This workshop gave participants the opportunity to explore and challenge the capabilities of Large Language Models (LLMs). Addressing questions such as: What can LLMs do and how can they mislead us? How would you tell an AI to behave like you do? How would you probe the AI’s values to see if its values sync with yours? Drawing from CCC’s current research, we showed ways in which we use LLMs to bring people together in constructive ways.
Media coverage
As unbiased as they may seem, AI chatbots can harbor outdated values. In her TEDxBentleyU talk, Belén Saldías explores this issue and reveals what's happening behind the scenes of the AI technology we see so much of today. In her talk, Belén will walk us through some stunning AI capabilities and show the impact of hidden challenges.
Experience
Conferences & Events
News and service
Main Printed Media
Other invited lectures & presentations
Working papers
Saldias, B.*, Gillani, N.*, Priscilla-Makini, S.*, Hughes, M.*, Roy, D. Exploring a human-bot blended storytelling model for youth mentorship. Working paper.
Saldias, B., Barassi, V. Challenges and Opportunities in Designing Child-Centered Content Exposure and Moderation. Working paper.
Saldias, B., TBA, Roy, D. Rationalizing Natural Language Content Moderation. Working paper.
2022
5. Saldias, B., Foster, G., Freitag, M., & Tan, Q. (May, 2022) Toward More Effective Human Evaluation for Machine Translation. Proceedings of the 2022 ACL Workshop on Human Evaluation of NLP Systems (HumEval). ACL.
paper, video, poster, slides, citation
2020
4. Saldias, B., & Roy, D. (July, 2020) Exploring aspects of similarity between spoken personal narratives by disentangling them into narrative clause types. Proceedings of the 2020 ACL Workshop on Narrative Understanding, Storylines, and Events (NUSE). ACL.
Work also presented at WiML @ NeurIPS19, WiML @ ICML20, CODI @ EMNLP20.
paper, video, short video, slides, poster, repo, @medialab, citation.
2019
3. Bhargava, R.*, Chung, A.*, Gaikwad, N.*, Hope, A.*, Jen, D.*, Rubinovitz, J.*, Saldias, B.*, & Zuckerman, E.* (2019, November). Gobo: A System for Exploring User Control of Invisible Algorithms in Social Media. Proceedings of the 2019 ACM Conference on Computer Supported Cooperative Work and Social Computing (CSCW). ACM. *All authors contributed equally to this research.
paper, poster, @medialab, citation.
2. Saldias, B., & Picard, R. (2019, September). Tweet Moodifier: Towards Giving Emotional Awareness to Twitter Users. Proceedings of the 2019 International Conference on Affective Computing and Intelligent Interaction (ACII). IEEE.
paper, slides, poster, news, @medialab, citation.
1. Saldias, B., Protopapas, P., & Pichara, K. (2019, May). A Full Probabilistic Model for Yes/No Type Crowdsourcing in Multi-Class Classification. Proceedings of the 2019 SIAM International Conference on Data Mining (SDM) (pp. 756-764). SIAM.