Flagged for Being Queer
USC ISI researchers uncover bias in content moderation, specifically, how LLMs misinterpret gender-nonconforming speech.
The Surprising Ineffectiveness of Countering Hate Speech on Reddit
A USC ISI study reveals that counterspeech fails to change user behavior in hate-based subreddits – except when it is hostile
USC ISI Shines at LA Tech Week 2024
From AI Titans to cutting-edge labs, ISI researchers showcased future technology to the LA tech community.
UPCOMING EVENTS
Nov
15
AI Seminar – Bigger Isn’t Always Better: Why Larger Language Models Struggle with Subjective Reasoning
Yiorgos (Georgios) Chochlakis, USC
-
Nov
22
Do We Need Large Language Models for Time Series?
Vinayak Gupta, Lawrence Livermore National Laboratory
-
Jan
17
Towards Socially Aware and Safe AI Agents
Xuhui Zhou, Carnegie Mellon University
-