USC Information Sciences Institute Courses

AI for Fostering Constructive Online Conversations

Explore AI's role in promoting constructive online discussions about race, addressing content moderation biases, and developing NLP systems for social good across various domains.

• YouTube
• 1 hour 1 minute
• On-Demand
• Free Video

Causal Inference to Inform Curation Practices in Online Platforms

Explore causal inference methods for improving content curation on digital platforms. Examine case studies on Reddit, discuss implications for policy, and consider future research directions in the era of generative AI.

• YouTube
• 1 hour
• On-Demand
• Free Video

Event Extraction for Epidemic Prediction

Explore event extraction techniques for early epidemic prediction using social media data. Learn about epidemic event ontology, cross-lingual models, and innovative label projection methods.

• YouTube
• 1 hour 2 minutes
• On-Demand
• Free Video

Understanding LLMs through Generative Behavior, Successes and Shortcomings

Explore LLM generative behavior, successes, and shortcomings. Gain insights into model design, inference algorithms, and performance across diverse tasks requiring varied knowledge and capabilities.

• YouTube
• 1 hour 2 minutes
• On-Demand
• Free Video

Staring into the Abyss and Eating Glass - From Academia to Startup

Explore the journey from academia to entrepreneurship with insights on building a sports analytics startup. Gain valuable perspectives on challenges, similarities, and differences between academic and business worlds.

• YouTube
• 1 hour 7 minutes
• On-Demand
• Free Video

How to Steal ChatGPT's Embedding Size and Other Low-rank Logit Tricks

Uncover hidden information about API-protected LLMs through low-rank logit analysis. Learn to estimate embedding sizes, audit model updates, and identify source models efficiently.

• YouTube
• 48 minutes
• On-Demand
• Free Video

AI in Structural Biology: Roots of Success

Explore the history and impact of AI in protein structure prediction, focusing on AlphaFold 2's breakthrough and its implications for structural and cell biology research.

• YouTube
• 57 minutes
• On-Demand
• Free Video

DeLLMa: A Framework for Decision Making Under Uncertainty with Large Language Models

Enhance decision-making accuracy in uncertain environments using DeLLMa, a multi-step scaffolding framework based on decision and utility theory principles. Improve LLM performance by up to 40% in complex scenarios.

• YouTube
• 1 hour 5 minutes
• On-Demand
• Free Video

Multimodal Machine Intelligence and Health

Explore multimodal machine intelligence applications in medical sciences and health research, addressing challenges in creating trustworthy, inclusive AI for clinical and quality of life domains.

• YouTube
• 1 hour 16 minutes
• On-Demand
• Free Video

Informative Example Selection for In-Context Learning

Explore efficient approaches for selecting informative examples in In-Context Learning, enhancing large language model performance without additional training.

• YouTube
• 59 minutes
• On-Demand
• Free Video

The Data Provenance Initiative: A Large Scale Audit of Dataset Licensing and Attribution in AI

Explore a comprehensive audit of 1800+ text datasets, examining legal and ethical risks in AI data practices. Gain insights into dataset lineage, licensing, and the divide between open and closed datasets.

• YouTube
• 56 minutes
• On-Demand
• Free Video

Do Androids Know They're Only Dreaming of Electric Sheep?

Explore transformer language models' internal representations to detect and analyze hallucinatory behavior in AI-generated content, with insights on task-dependent patterns and evaluation methods.

• YouTube
• 59 minutes
• On-Demand
• Free Video

A Gateway to Trustworthy AI: Using Visual Analytics to Unmask Coincidental Correlations

Explore visual analytics techniques to identify and address spurious correlations in AI systems, enhancing trustworthiness and reliability in machine learning and data-driven decision-making.

• YouTube
• 1 hour 1 minute
• On-Demand
• Free Video

Self-Play Fine-Tuning Converts Weak Language Models to Strong Language Models

Explore a novel fine-tuning method for Large Language Models that improves performance without additional human-annotated data, using self-play mechanisms and theoretical guarantees.

• YouTube
• 1 hour 3 minutes
• On-Demand
• Free Video

Red Teaming Language Model Detectors with Language Models

Explore strategies to compromise LLM detectors using synonym replacement and instructional prompts, revealing vulnerabilities in text detection systems and emphasizing the need for robust safeguards.

• YouTube
• 48 minutes
• On-Demand
• Free Video