Explore the challenges and solutions of detecting harmful content at scale in this 51-minute podcast episode featuring Matar Haller, VP of Data & AI at ActiveFence. Dive into the complexities of online platform abuse, including brand and legal risks, user experience impacts, and the blurred line between online and offline harm. Learn about AI-driven content moderation, optimizing speed and accuracy, cultural sensitivity in AI training, and continuous adaptation to evolving threats. Discover strategies for testing and deploying machine learning models, monitoring hallucinations in transformer models, and balancing moderation efforts. Gain insights into improving production code quality and addressing AI detection concerns in the ever-changing landscape of online content moderation.
AI for Good - Detecting Harmful Content at Scale
Overview
Syllabus
[] Matar's preferred coffee
[] Takeaways
[] The talk that stood out
[] Online hate speech challenges
[] Evaluate harmful media API
[] Content moderation: AI models
[] Optimizing speed and accuracy
[] Cultural reference AI training
[] Functional Tests
[] Continuous adaptation of AI
[] AI detection concerns
[] Fine-Tuned vs Off-the-Shelf
[] Monitoring Transformer Model Hallucinations
[] Auditing process ensures accuracy
[] Testing strategies for ML
[] Modeling hate speech deployment
[] Improving production code quality
[] Finding balance in Moderation
[] Model's expertise: Cultural Sensitivity
[] Wrap up
Taught by
MLOps.community
