
Are Emergent Behaviors in LLMs an Illusion? with Sanmi Koyejo - #671
The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)
Partial Credit Shortage in Metrics
The focus is on applying abstracted academic metrics to real-world use cases and business scenarios. The key observation is the difference between metrics that correlate with emergent behavior and metrics that don't. This difference is referred to as 'sharpness' or 'harshness', meaning the metric doesn't give partial credit and follows an all-or-nothing credit assignment approach.
00:00
Transcript
Play full episode
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.