The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) cover image

Are Emergent Behaviors in LLMs an Illusion? with Sanmi Koyejo - #671

The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)

NOTE

Models as Systems with Surrounding Components

Machine learning models, particularly LLMs, are not just standalone entities but are complex systems that involve additional components like reachable augmented generation and various tool use. These surrounding components play a significant role in influencing the behavior of the models during tests. Evaluating these tools separately can offer insights and potential for mixing and matching in different use cases, such as evaluating a rag tool independently based on existing evaluations of embedding quality which could correlate with its practical performance.

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.
App store bannerPlay store banner