The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)
Are Emergent Behaviors in LLMs an Illusion? with Sanmi Koyejo - #671
Sanmi Koyejo joins the TWIML AI Podcast to discuss his research on emergent abilities in large language models, questioning if they are a mirage caused by evaluation metrics. He also covers his work on "DecodingTrust," assessing LLM trustworthiness in areas like toxicity and…