
Is It Time to Rethink LLM Pre-Training? with Aditi Raghunathan - #747
The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)
00:00
Exploring Limitations of Language Models
This chapter discusses the limitations of current large language models (LLMs) and reflects on insights from recent research. It emphasizes the contrast between the impressive capabilities of these models and their failures, advocating for improved understanding and safety mechanisms. The conversation also explores the dynamics of fine-tuning, overtraining, and the challenges of unlearning harmful information to enhance the reliability of AI systems.
Transcript
Play full episode