Studying Machine Intelligence with Been Kim - #571
The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)
00:00
Decoding Interpretability in Machine Learning
In this chapter, the speaker explores the vital topic of interpretability within machine learning, emphasizing the balance between engineering and scientific inquiry. They discuss various interpretability methods, their limitations, and propose a hypothesis-driven approach to understanding neural networks. By examining the connection between Gestalt psychology and machine learning, the chapter seeks to highlight new methodologies for revealing biases and enhancing model transparency.
Transcript
Play full episode
