AI Dashboard Design
Martin Wattenberg discusses the concept of large language models as aggregating statistics versus building an internal model of the world. He explores the idea of probing the model's activations during inference to understand how it encodes aspects of the world. Wattenberg also shares research on using the game Othello to experiment with these concepts.In this clip
From this podcast
The Gradient
Martin Wattenberg: ML Visualization and Interpretability
Related Questions