Extracting Concepts from GPT-4

Using new techniques for scaling sparse autoencoders, we automatically identified 16 million patterns in GPT-4's computations.

Source: OpenAI — Published June 6, 2024 at 00:00 UTC — Category: Models