This startup’s new mechanistic interpretability tool lets you debug LLMs

Goodfire, a San Francisco-based startup, has launched Silico, a tool that allows researchers to adjust AI model parameters during training. This tool is claimed to be the first of its kind, enabling developers to debug throughout the entire development process. Goodfire aims to make AI model building more scientific, addressing the gap between model understanding and deployment. CEO Eric Ho emphasizes the need for precision engineering in model training, advocating for mechanistic interpretability to understand model behavior.