Interpretable and Auditable AI Systems
We are building a new class of interpretable AI systems and foundation models that humans can reliably debug, trust, and understand.
Understand which part of the prompt is responsible for the output.
Understand what factors are responsible for the output.
Understand which training data is responsible for the output.
The Problem
Current AI systems and foundation models:
The Problem
produce explanations and justifications that are unreliable and unrelated to their output.
cannot be reliably debugged and fixed.
are difficult to control and align with current approaches.
Our Solution
LLMs & Foundation models engineered to be interpretable
The Problem
Produces human-understandable factors for any output it generates
Produces reliable context citations.
Specifies which training input data strongly influences the model’s generated output.
About Our Team
Learn moreon interpretable machine learning with PhDs from MIT, UMD, & MILA.
on interpretability published at top ML conferences
Developed the first