The Comprehension Normalization Method For AI Networks
Complex systems are difficult to understand because their behavior emerges from the relationships among their many parts. The Comprehension Normalization Method (CNM) is a patented method that exposes otherwise hidden inner relationships that drive the behavior of the system. Originally developed for biological networks, CNM is now being applied to artificial intelligence, where understanding changes to a model’s internal organization can help reveal vulnerabilities in the model’s functional capabilities.
The Comprehension Normalization Method for AI
CNM enters AI as a mechanistic interpretability tool - meaning, within the black box of how AI works, CNM can find causal structures that drive AI decision-making. Unlike traditional AI evaluations that measure the behavior of the results (like checking temperature or a rash) CNM measures changes to the model's internal structures (like conducting an MRI would).
Legacy Comprehension Normalization Method
CNM was born as a network analytics tool for systems biology and language. It was developed to help find causal mechanisms of disease and was used with both genomic data and imaging data for diseases like Alzheimer's, schizophrenia, and obesity. Networks exist across fields with preserved mathematical principles. Today the network analytics method is being used in the networks of AI.
Awards & Accreditations
ELabNYC
First Growth Venture Capital (VentureCrush)
