MENLO PARK — Facebook announced a collaboration between Captum and Fiddler Labs, aimed at advancing the field of Explainable AI (XAI). With AI models being increasingly applied in critical areas such as healthcare, hiring, and finance, understanding the reasoning behind their decisions has become essential. The partnership between Captum, part of the PyTorch ecosystem, and Fiddler’s Explainable AI platform seeks to tackle this challenge by providing new tools to interpret the inner workings of machine learning algorithms.

As AI systems become more complex, it is becoming harder for developers and engineers to fully grasp how these models make decisions, raising concerns about trust and transparency. Captum, introduced with PyTorch 1.3, offers cutting-edge interpretability algorithms designed to provide greater insight into ML models. These tools help engineers build AI systems that are more understandable, reliable, and predictable.

Fiddler Labs’ platform enhances the ability of engineers to diagnose AI models by delivering insights into their decision-making processes, helping mitigate issues like bias, mispredictions, and fairness. By integrating Fiddler’s platform with Captum, developers can significantly improve their understanding of AI behavior. Fiddler can analyze models, identify data drift, detect outliers, and monitor live traffic, providing comprehensive oversight over the models in production.

This partnership aims to promote the widespread adoption of Explainable AI across various industries by making AI systems more transparent and interpretable. By offering a more robust set of tools, the collaboration is expected to foster both research and practical applications of XAI.

For more details on the collaboration and how it benefits AI explainability, read about the Fiddler and Captum partnership.