Responsible AI: Debugging AI for errors, fairness and explainability

I had the opportunity to deliver a webinar about responsible AI for the ODSC audience: “Debugging AI for Errors, Fairness and Explainability”. First, we looked at the current state of machine learning where AI innovations are occurring at a rapid rate; companies are trying to adopt AI in their business processes and solutions; society expectations for AI are growing, and governments are starting to regulate AI.

Then we explored the role Microsoft is playing in making responsible AI principles and practice tools for the data science community and organizations to use as part of their machine learning life cycles. We reviewed some of the open-source tooling advancements researchers, academics and Microsoft have developed for machine learning professionals to better understand model behavior, discover and mitigate undesirable issues from AI models.

Finally, we did a deep-dive into Azure Machine Learning’s Responsible AI dashboard components that helps data scientist and AI developers debug and mitigate model issues by using its error analysis, data explorer, fairness assessment, model interpretability, counterfactuals/what-Ifs and causal analysis components. And how this holistic tool helps data scientists and AI development teams get started integrating responsible AI analysis into their everyday development practices.

To see other great ODSC on-demand webinars or upcoming events, visit: https://app.aiplus.training/courses/Responsible-AI-Debugging-AI-models-for-errors-fairness-and-explainability.