Introduction to Causal AI

In the evolving landscape of AI, especially with the capabilities of Generative AI, there is still an underlying shift from correlation-based approaches to causal inference models marks an inflection point. Filling in the large gap between correlation-based methods common in machine learning and physics-based approaches, Causal AI introduces a methodology that relies upon two key components: (1) algorithmic knowledge discovery to uncover causal relationships and (2) building the causal model. The presentation will provide an overview of key concepts within the causal inference domain and discuss how the role of experimentation plays a larger role in determining the performance levels of causal models. Finally, the power of causal modeling will be demonstrated through two real-life use cases, one on manufacturing root cause analysis and another on the topic of process optimization.

Objective

• Introduce a new area within AI, called Causal AI, that finally brings causality and therefore explainability into the modeling process
• Demonstrate potential use cases for Causal AI where physics-based approaches may not be practical

S​peakers

Mahmood Tabaddor PhD

Director for AI Solutions for Industry X, Accenture

Dr. Mahmood Tabaddor has been involving in modeling and simulation in a variety of industries covering engineering and manufacturing for over 25 years. He is currently responsible for AI and Gen AI applications within engineering and manufacturing at Accenture. He is a member of the America Steering Committee for NAFEMS supporting AI initiatives.