In this episode of A Tale of Two AIs, host Razi Raziuddin sits down with Eric Siegel, CEO of Gooder AI and founder of Machine Learning Week, to explore why so many predictive AI projects never make it to production—and what needs to change.

Eric has spent decades helping enterprises move machine learning from theory to measurable business impact. In this conversation, he shares why technical performance isn’t enough, why deployment must be “sold” internally, and how predictive AI is poised for a renaissance in the age of generative systems.

They discuss:

• Why ML evaluation (AUC, lift, ROC) isn’t the same as ML valuation

• The myth that a “good model” will automatically get deployed

• Whether GenAI has actually fixed anything fundamental in data science

• Why predictive AI may be entering a renaissance

• And how predictive models can serve as a reliability layer for GenAI agents
  

The core idea: enterprises don’t run on certainty — they run on probabilities, and probabilities aren’t going anywhere.