This video explains one aspect of AI (generative AI mostly) that keeps me from worrying too much that the “robots will take our jobs”. The contrast the Dickerson draws is between knowing and understanding: AI will never understand anything.
In statistics, an AI tool could someday perform a thorough data analysis. Maybe even perform it well. But I can’t image that it will be able to fully understand the nuances of the application, the generation of the data (i.e., collection), and how those things fit in to the many choices of data analysis. It will miss things, just like amateur statisticians miss things (probably worse). Expertise will be needed to find those errors.