There are a variety of reasons why I think statistics and statisticians/psychometricians will remain useful, even as technology advances. Not only are statistical results more parsimonious and easier to wrap one's head around than machine learning results, but many of the decisions statisticians and psychometricians make are still somewhat subjective.
If I give a dataset to 12 statisticians, along with the questions I want to answer and hypotheses I want to test, I'm likely to get 12 different sets of analyses. Some of the decisions we make are more art than science, or even simply a matter of preference. Some statisticians put more trust in the robustness of statistical results to deviations from the key assumptions. There are a variety of explanations as to why the 12 statisticians would come back with 12 different analysis approaches, though some approaches may be more justifiable than others.
Enjoy the video and let me know your thoughts and reactions!