How often do we still use old statistical measures nowadays?

How useful are statistical measures like AIC, BIC, adjusted R-squared etc, in picking the most efficient ML models in the present days? Or were they completely overtaken by the Cross Validation methods?