Statisticians are cautioning fellow scientists not to make assumptions about the accuracy, uncertainty or reproducibility of scientific discoveries made with today's machine learning models.