In a sense, we have completely replaced a lot of scientific processes with the power of statistical methodology and essentially outsourced part of our scientific thinking and practice to statistical inference. And now we're confronted with the risk of the same thing happening with AI.