Hemelrijk's definition of random variables as completely unpredictable phenomena is contrasted with the practice of prediction in the face of uncertainty. General observations on the latter, and on the conflict between the conditional predictions supplied by scientific research and the unconditional predictions needed by policymakers.