He's pointing out that we have empirical evidence for the harm of certain substances and it's a different kind of harm than the largely unintended consequences of, say, cars, which provide us with far more direct advantages at reasonably low risk. The risk is also intentional and, to a degree, controllable with wild animals and intoxicated drivers being some of the major exceptions. (Inclement weather is at least partially controllable, as long as "don't drive" is a choice.)