There are a few counterarguments. One is that if a fit driver (say she also works out in a gym) switches to biking, that will save a lot of energy, because there won't be the same longevity effect. So his argument only applies to people who go from no exercise, to regular bicycle commuting. Ulrich also didn't take into account that bikers may own fewer cars (so less energy for manufacturing) and require less roadspace to be built for them (so less construction energy spent on their behalf). When I emailed him about those last two points he sent a very polite reply trying to estimate what the effects would be.