I think the per se limits have to be chosen carefully. I know, for example, that a heavy user could have a low level of THC, a nanogram per millilitre, that's detectable for a long time after they last use. You certainly don't want a per se level that's going to be catching everybody up in the net.
It's a difficult problem because on the one hand, you don't want the levels so low that you're detecting chronic users who might not be impaired. On the other hand, unlike alcohol, the THC levels can drop off quite quickly with time. It's hard to go from a level that may be four hours after a crash and to estimate what it would have been at the time of the crash. You can't back extrapolate the way we've learned to do with alcohol. If you have a level that's too high, you might have someone who was quite impaired at the time of a crash, and two hours later, by the time you get a chance to get blood and measure the level, their level has dropped below that. It's sort of a balance.
Other countries have used a slightly higher limit. Norway uses three nanograms per millilitre. I still think that, on balance, two is a reasonable level. I should put in a disclaimer that I don't have a lot of experience with seeing people who have used cannabis, measuring their levels, and studying their impairment. That's not my line of research. It's more at a population level, looking at the risk of crashing for most drivers.
I have one additional comment. To use the analogy with alcohol, there are people who are less impaired at a high alcohol level than others, yet we've accepted that, on balance, most people above .08 are not fit to drive, and it makes sense to have a per se level for alcohol as well.