... But actually to continue the work that we're already doing to implement self-driving cars. I mean, it's not just drugs, it's also tiredness, distractions, and simple mistakes. We can make cars safer by removing human drivers and replacing them with extensively tested machines. It will take a bit more time, but we'll get there.
While I certainly recognize potential benefits of autonomous driving (and automation, in general), relevant global or nationwide deployments are still a very long way to go; it's not "a bit more time", but most likely decades. Imagine how many lives could be saved with simpler sensor-based solutions (which, by the way, could very much help with the problems of "tiredness, distractions, and simple mistakes") in the meantime ...
With respect to self-driving, search YouTube for FSD beta. You can see what Tesla has been up to. It's not perfect but it keeps getting better. It's not going to take several decades.
I'm well aware of the progress in autonomous driving space (including Tesla's - though, I would argue that Tesla is not the leader in this space; I think that it's Waymo). However, more generally, my point was not about limited fleet use cases, but rather global-scale deployments and massive adoption. That, I believe, is certainly not going to happen soon.
The solution to this problem seems quite trivial to me: to equip all cars with alcohol (and, perhaps, some drug) sensors [not breathalyzers, but rather some more sophisticated remote-sensing devices], strategically located throughout interior to prevent cheating / gaming the system, and integrate them with ignition and engine. As long as the system determines that a driver is drunk or under influence of drugs, it prevents the car from starting or (in case someone starting drinking or using drugs after stopping a car, but leaving it idling) moving.