I dunno where you’re getting your numbers for the calculation you’re doing, that would probably be something good to include

Absolutely fair. In fact, since yesterday night I tried to do more variations of my calculations for different types and now I, at least, heavily doubt my own conclusions and, at most, disagree with my conclusions entirely. Especially taking into consideration some of the aspects that people like you mentioned. But here’s my approach. When I was saying “efficiency” I specifically meant percentage of useful work done relative to weight (I know it’s not a be-all, end-all metric, but that’s what I chose). For example, a 2 ton car carrying one 70kg person has efficiency rating of:
(70 kg / 2,070 kg) * 100 = ~3.38%
Then I did these calculations for 1, 3 and 5 passengers, which makes 3.38%, 9.50% and 14.89% respectively. Then I took a random bus (curb weight = 12000 kg, max capacity 40 passengers), and repeated the calculations for 1 passenger, half and full occupancies. That came out to 0.58%, 10.45% and 18.92% respectively. Seeing that at half occupancy, cars are basically as efficient as busses, and knowing that on average busses are not even half-loaded (around 40%) I concluded that cars are in fact very efficient, given that you use them properly.
But of course that isn’t the whole picture. Some issues with my numbers that I found:

• average car is much lighter than 2000 kg (regular sedans are about 1500 kg, and a typical European car is around 1100 kg)
• busses at that weigh actually have much more occupancy
• it’s unfair to compare half occupancy, because statistically cars have 1.2 passengers on average.

Taking these things into account, I (mostly Claude) made this calculator. It even has rough numbers for certain cars and bus types. Using that calculator I can clearly see, that busses win, even when lighter cars are reasonably utilized.

talking about autonomous vehicles

This was a sci-fi hypothetical anyway, even optimistically, I don’t think we will have truly self-driving cars for another 5-10 years. I agree with a lot of what you said, but we can’t really apply today’s approach to that future sci-fi scenario. For example, if we have a swarm of hive-mind public cars that anticipate each other’s moves, then those potentially could be way more efficient than route based traffic. But I don’t wanna fixate on the hypotheticals.

Regarding your last paragraph, I don’t own a car, mostly walk and use trams. But I live in Europe, and here in Warsaw, we don’t really have a car problem. Sure, the work commute hours are a bit loaded, but otherwise, public transport is really good and a car is barely needed. So yes, until further notice, avoid cars if possible.

Thank you for such a lengthy and detailed response!