No, they'd actually be substantially worse. Personal cars are poorly utilized, but they at least spend all of their time carrying at least one person (the driver), and stay off the roads when they aren't in use. A taxi will, necessarily, spend less than 100% of its time carrying a passenger; if it's being used primarily by commuters, it's likely to spend at least 50% of its time with no passengers at all (because it's deadheading back into a city to take workers home, or vice versa).
Having everyone drive cars everywhere isn't sustainable, but making those cars into taxis isn't the solution either.
One option is to mix public transport with driverless taxis - consolidate public transportation into fewer, bigger routes (knocking down costs), then use driverless taxis for short trips to the station. A little inconvenient to add 1-2 extra transfers to your trip, though.
There's also the opportunity to automate carpooling, but that's not quite the same - it's just moving a little further on the flexibility <-> efficiency scale of transportation (where individual vehicles are furthest left and trains are furthest right). Even if you're just sharing a taxi with 2-3 other commuters, you're still sacrificing the convenience of leaving at the time of your choice and having privacy during the trip - though of course for a big price discount.
I am curious how much cheaper the robotaxis will actually be for the user (right now they’re all at a loss)
Assuming that a cab can make the same income in the dead of the night as it does during peak periods is a rather bold assumption.
How much revenue a cab, whether human-operated or autonomous, generates at 1am vs 1pm is entirely irrelevant.
And you're assuming that all cabs are operated 24 hours a day, in three shifts. They are not.