Let me take a stab at it as well, because I think it could be simpler.
Let's assume the driver drives the exact same model as the autonomous one, except he saves $10k upfront on the sensors. So his costs are 12k/year. A decent alternate job would pay $10/hour, and a person could probably work 9 hours a day * 6 days a week * 50 weeks, which comes to $27k. So for the taxi driving gig to be worthwhile, the driver needs to make somewhere in the region of $35-40k ($27k + $12k).
As mentioned earlier, the costs of the self-driving car is $14k/year. So that's a difference of 2.5x-3x. Still significant enough that the a taxi company that offers rides at 30-40% of the cost of its competitors will just win.
On factor I have not seen fully addressed is the ability of driverless cars to run near 24/7. Given than demand is variable and human drivers could share cars like taxis what is the worth in practice?
Right now an Uber driver who is probably online 60 hours a week, needs to support a car priced around 25k (amortized over 5 years), gas (which is more expensive than electricity), maintenance and 50k-70k/yr driver income. Assuming lower end we are looking at 5k car + 10k fuel + 5k maintenance + 50k driver income. So cost to support a driving car is 70k/yr for lower hours of operation.
So we are looking at a 5x cost and lesser hours of operation with human driving vs autonomous driving.
These are off the napkin numbers so please don't take them literally. It is meant to highlight that driver income is the biggest factor.
Whoever has self driving tech, they can start undercutting Uber with the most profitable markets (NYC, SF, Boston, LA etc) and expand from there.
I feel that we are at least 5-8 years away from main stream self driving. But I could be wrong.