Hacker News Reader

Comment by paulcole | Hacker News Reader

> First, they were gonna make the jump to be an autonomous robotaxi service

Anyone who fell for this is a fool. Nobody could answer the question, “How is self-driving cheaper than exploiting human drivers like they are now?”

View on HN

The financial benefit seems pretty clear, conditional on self-driving tech working. The capital cost of the hardware would be amortized. Let’s say a driver costs $10/hour. If the self-driving car is active for 16 hours a day it saves you $58k a year in driver wages. Then over more years that number multiplies. It seems totally realistic for self driving tech to cost five figures per vehicle. With mass production the cost of hardware would go down like it always does. I also think it makes a lot of less driver-time-inefficient routes more viable, like driving someone from the city deep into the suburbs with little chance of a return fare. This grows the addressable market.

There are also ethical, legal, and P.R. disadvantages to exploiting human drivers that seem to be getting more significant over time. Personally I’m not super optimistic about the near-term technical viability of self driving cars but it does seem to make Uber’s prospects look better to investors given some possible paths the future might take.

So instead of paying a bunch of people $10/hour, Uber is going to pony up front for a huge number of vehicles and the staff it takes to maintain them? Or are scads of private owners going to hand over their car to Uber to run as AI-taxis for $10/hr?

$10/hour is a rough estimate for the after expenses earnings for Uber drivers. The gross Uber pays is more like $16/hour so that’s the number to compare if you are going to look at Uber’s all-in cost to run a fleet. Using the 16 hours/day estimate it comes out to $93k a year of driver payments saved, which seems like it should easily pay for the purchase and maintenance of a car (especially if the car lasts multiple years). Uber has access to capital so they can handle the “up front” nature of the expense. Paying people to drive a car around all day is expensive and if Uber could avoid that they would have a better business.

Yeah, 16 hours of driving a day, 7 days a week, 365 days a year for multiple years does seem reasonable for a car. Feeling like a fool for doubting Uber’s self-driving end game.

I never understand why people expect there won't be self-driving cars for sale to the public which voids the need for self-driving car taxis? I would prefer buying such as car than keep taking taxis. In the end I think it would be much more convenient if you could call via app your own car in a similar way :)

That's the kind of future I am imagining

> Nobody could answer the question, “How is self-driving cheaper than exploiting human drivers like they are now?”

Huh?

There are SEVERAL obvious advantages.

* AI drivers don't get tired - they can drive 24 hrs/day if there's demand. This lowers cost because the (fixed) cost of the vehicle is amortized over more rides.

* They don't get a cut. See above.

* AI doesn't incur hr costs and can't sue you or try to unionize

Who owns the AI-driven car? Who maintains it?

Either Uber is spending a shitload of money to build, maintain, and replace their own fleet or they‘ll be leasing cars from private car owners for about what they paid human drivers in the first place.

Consider a very simple city model which requires 10,000 cars to meet demand during rush hour (3 morning hours, 3 evening hours) and 4,000 cars to meet demand the rest of the time.

A self-driving model requires the up-front purchase of 10,000 cars, 6,000 of which sit idle 18 hours per day, earning nothing. Meanwhile Uber and Lyft have zero up-front costs, and only pay drivers when the company itself is earning revenue (and presumably, a marginal profit).

And, furthermore, assuming the availability of self-driving at the highest degree of difficulty--namely relatively dense cities.

Reply on news.ycombinator.com