Uber CEO warns robotaxis can't find a fast route to commercial viability Uber CEO Dara Khosrowshahi is warning that it's not yet possible to commercialize autonomous vehicles due to high costs, seasonal demand fluctuations, and the need to prove "superhuman" safety. Khosrowshahi took the unusual step of including an update on his company's ambitions for autonomous vehicles (AV) in the prepared …
Uber doesn't own it's vehicles, or technically employ the drivers, so it's overheads in those areas is vastly reduced.
Running robotaxis it'll own the vehicle/device, be responsible for insurance, fuel/charge and obviously maintenance. So per hour running costs need to be low to offset the huge rise in costs.
“ All those rides and deliveries saw $12 billion arrive in Uber's coffers, a 20 percent year-over year increase. Net income was $6.9 billion, but $6.4 billion of that was a "benefit from a tax valuation release." Another $556 million came from "net unrealized gains related to the revaluation of Uber's equity investments."”
So looking at that - apart from tax and equity wheezes they basically made fuck all income despite all the effort and misery inflicted on their workers.
Sweet.
...does this whole scam unravel?
Autonomous vehicles are very much in the same line as all of the other tech industry's recent 'breakthrough ideas' - an attempt by corporations to insert themselves between the consumer and the thing they want to do, to minimal (or even negative) benefit to the consumer and usually an additional cost.
And it's so important to them to force themselves into that position that they will promise the earth just to get there. Remember the idea that Robotaxis would make their owners tens of thousands of dollars a year? Not only was that blatant economic illiteracy, but it was purely being dangled in front of people to get them to hand over their assets for the benefit of yet another corporation.
Additionally, autonomous vehicles are yet another example of "privatise the profits; socialise the costs". They only work - in the sense they only achieve the safety record they do - because actual human drivers are rather good at avoiding the dangerous situations created by the robots. Dodging them when they do stupid things, in other words. There is no evidence that autonomous vehicles can do the same, so they simply don't scale.
Imagine putting a toddler in a pedal car on a suburban street. They'll probably be OK, because drivers will spot and avoid them. But make every vehicle on that street a pedal car with a toddler in charge and they'll be bumping into each other all over the place. Same basic principle.
The current Uber model is definitely socialize the costs. Kid gets run over by a Uber driver while getting off a school bus. Uber drivers insurance and Uber driver pay. Not Ubers fault at all, no Metter how hard the drivers have to hustle. AV would be the AV companies responsibility.
That's been the fundamental logistical issue for all public transport providers ever since the existence of public transport: responding to traffic peaks in periods of high demand and maintaining a usefully responsive service in periods of low demand. Has Uber just noticed?
The same Dara Khosrowshahi is reportedly a fan of workers returning to the office; ironic, as the most significant driver of peak demand is the daily commute. Perhaps he hopes to make up on volume for what he is losing on margin.
While I don’t disagree with your observation, I don’t think AVs are a solution
For one, I would find it frankly terrifying to be charging along a motorway with nobody but a buggy computer in control. I would find it scary enough pootling around town. And I doubt I’m alone in that.
And I don’t think the whole issue of who is at fault when someone gets killed has been addressed. Not a technical issue, but not a resolved issue either.
If AV's were actually safer, Tesla would have been funding university researchers to write a bunch of papers proving this years ago.
The fact that that has not happened, that all we have had a vague assertions of "4x better" with hand wavy stuff, is prima facie proof that they are actually worse. Worse enough that you can't even get academic shills to prove them better for you.
One particular stretch or road locally has signs up saying so many casualties in 3 years (which 3 years not stated). What is the correct question to ask about that?
1. None (apparently the local authority's)
2. Who are the anomalously large population of selfish pricks?
3. What's the matter with the road?
In answer to 3 it could be 5 out of 7 junctions having bad sight-lines, 2 being virtually blind in one direction another which really needs to be a roundabout, probably double and was far from being improved by being improved a few decades ago and yet another being quite deceptive at night in that a badly placed reflective speed limit sign makes the main road appear to go on straight on while it actually bends to the near side.
The reason for these, by the way, are that in the age of the horse and cart and stage coach it didn't matter if the turnpike cut at odd angles across the old roads used by pedestrians and pack ponies.
But it's OK, if we just put up signs warning drivers that this road may be dangerous in unspecified places for unspecified reasons all will be well.
Just for added joy, a few miles away there's a left bend sign just at the point where the road bends to the right.
AV's really aren't going to fix issues like those.
There is nothing particularly dangerous about bad sight lines, as long as they are clearly bad and therefore engender caution. The real stinkers are bad sight lines which seem to be good ones. Tom Scott has a rather nice analysis of one here.
The corollary is that you can improve road safety by making sight lines which feel worse than they actually are. For example, by arranging things so that you think you can't see the road ahead, while in fact you can see anything more than four inches high on it. Tough shit for hedgehogs.
Note: Of course in a perfect world all drivers would drive cautiously and follow all road signs, but pragmatic deign takes account of human failings. And let's not forget that Teslas use visual input only for their "self driving" and so will have exactly the same problems.
"AV's really aren't going to fix issues like those."
They could, because they can maintain a speed appropriate to what's actually visible to them (via whatever sensors), in the way that humans simply don't... they just plow on because their time is more important than the safety of the others on the road.
Whilst AV's might in future be how we reduce Road deaths with technology, We are not there yet. As even Uber are admitting the superhuman safety record is not yet proven. In the mean time Automatic Emergency Breaking, Lane Keep Assist, and dynamic cruise control have advanced exponentially in recent years. Focusing on deploying the mature technologies like this, whilst being honest about their capabilities seems to be a much better bet for reducing fatalities in the near term. Longer term we already have electrified autonomous capable (many lines already have autonomous trains as there are less unknowns its already safe) intercity passenger/freight options in the form of Rail. Moving traffic from road to rail in Europe and America would make for fewer direct emissions (could help reducing global emissions as part of a low carbon electricity market), reduced traffic would result in safer roads. We are not investing effectively in additional rail capacity because the tech bros prefer to fritter away billions on automation and making futuristic technologies that will enrich the 1% if and when they mature to dealing with Rail regulations and unions, and governments are all claiming to be too poor to invest or are just inept (see HS2).
I mean we could get a significant improvement by handing over all motorway driving.
- Motorway driving is the simplest to automate
- It's also tiring, tedious and high speed
By handing that over to an AV, and then getting control back when you leave the motorway then the driver can be fresh for the non motorway leg of long journeys.
We don't need, and probably shouldn't aim for, a complete end to end "in one hit - nothing else will do" approach.
I find that motorway driving is only tedious late at night, although it's always tiring. Most of the time, there is plenty to think about because there are too many muppets with poor lane discipline (in *every* lane) and, much, much, worse, far too many idiots who try and intimidate the people who are clearly in the wrong (middle) lane by driving too close at high speeds. (Yes, they are annoying, but sitting on the bumper of someone who clearly isn't paying full attention to the road is not really a good idea, for obvious reasons, as is passing them on the wrong side or pulling in to their lane a car length or two in front of them.)
you beat me to it. Another advantage of AVs on motorways is that the cars could follow each-other at very short distance – like 2m – reducing aerodynamic drag and thus improving gasoline consumption by a lot (I have read that it could be up-to 40% !). The cars could communicate wirelessly (or even by optical means if they're in very close range) and the last car in the line would have the same information about dangers ahead as the first car : in case of emergency braking, all cars could brake at the same time !
Yes I know, we could make roads out of 2 steel rails and put the cars onto them ... the problem being to get on and off of this though.
> By handing that over to an AV, and then getting control back when you leave the motorway then the driver can be fresh for the non motorway leg of long journeys
There is no way that I would be fresh for the non motorway leg of long journeys - instead I would be a nervous wreak always on the edge of taking over. If I have to drive at all I will drive it all or share the driving with the wife which works best as she loves straight roads and I love the bendy up and down ones
a solution looking for a problem
sort-of. While I agree with that on general roads, there is one particular problem that this could may-be solve : long hours of driving on highways. Having to be concentrated for hours driving at 120km/h with other cars driving similar speeds, and sometimes overtaking trucks driving at 90km/h, while I could be playing with the kids, would be nice. Highways have the special case that everybody is driving in the same direction, there are no crossings, no pedestrians, no cyclists, no children playing ball on the side, no dogs ... and at the same time it is quite easy to imagine integrating some sort of guidance system into the road itself, or the sides of it.
Because frankly the main problem will come from the maintenance of the cars : what happens when some sensor dies, or gets dirty ? Or if the computer is hacked, or an OTA update halfly borks it ? On a highway this would have no consequences other than you have to switch to manual. But in a European town with tiny crowded streets ?
Highways have the special case that everybody is driving in the same direction, there are no crossings, no pedestrians, no cyclists, no children playing ball on the side, no dogs ...
Wildlife, escaped farm animals, HGVs shedding remould casings from tyres, Audi drivers being complete wankers.
> All of which are pretty easily handled
Agree: AV sees anything out of the usual an performs an emergency stop and waits for road to become clear or for (human) passenger to take necessary action to restore the road to something the AV can handle, and press the resume button…
But "pedestrians, cyclists, children playing ball on the side, and dogs" are not easily handled? Or are you saying that performing an emergency stop when anything out of the ordinary happens is fine at 70mph but dangerous at 30mph.
I expect autonomous Audis will have a "wanker mode".
There are several problems it solves. Or will solve, when it is viable.
1) I want to go home from the bar/party but I'm currently drunk
2) I'm a salesman or other role where I have to drive around between appointments and I want to catch up on my email and phone calls when I'm in the car without endangering myself or others
3) I want to be somewhere tomorrow that's 1000 miles away, but don't want the hassle/expense of air travel but also without the issues raised by having to be behind the wheel paying full attention 15+ straight hours and arriving at my destination with little or no rest
While it isn't necessarily a "problem", I'd personally prefer not to have to deal with driving anytime I'm in the car. If it is able to drive itself and I felt 100% confident it could do so safer than I could even when paying full attention, why would I not let it? I'd pay a premium to get that feature, just like I would pay more to have cruise control or seat heaters in my car.
> problems it solves. Or will solve, when it is viable.
Solutions exist now
1 is a taxi or public transport where it exists (and yes we have 24 hours buses around here in the UK)
2 is an edge case, sound like you need to reorganise your life priorities
3 who the heck drives 1000 miles in one go ? As you may guess I'm in the UK and even Land's End to John o' Groats is quoted at only 846 miles. For that distance I'd take a train/plane and hire a car at the destination unless I had a couple of days to break the journey up.
My view is that most existing drivers would hate being in a car that is driving itself and being expected to take over if its computer craps itself , unless of course you have a level 5 car and they would not have a steering wheel at all.
1. Not everywhere has any sort of transport available 24 hours a day, even Uber isn't an option if there are no Uber drivers currently logged in. Or you might have to wait an hour or more because the few there are are busy - then you'll pay higher prices and still have no guarantee if or when they will actually show up. So no, that's not AT ALL the same as having your car available to drive you directly home any time you want.
2. Was an example. That isn't me but that is true of a lot of people. I like you simply dismiss their needs as unimportant.
3. Few will drive 1000 miles in a go - that's typically a flight. But if you don't want to fly because it is expensive (especially for last minute travel) or a pain in the ass with all the security theater, having the OPTION to have your car drive you 1000 miles would be nice. For my current car that would be around $100 of gas, or $200 round trip which is cheaper than any discount airline, and much cheaper with an EV. It would also be FAR more comfortable if my car was totally autonomous, since I could lie flat and sleep - if you compare the cost to the first class ticket required for THAT it is even more stark of a comparision.
You're a moron if you think "most existing drivers" would hate a car that drives itself. Most people DO NOT enjoy driving. Even those who do enjoy driving don't enjoy ALL the driving they do. Sure on a nice day on a scenic road it is great. But driving on a freeway? NO ONE enjoys that.
CSB time:
I have avoided taking Waymo because I was dubious about the technology, and I wanted to put at least some money into the pockets of human drivers. Then I had two back-to-back Lyft drivers who had issues, one with driving safely and the other with navigating (plus he claimed to be deaf and clearly was not). The next day, I took my first Waymo ride, and . . . it was great! The vehicle braked, accelerated, and turned smoothly and with consistent attention to traffic rules. It took appropriate routes and successfully navigated around standard hazards such as double-parked cars. I felt substantially safer in that car than I have in numerous human-driven vehicles. My only real fears taking a Waymo are getting stuck behind some truly difficult obstacle such as an unmoving garbage truck and having my vehicle interfered with by hooligans. I'll still prefer human drivers for the reason stated above, but Waymo, at least, has raised the bar on autonomous driving.
The problem is that you have to choose between fulfilling peak demand or average demand.
If you fulfill peak demand customers will be happy, but the economics don't wok since the robo-cars are mostly idle. And also it will mean the same number of cars on the road as today...
If you fulfill average demand you customers will be pissed they can't get a robo-car, so once again the economics don't work since too few customers are paying.
Hmmm, it's like maybe a different mode of transportation may be the actual solution...maybe something public? Network of well-built public transport? <irony>I wonder when someone will invent that concept...</irony>
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
