Uber lays off 100 safety drivers as it scales back self-driving tests
- Thread starter JournalBot
- Start date
To be frank, Uber's reduction in quantity does not necessarily signify to me they're aiming for any sort of quality. I would accept in good faith this argument from most other organizations, but not this one. It seems as likely as not that they're just being cheap or playing a PR game.
Upvote
67
(69
/
-2)
So basically they did a shit job and have to rethink their strategy.
Waymo is so ahead other players that it's probably better just to trust them and license their technology, and end with this pantomime.
I remember a leak last year with regards Uber disengangements, where they had like one disengangement every 0,6 miles or something like that. This year I think it was like once every 13 miles. Ok they improved quite a bit....
Until you realize Waymo cars drive on average for FIVE THOUSAND miles before being disengaged. And they also drove like two orders of magnitude more miles. And that has been true since like 2016.
It's simply ridiculous. A ridiculous difference.
It being ridiculous is also consistent with the ridiculous company Uber is. Absolutely abhorrent human beings doing unethical business.
Waymo is so ahead other players that it's probably better just to trust them and license their technology, and end with this pantomime.
I remember a leak last year with regards Uber disengangements, where they had like one disengangement every 0,6 miles or something like that. This year I think it was like once every 13 miles. Ok they improved quite a bit....
Until you realize Waymo cars drive on average for FIVE THOUSAND miles before being disengaged. And they also drove like two orders of magnitude more miles. And that has been true since like 2016.
It's simply ridiculous. A ridiculous difference.
It being ridiculous is also consistent with the ridiculous company Uber is. Absolutely abhorrent human beings doing unethical business.
Upvote
58
(59
/
-1)
IncorrigibleTroll
Ars Tribunus Angusticlavius
There are plenty of good arguments for having a few different players (that eventually merge into an open standard). Waymo, while doing very well so far, could have some blind spots or oversights in their testing that are concealing problems, and proprietary standards aren't great.
There aren't many good arguments for why Uber should be one of those players, though.
Upvote
50
(50
/
0)
.
Uber has no choice in this matter. If they are unable to produce self driving cars themselves they will not be a tech company any more. If they are suddenly valued as a taxi company todays lofty valuation will collapse and venture capital will dry up.
Upvote
41
(41
/
0)
Definitely agree. I was trying to mock Uber, not competition to Waymo, which is obviously positive in every regard
Upvote
13
(13
/
0)
No, they're totally disrupting the status... quote... thing... and, you know, creative destruction, innovation, move fast, break shit. And killing people is really just breaking stuff, if you look at through the right sort of tech squint... or... something...
LOOK! OVER THERE! TRUMP JUST TWOTE!!!!
=============
I think what we're seeing is that self driving cars are hard - and some of the self evidently "That should work because humans can do it!" type approaches aren't something we can actually make work yet (pure camera based, or even camera and radar). Waymo's successes are heavily based around having an absurd amount of data about their environment, and about what their environment should look like. It's easier to find the things that don't belong when you have cm precision maps of what the environment is. They're using gas cars, and I expect a large part of that is due to the server racks wedged in somehow - if you're drawing a few kW sustained to run your compute, an electric car's battery just doesn't last for a full day's driving.
We'll see who comes out ahead, but at this point, I think it's fair to say that Uber is not going to be disrupting the self driving taxi market any time soon. I'd place far more money on them getting "disrupted" in the next decade - at least in certain areas.
Living out in rural farm country, I don't expect self driving cars to be a factor for me or anyone else in this area for a long, long time...
Upvote
2
(9
/
-7)
I don't think anyone is going to open source their self driving stuff (at least not working self driving stuff - I can see Comma or whatever they're calling themselves now releasing code). It's far too valuable a secret sauce, and it's certainly so custom to the hardware in use that the code, alone, wouldn't be particularly useful.
As far as blind spots in testing, it's possible, but "miles in the real world" beats just about every other way of generating real world situations. They had an issue a number of years ago with a cyclist doing a track stand - not common, but put a couple hundred thousand miles a year on, and you come across those. So I'm not that concerned about them missing stuff, because they're putting literally millions of real world miles on their vehicles, collecting data, and simulating from there. Being an Alphabet company, I'd wager they still have every bit of sensor data, from every drive, stored - because they can, and because it's useful to test new algorithms.
Upvote
11
(11
/
0)
I agree, Uber believed that they either had to figure out self-driving first or they would be out of business. Which is why they were playing fast and loose with quite a few best practices. That doesn't absolve them of the terrible decisions, just explains how a company can make such poor choices.
Upvote
8
(9
/
-1)
Upvote
20
(20
/
0)
IncorrigibleTroll
Ars Tribunus Angusticlavius
Irresponsible, sociopathic assholes in charge?
Upvote
26
(26
/
0)
I don't agree with this. Self-driving cars and ride-hailing networks could easily prove to be complementary networks. A decade from now we could easily have a bunch of competing self-driving car makers. Customers aren't going to want to keep 10 different apps on their phone to decide which one can get a car to them quickest. They're going to want to have one app that can provide service from whoever has the quickest and cheapest service.
This is the vision Lyft is pursuing and it seems entirely plausible to me. They probably won't be able to get the 20 percent cut they get today, but a 5 or 10 percent cut of a much larger industry could be a very nice business to be in.
Upvote
21
(22
/
-1)
This just confirms Uber's original problem. Someone came up with the (somewhat reasonable) idea that a huge number of road miles would demonstrate safety (they tested the heck out of this product). Uber said the heck with testing, lets just run up the miles of of this so we can check that box.
Upvote
14
(14
/
0)
It's interesting how a single "safety drivers" failure to meet the prime directives led to 100 "safety drivers" getting the axe. As the ripple becomes the wave...
Upvote
4
(6
/
-2)
Yep, provide a platform that legit taxi/transport services can tie into. I still think law makers (and the community) will eventually find that what Uber is currently doing should be classified as a taxi service and taxed/regulated as such. They are just biding time until they figure out exactly what to do next.
However, Uber has proven time and time again that they aren't exceptionally great decision makers. As such, whether they can make that shift in their core offering or not is very much up for debate. And if they don't they'll be remembered as a case-study of how to fail epic.
Upvote
10
(10
/
0)
Running with this idea,
Expedia (which also owns Travelocity) has $18.5B assets on $10B annual revenue (on which it turns a $380M net income).
Uber has $15.7B assets on $6.5B annual revenue (on which it turns a $2.8B loss).
The bottom could absolutely fall out for Uber even if they saw an opportunity to serve as a car-distribution network for other operations.
Upvote
7
(7
/
0)
It's a vertical integration question.
As you say, there's no particular reason the ride hailing network should also be the developer of autonomous driving capability. Just as Uber or Lyft don't build their own cars, there's no reason for them to build the software in those cars.
And while, as a consumer, right now, I would care a lot whose software is controlling my robot taxi (because only one or two of the software developers seem to have something that works), in the long run, I shouldn't care about that at all. I want my robot taxi to show up when I ordered it and to be in clean and working condition. Thats what the taxi network needs to do. I don't care about their maintenance costs, what suppliers they pay, etc. I do care that they've got that all figured out so the robot taxi shows up and gets me where I'm going.
Short term: vertical integration can give you an advantage. Long term: it's almost always a disadvantage ...
Upvote
1
(3
/
-2)
Considering Uber's reputation, I can't help but think they are just brushing the whole mess under the carpet. Fire a few people, rebrand and get back to what they were doing before.
Upvote
9
(9
/
0)
Statistical
Ars Legatus Legionis
Well like he said valuations will collapse. Uber currently takes 30% and the concept has been sold to investors that they will be getting 30% off the top of essentially all tax traffic in the world. Dropping to a 5% cut as a middleman running an app means Uber loses 90%+ of its value.
For the record I think Lyft's strategy is more likely. Just like airline booking sites that exist as a middle man making a couple percent connecting travelers and operators there is a business model there. The same can be done for AV fleets but it is going to be a lot less profitable that what Uber has promised investors.
Upvote
10
(10
/
0)
If 100% of your testing is at your challenging test track dodging fake pedestrians and on-purpose-idiotic drivers, it _can_ be better than real-world testing where you're doing highway milk runs from Phoenix to Tuscon just to rack up miles where an exciting event only occurs every couple hours.
Of course, it could easily be a whole lotta PR spin too!
Upvote
3
(3
/
0)
Upon first glance, I read the title to this article as "Uber lays off 100 safety drivers as it ramps up self-driving tests." An odd brain fart, but totally logical and consistent with Uber's overall reputation.
Upvote
5
(5
/
0)
Id never trust a system that expects me to jump in and take over if it makes a mistake. Consider that if it normally takes an active driver X seconds to react to an event, if that driver is not actively driving it will take at least twice that long to realize the car is screwing up and grab the wheel and brake. It may be safer in general but in the times where it cant do its job the risk has skyrocketed. This goes not just for Uber but for any "autopilot" system.
edit: and no I dont use cruise control either
edit: and no I dont use cruise control either
Upvote
2
(3
/
-1)
Uber's management was way too optimistic about their driver tech and the recent accident goes to show their over confidence can be costly if it wasn't done in a controlled environment. This is what happens when you let business people make technology decisions in the physical world you have consequences. Unlike the virtual world you can throw software out there without much exposure to risk.
Upvote
2
(2
/
0)
Uber's driver vetting process is a joke. Plenty of sexual predators are driving for them. If you read the horror stories their investigation process is a complete joke. It's like dealing with eBay you have to wait for a customer rep to consult you and they don't want the authorities to know at all.
Upvote
1
(1
/
0)
Whoever in a company makes up stupid titles like "mission specialist" is dead weight and should be fired.
Upvote
3
(4
/
-1)
I agree with you about everything up until the final point. In this particular case, you are correct. Vertical integration gives Uber little in the long run, but it's not a general rule. Samsung is a great example where deep vertical integration (microcode, chip design, chip fab, display, screen, semiconductor storage, phone) gives it an *enormous* competitive advantage.
But transportation is a commodity, and I think that it's much less useful in commodity supply chains, in general.
Upvote
3
(3
/
0)
Then you'd never trust an airplane autopilot either, because they *by design* dump control of a plane back to the pilot when certain parameters are exceeded. Unfortunately, they have a tendency to happily maintain control right up to the bitter end of the design envelope, then disengage and dump an uncontrollable plane into the hands of the unsuspecting pilot (see the accident involving American Eagle Flight 4184).
So, upon reflection, I guess you're right.
Upvote
0
(1
/
-1)
To Uber, getting self-driving down is an existential problem. They don't necessarily have to be first to market, but they can't be very far behind first or the money will dry up and they will cease to exist. That incident in March was a serious blow to the company, one that I'm not sure they'll recover from.
I won't mourn the loss of Uber. They have been an unethical company all along, and I refuse to use them (I always use Lyft instead when I need that service).
What I do lament, however, is that this represents one less viable self-driving option. Once autonomous vehicles are a common thing, there will most likely never again be an independently developed system because the cost to develop such a system will be far too high to be able to compete with the established players. That means that the ones in development now are the only ones that there ever will be.
I'm aware of 5 different systems in development (there might be others): Uber, Waymo, GM Cruise, Apple, and Tesla. My hope is that all 5 of them reach the finish line as a viable option for self-driving cars.
I won't mourn the loss of Uber. They have been an unethical company all along, and I refuse to use them (I always use Lyft instead when I need that service).
What I do lament, however, is that this represents one less viable self-driving option. Once autonomous vehicles are a common thing, there will most likely never again be an independently developed system because the cost to develop such a system will be far too high to be able to compete with the established players. That means that the ones in development now are the only ones that there ever will be.
I'm aware of 5 different systems in development (there might be others): Uber, Waymo, GM Cruise, Apple, and Tesla. My hope is that all 5 of them reach the finish line as a viable option for self-driving cars.
Upvote
2
(2
/
0)
I'm aware of the context, but in isolation how is laying off 100 safety drivers supposed to improve safety track record for Uber - it's just a cheap PR move
Upvote
1
(1
/
0)
If Uber is going for quality testing, they sure are going about it in a stupid way. Firing everyone and expecting them to reapply for their jobs isn't how you select the best workers, it's how you select the most desperate workers. If anyone was any good at that job, they would go somewhere else where they are treated much better as employees.
But Uber's gonna uber. Maybe Dara will shoot another commercial where he talks about how he wants to listen more, and the suckers will forget how abominable this company is.
But Uber's gonna uber. Maybe Dara will shoot another commercial where he talks about how he wants to listen more, and the suckers will forget how abominable this company is.
Upvote
3
(3
/
0)
