Engineers were reportedly encouraged to limit "bad experiences" to one per trip.
Read the whole story
Read the whole story
Report: Uber self-driving team was preparing for CEO demo before fatal crash
- Thread starter JournalBot
- Start date
I mean, I'd think that being in a self-driving vehicle that murdered someone would qualify as a "bad experience," but I guess this is why I don't work for Uber.
Upvote
143
(144
/
-1)
You could say that about the whole self-driving industry. And that's an effect of the design of these projects; keep as much as possible under wraps to avoid snooping from your competitors.
The stakes are astronomically high for self-driving; the first company to make it to market with a solution that completely replaces a human driver, and is proven safer, will control the world's public transportation and shipping, no doubt about it.
The problem is that nobody wants to share that victory, so information regarding such projects has to be guarded with many layers of security. And often that leaves your own employees in the dark as well.
I submit that we can build a self-driving system that really does make the roads safer, but we won't get to that point if the technology is not developed openly and held accountable. Safety isn't a competition, you fail at safety if your inventions hurt people.
Upvote
42
(51
/
-9)
We've come such a long way. First we improved driver safety by enhancing the systems around the driver, then we introduced semi-autonomous driving systems with the driver as the safety mechanism, and now we are firmly in the midst of drivers backing up the autonomous systems as automated safeties are being turned off.
Upvote
-16
(5
/
-21)
Upvote
142
(143
/
-1)
I know exactly what the author meant and can't think of a more concise way of expressing "a fleet stopped for safety reasons while Uber does a rethink", but when applied to a ground-transportation technology, the term "grounded" made me giggle.
Upvote
25
(30
/
-5)
I wonder if it would be a good idea to build a safety town for self driving cars. A large enough area with varying road types and intersections. Automated dummies that move across the road, several self driving cars driving around, street lights that go out suddenly or flicker. Sure it's not exactly real world testing and would probably cost a lot and take a while to build but no one is going to die. A desert state with plenty of open space could build it and then rent it out to the car companies.
Upvote
8
(13
/
-5)
The companies seem to all be acting as if that's true but personally I don't see why. Sure there's some market advantage to being the first to market. But I don't see why I would expect the first company out of the gate to get anything close to a monopoly market share long-term. There isn't a lot of network-effect to the technology: each product more or less stands on it's own. There isn't a lot of lock-in either. I can buy a car from company A this time and easily choose company B next time. I think it's just as likely that the market will end up with a half dozen more or less interchangeable vendors.
Upvote
59
(59
/
0)
These basically already exist, except without the pedestrians, which are the most important part. Also bear in mind that with machine learning, more data is better. Without a huge amount of road to turn into data, you could over-train.
However, I do like the idea of such a town as a part of a test to pass before self-driving cars are allowed on public roads. At least perform acceptably in limited circumstances before you’re road-legal. You know, like teens.
Upvote
45
(46
/
-1)
Upvote
36
(36
/
0)
I don't know what it is about corporate hand jobs like this, but every time I read one, I feel like I need a shower.
Upvote
28
(29
/
-1)
Canterrain
Ars Scholae Palatinae
…
lemme see if I understand this correctly.
They wanted the CEO of the company to be able to take a test ride.
And they realized that the safety measures were making the ride less than comfortable.
…. so the answer was to make the ride less safe and risk the life of the CEO?
Really? Seriously?
Only Uber.
lemme see if I understand this correctly.
They wanted the CEO of the company to be able to take a test ride.
And they realized that the safety measures were making the ride less than comfortable.
…. so the answer was to make the ride less safe and risk the life of the CEO?
Really? Seriously?
Only Uber.
Upvote
89
(89
/
0)
There are two area's where I could see being first having a big snowballing effect.
1) Hardware costs - Current systems are extremely expensive which is why Waymo's starting with taxi cab's. The first one to market can be the first one to start benefiting from economies of scale to bring costs down so that could be a major advantage. On the other hand some of the most expensive parts, like the LIDAR or compute cores, seem to be coming from third party vendors for most systems so that might not really benefit the early mover.
2) Training data - They all seem to be relying on lots and lots of training data to get the systems working right. That's why they all talk about miles driven or why Tesla includes their hardware in all cars. First one to get a system out there and working will get the benefit of getting the most "free" training data from having the systems out there and in use to further refine their systems.
Upvote
19
(19
/
0)
That has been the underlying justification behind the astronomical valuations placed on Uber.
Problem is, I don't buy it. There doesn't appear to be any magic sauce behind autonomous vehicles, just a long hard slog in building experience and refining the sensors, programming and logic to cover an ever wider range of driving situations. Given the slow pace at which driving infrastructure, rules and regulations evolve, there's no reason to believe that second and third placed companies in the market won't have time to catch up with any first mover advantage.
Even if the competition was slow in catching up with some magic break through that enabled autonomous driving, I also don't believe that regulators the world over would allow but a single dominant player.
Upvote
38
(38
/
0)
Upvote
64
(65
/
-1)
Crushing financial liability is what will change corporate behavior. How about $100M for the victims family?
Upvote
31
(33
/
-2)
But... but... but... my Agile Coach is always telling me that remote working is better for teams than having everyone stuck in the office.
Upvote
-3
(13
/
-16)
I honestly don't think they'd even notice unless you add another couple of zeros onto the end that.
Upvote
23
(25
/
-2)
Yes, this. Very much so.
The same sentiment can be applied to a number of other instances as well. A capitalist money-driven economy is very good at some things, but demonstrably fails in many cases where there are high stakes but where cooperation would be more beneficial than competition. I wouldn't be surprised if the game theory geeks have a perfect solution to the problem, but if they do, they haven't told the software industry about it.
Upvote
7
(9
/
-2)
Upvote
19
(22
/
-3)
Only Uber? Hardly.
Greasing the wheels to keep the CEO in their smile-bubble is deeply ingrained into the U.S. workplace. Hack-jobs like this happen all the time at companies all over the country. The only thing special about Uber's hack-job was it turned out to be lethal.
Upvote
29
(29
/
0)
Upvote
15
(16
/
-1)
Just be glad the spec didn't call for maximizing the number of paperclips.
Upvote
9
(10
/
-1)
Maybe something like "waiting for the green light" for a more road appropriate metaphor? Stuck in neutral?
Upvote
2
(3
/
-1)
[/quote]
The companies seem to all be acting as if that's true but personally I don't see why. Sure there's some market advantage to being the first to market. But I don't see why I would expect the first company out of the gate to get anything close to a monopoly market share long-term. There isn't a lot of network-effect to the technology: each product more or less stands on it's own. There isn't a lot of lock-in either. I can buy a car from company A this time and easily choose company B next time. I think it's just as likely that the market will end up with a half dozen more or less interchangeable vendors.[/quote]
The company that gets it right will then license out the software to all automakers who won't want to be left out of the self driving group. They don't want to gamble on an uncertainty, so they'll go with what is proven.
So yes, it is a race.
The companies seem to all be acting as if that's true but personally I don't see why. Sure there's some market advantage to being the first to market. But I don't see why I would expect the first company out of the gate to get anything close to a monopoly market share long-term. There isn't a lot of network-effect to the technology: each product more or less stands on it's own. There isn't a lot of lock-in either. I can buy a car from company A this time and easily choose company B next time. I think it's just as likely that the market will end up with a half dozen more or less interchangeable vendors.[/quote]
The company that gets it right will then license out the software to all automakers who won't want to be left out of the self driving group. They don't want to gamble on an uncertainty, so they'll go with what is proven.
So yes, it is a race.
Upvote
1
(6
/
-5)
There's some kind of psychotic resistance to planning for failure with some of these people. It's always bright, cheery, positivity in their product design. This extends to plenty of software design I work with, where visual designers just don't want to show error messages no matter what happens, instead simply doing various auto-retry systems, or just NOT addressing the situation (which might mean clicking a button means literally nothing happens).
I feel like if I worked at a nuclear plant, someone might say to me "Hey, so about the alarm bell that screams 'MELTDOWN IMMINENT, EVACUATE IMMEDIATELY.' you were testing - it was really irritating to some people, and a lot of people were really scared about it. We were wondering if we could tone it down or just get rid of it since we don't want to have meltdowns anyway."
I feel like if I worked at a nuclear plant, someone might say to me "Hey, so about the alarm bell that screams 'MELTDOWN IMMINENT, EVACUATE IMMEDIATELY.' you were testing - it was really irritating to some people, and a lot of people were really scared about it. We were wondering if we could tone it down or just get rid of it since we don't want to have meltdowns anyway."
Upvote
66
(67
/
-1)
Always evergreen: Uber has the worth ethics of any large tech company. The company deserves to go off a cliff
Upvote
10
(12
/
-2)
Uber partnered with Carnegie Mellon University to build just this in Pittsburgh. Then Uber hired away 40+ CMU profs, and generally left the partnership to rot. Links in order of how the partnership unfolded:
https://www.cmu.edu/news/stories/archiv ... rship.html
http://www.thedrive.com/tech/15241/uber ... iving-cars
https://www.wsj.com/articles/is-uber-a- ... 1433084582
http://fortune.com/2016/03/21/uber-carn ... rtnership/
Upvote
17
(17
/
0)
Glad to not be seeing all the comments about how automated driving is so much safer than human driving. At least not today.
Upvote
1
(8
/
-7)
I’m not sure if you missed it, it says right there they were planning on using “safety drivers”
Upvote
0
(0
/
0)