Uber vs. California - Robot cars not ready for roads?

Uber just went ahead and started using self-driving cars a couple of regions, including San Francisco. These special Volvos have a “safety driver” behind the wheel to watch for errors, but they are 100% self-driving.

The clash is that California says Uber needs a special testing permit. Uber says they don’t because of the presence of the safety pilot.


The California DMV encourages the responsible exploration of self-driving cars. We have a permitting process in place to ensure public safety as this technology is being tested. Twenty manufacturers have already obtained permits to test hundreds of cars on California roads. Uber shall do the same.[/quote]


Finally, we understand that there is a debate over whether or not we need a testing permit to launch self-driving Ubers in San Francisco. We have looked at this issue carefully and we don’t believe we do. Before you think, “there they go again” let us take a moment to explain:

First, we are not planning to operate any differently than in Pittsburgh, where our pilot has been running successfully for several months. Second, the rules apply to cars that can drive without someone controlling or monitoring them. For us, it’s still early days and our cars are not yet ready to drive without a person monitoring them.

But there is a more fundamental point—how and when companies should be able to engineer and operate self-driving technology. We have seen different approaches to this question. Most states see the potential benefits, especially when it comes to road safety. And several cities and states have recognized that complex rules and requirements could have the unintended consequence of slowing innovation. Pittsburgh, Arizona, Nevada and Florida in particular have been leaders in this way, and by doing so have made clear that they are pro technology. Our hope is that California, our home state and a leader in much of the world’s dynamism, will take a similar view.[/quote]

In the middle of this, pedestrians and other drivers are seeing stuff like this:

That’s an Uber self-driving Volvo shooting through the red light. Uber, of course, says that’s “driver error” because the safety person didn’t catch the robot’s error.

Leaving aside the issue as to whether they are legally allowed to do this, and just talking about the specific incident:

To be fair, this is a weird stoplight in the middle of the block which appears to be for a pedestrian crossing, so it’s not like it blew through a regular intersection. Should it have stopped? Of course. Was that a problem with the car? Absolutely. But I don’t think it’s a sign that self-driving cars are useless. (As an aside, did you know the guy reporting it works for a cab company? Just sayin!).

I also think this is how errors are going to be caught - by actually using the cars on real roads. Issues will come up, they’ll get fixed. Finding stuff like this in testing seems to me to be a good thing.

Also, a lot of the people complaining seem to miss the fact that we’re talking about using these as opposed to having people driving - people who typically are fiddling with the radio, talking on the phone, gesturing at pedestrians, talking to passengers, etc. Thinking there won’t be any fatalities associated with autonomous cars is silly… what we should be doing is comparing the fatalities of using self-driving cars versus the fatalities when using human-controlled cars. I have no problem figuring out which solution will eventually end up with fewer fatalities.

I like how the people are literally there to take the blame. Seems like SOP in a nutshell, really.

I don’t think self-driving cars are useless. I do think self-driving cars have flaws and the “safety drivers” are prone to doing what most humans do when they’re given a boring task like watching a robot car drive. They get bored and drift off to doing other things, missing dangerous errors. I’m not sure how any company is going to solve that issue.

Take the infamous Tesla autopilot guy. He died because he stopped paying attention to what the car was doing. The computer mistook the large white side of the truck trailer for the open road and barreled right into it while he was watching a movie. Tesla, of course, says that’s not what the autopilot is for, but again, they’re asking bored humans to watch a robot drive for them. Not gonna happen.

I agree with you completely. When I get into my robot car I’m going to flip up the lid on my laptop and play games until I arrive at work. So the software better be pretty freakin’ good!

While my chosen activity would be different, I’d read, the idea is the same. So, yeah, full autonomous or full driver. In between madness lies.

The difference here, though, is that this is literally the driver’s entire job. They are paid to be monitoring and controlling the AI driver, not playing games or reading or watching movies, and they are in control of an unproven test vehicle. If they are not watching for problems they are shirking their duties and putting their lives at risk, just like if the captain & co-pilot of an airplane sat there watching movies after take-off. If there’s blame here, it is to blame Uber for not properly training/vetting the driver and the driver for not doing his job. Or maybe it’s blown out of proportion and that driver would have done the same thing if he’d been actually driving the car, because the crosswalks looked clear and there were no cops.

It’s also interesting to me that Uber says there haven’t been any issues in Pittsburgh.

I’m sure the pedestrians that get mowed down will be pleased knowing they died in the service of progress.

This is getting pretty far from P&R territory, but self-driving cars are so far from being ready for actual use (no driver on board) I can’t really understand why Uber is even testing them now (other than wanting to piss off regulators, which is obviously a big part of Uber’s business model)

Before we see self-driving Ubers taking real fares, we’ll see:

  1. Self-driving campus shuttles (this is starting to happen now-- note these are slower and more cautios than a human operated vehicle would be, and only support a limited route)
  2. Automated trucks in convoy with driver-operated trucks (this is just starting testing)
  3. Self-driving airport shuttles (like 1, but have to travel in much more challenging conditions)
  4. Automated trucks that require an operator but that can autonomously pull over safely in most highway conditions, so the operator can be inattentive/sleeping.

Self-driving Ubers are probably 5 - 10 years after 3, which probably won’t be available for 5-10 years from now.

Uber needs to get to autonomous cars as fast as possible. They’re bleeding billions of dollars with their current business model. The only way they keep investors is by promising that they can get to driverless cars on the roads before all the other guys.

That is incorrect. According to Uber, that car was not in self drive mode at the time, and that the driver was in full control running the red light.

Is that better or worse?

Way better. This was the stupid human on board making a mistake, the computer wouldn’t have done that.

Yes, but is this better or worse for Uber’s assertion that they don’t need whatever permit California is saying they need?

Unfortunately, computers have stupid humans programming them.


I see your point now. Yeah… not better indeed.

Let us have our bad drivers ride around in your city running red lights, great!

I think it is funny that likely these tech companies are going to get lobbied and regulated out of working in liberal hippy California or San Fransisco.

Yeah, requiring a permit to test self-driving cars on public roads sure is some absurd nanny-state overreach…

Why do things have to go in that order? What’s basis of your 5-10 year assertion for airport shuttles? What’s the basis for your 5-10 year assertion for Ubers following those shuttles? If the low end of both those arbitrary ranges comes to pass (so 10 years to full automation), when should the cars begin road testing? Its not like Uber is the only one with self-driving cars currently being tested.

Things that are technically easier and have a surer path to economic return will happen first. That’s how you get the order these things will happen. The hard part of the autonomous vehicle problem remains unsolved, though it seems like companies are happy to soak up investor dollars pretending it isn’t.