A Lesson of Tesla Crashes? Computer Vision Can’t Do It All Yet

[FrodosPet]

A Lesson of Tesla Crashes? Computer Vision Can’t Do It All Yet

Jitendra Malik, a researcher in computer vision for three decades, doesn’t own a Tesla, but he has advice for people who do.

http://www.nytimes.com/2016/09/20/science/computer-vision-tesla-driverless-cars.html?_r=1

By STEVE LOHR | SEPT. 19, 2016

“Knowing what I know about computer vision, I wouldn’t take my hands off the steering wheel,” he said.

Dr. Malik, a professor at the University of California, Berkeley, was referring to a fatal crash in May of a Tesla electric car that was equipped with its Autopilot driver-assistance system. An Ohio man was killed when his Model S car, driving in the Autopilot mode, crashed into a tractor-trailer.

Federal regulators are still investigating the accident. But it appears likely that the man placed too much confidence in Tesla’s self-driving system. The same may be true of a fatal Tesla accident in China that was reported last week. Other automakers like Ford, which last week announced its plan to produce driverless cars by 2021, are taking a go-slow approach, saying the technology for even occasional hands-free driving is not ready for many traffic situations.

~ snip ~

Yet the recent advances, while impressive, have been mainly in image recognition. The next frontier, researchers agree, is general visual knowledge — the development of algorithms that can understand not just objects, but also actions and behaviors.

~ snip ~

Self driving cars are definitely coming. But in 2016, they are NOT ready for prime time.

To my knowledge, there are no known accidents in the Pittsburgh test. But there have been some issues with stop lights and pedestrians (citations coming in a future post as I forgot to bookmark them). Even the most enthusiastic people in the industry are starting to sound more cautious about their current capabilities, and the timeline to letting the cars roll in the wild.

Keep supporting the tech. Someday, it will help keep future generations safer. And from a personal stance, I would love to score a 6 figure job in development management. But it is not there yet.

[deathrind]

1. The auto pilot feature

Is not a recline your seat and take a nap feature...you still need to pay attention.

Teals should just remove this feature from their cars. People obviously are not equipped for it.

[FrodosPet]

2. I think renaming the feature is the first step

Done right, this could help prevent accidents. But I believe it needs to be renamed to emphasize the driver assistance role, not as a driver replacement.

[deathrind]

13. You are probably right.

[killbotfactory]

3. considering the 2.3 millions serious injuries, and 19,000 people killed in the US while driving.

Annually. I don't think we shouldn't rely on people to drive, and greatly overestimate our ability to drive safely.

http://www.newsweek.com/us-traffic-deaths-injuries-and-related-costs-2015-363602

[FrodosPet]

4. Someday, with enough processing power, computers will be better driver than humans

Today is not that day. It's coming. There are some amazing things being tested and developed as we speak. But it requires an exquisite level of mapping, literally down to the centimeter.

And once it is developed, it will take years before it becomes the primary mode on most passenger vehicles.

I am not trying to be a luddite. I support self driving technology as a major lifesaver. WHEN it is ready. But a number of intelligent, connected people are saying there is still a number of systems that need to be improved before the cars are turned out in the wild. I tend to believe them.

[killbotfactory]

6. Autopilot was never meant to be self-driving

The name throws people off, and it's overall effectiveness has lulled a couple people into fatal accidents, but it is not something that is a replacement for drivers taking their attention away from the road. Maybe the autopilot feature tricks people into believing they don't have to focus on the road, but hell... even without something like Tesla's autopilot, people still dick around on their smartphones while driving, or take phonecalls while driving, or get into intense conversations with their passengers, or see something interesting off the side of the road... all the goddamned time. 99% of the time driving is mundane and boring, so people don't pay attention.

[FrodosPet]

8. Hence, the need for the name change

The name deceives

[LanternWaste]

10. It's not deception, it's simple branding.

It's not deception, it's simple branding.

[FrodosPet]

12. So why don't they call it "Self Driving" then?

After all, the name is just branding, and not a description of what it actually does.

[Egnever]

5. Teslas are not self driving cars.

Using them as an example is not really fair. They use a very tiny portion of what self driving cars use.

They were never ever intended to drive themselves. They have driver assist. Really they have a fancy cruise control with lane detection and in theory collision detection. Basically a bit of an evolution on the more modern cruise controls that that are in higher end cars now.

They mental midgets compared to the google cars. Google cars ARE self driving cars. They are certainly not there yet for all conditions but in fair weather climates with fairly maintained roads they are all but there. At this point they just have to prove them. I am sure they will continue to evolve as well.

[FrodosPet]

9. You and I know that Teslas are not self driving cars

But between the accidents and the YouTube videos, the evidence is out that people are filtering. The name "Autopilot" has GOT to go. Maybe "Advanced Driver Assist". Not as cool, not as marketing friendly, but much more accurate.

You may disagree, but this issue is setting back public acceptance of self driving technology.

[Egnever]

14. Agreed autopilot is ridiculous

It gives the wrong impression.

[nationalize the fed]

7. 9/26: Another Tesla Crash, Autopilot Blamed Again

Tesla’s Autopilot has been held responsible for many crashes as the company tries to defend itself, saying it was the driver’s fault every time. However, once again a crash has taken place, this time with a Model S, and again Autopilot has been blamed. In the crash, in which a Model S is seen slowly pulling into the parking lot adjacent to a gym, it suddenly accelerates through the gym’s façade.

Owner disappointed with Tesla

According to Electrek, the owner of the Model S was hugely disappointed with the way Tesla responded to the incident. The owner claims that Tesla’s Autopilot system malfunctioned, leading to the incident. Tesla reviewed the car logs and concluded that the driver accidentally pushed the accelerator pedal, leading to the crash...snip

“I am amazed and wildly disappointed by the way Tesla has handled this and their complete unwillingness to even talk to me about it. Of course, they immediately blamed it on the driver and claimed their online computer tells them that,” he said...

...In the past, Tesla owners have complained on several occasions of the car exploding into full acceleration on its own. In June, Electrek reported on another incident, in which a Model X crashed into the side of a building in California...
http://www.valuewalk.com/2016/09/tesla-gym-crash/

It's just a matter of time before this "autopilot" nonsense ends up in a major lawsuit. A lawsuit that Tesla- burning ~$300 million per quarter- can't afford.

And by the way, anyone that thinks these cars are trouble free should spend a few minutes looking at the biggest Tesla owners forum: http://teslamotorsclub.com

[FrodosPet]

11. How Tesla's Self-Driving Autopilot Actually Works | WIRED

?list=RDUgNhYGAgmZo