Tesla has a self-driving strategy other companies abandoned years ago • Hacker News Board

Re: Tesla has a self-driving strategy other companies abandoned years ago

#151

Post by ec109685 » Sat, Mar 09, 2019, 4:03 AM UTC

bunderbunder wrote: post #122

Animats wrote: post #49

Waymo keeps plugging away. Each year, for the last few years, the number of miles between disconnects they report to the CA DMV has doubled. Three more doublings, and that number will be bigger than average miles between accidents for human drivers. Then they can ship a product. Nobody else is even close. This problem is slowly being solved, by normal engineering practices. The "fake it til you make it" players are b…

> Three more doublings, and that number will be bigger than average miles between accidents for human drivers. Then they can ship a product. They can ship a product that can drive under the conditions they're testing under. Which would be appropriate for autonomous taxi service in certain municipalities, but presumably not for the general market. Granted, that's been their business plan for a while now.

So can Tesla. They don’t have to enable full automous driving in all conditions.

Re: Tesla has a self-driving strategy other companies abandoned years ago

#152

Post by ec109685 » Sat, Mar 09, 2019, 4:11 AM UTC

devy wrote: post #64

The top editor promoted comment[1] on the article summed up very well. The attention problem is well known in engineering. It is very hard to get a human to concentrate on something that will turn up good more than 99% of the time, even when there's serious or fatal consequences of failure. Trains are the classic example - tracks are amost always clear, signals are almost always correct which means you have to devise…

I am not sure how Tesla’s system is much different than Waymo’s. In both cases, the cars can steer themselves under limited circumstances. Getting to full auto driving in crazy situations like single lane roads or unprotected left turns will take a long time. But using it in the situations it can handle today is very useful. At least with the Tesla, you can actually drive in places it doesn’t support auto pilot, unlike Waymo. The Tesla autosteer won’t engage.

Re: Tesla has a self-driving strategy other companies abandoned years ago

#153

Post by SEJeff » Sat, Mar 09, 2019, 4:12 AM UTC

bombas wrote: post #91

leesec wrote: post #84

Earlier quoted context omitted.

What do you call it when a Tesla drives me 90% of the way to work where I don't do anything but observe? It's not full self driving but it definitely is driving it self and providing value.

^Exactly. I think Tesla's marketing is shit around this. Elon needs to shut the hell up on twitter as well. Stop coming at me with Waymo is 10 years ahead of Tesla... Ok great, are they shipping & delivering any value to paying customers? Waiting....... In 5 years time, do you really think Waymo will have paying customers and have gained regulatory approval? Hell no... Tesla at least realizes they can't get passed th…

Very much this. I see it as a waterfall vs agile approach when you use software development methodologies as a comparison. Tesla put out a minimum viable product, and is quickly iterating on it. It isn't the ivory tower dogmatic example of perfection, but it is really good, and is tons of fun. Waymo on the other hand, is trying to just skip putting their products seriously in the hands of users and are doing limited trials in limited mostly perfect driving conditions (like Phoenix, AZ).

The best thing is that as much as people want to argue on the internet about "which is better", it is somewhat inevitable they'll both succeed in their own way. Waymo will eventually crack the nut for Level 5 autonomous driving with Lidar just like Tesla will crack the nut for their full self driving (defined as driving from the east to west coast without any involuntary disengagements) with machine vision. I actually (potentially incorrectly) suspect Tesla will win this war first as they simply have a multiple of any competition when it comes to amount of driving data. How many more hundreds of millions of miles has Tesla autopilot logged compared to Waymo? Reminder: Tesla is building ~5,000-6,000 cars per week and tends to sell them pretty well. Their data moat only continues to widen. Once they figure out the right models for really pulling it off, they'll make HUGE leaps with the data they have. Some nights, my Model 3 will upload 10G of data up to Tesla. Hopefully, it is from Shadow Pilot to train the fleet.

TL;DNR: There is nothing stopping both Tesla AND Waymo from solving full self driving using competing technology and different approaches. It is an exciting time to be alive!~

Re: Tesla has a self-driving strategy other companies abandoned years ago

#154

Post by mbreese » Sat, Mar 09, 2019, 4:13 AM UTC

zaroth wrote: post #99

mbreese wrote: post #95

Earlier quoted context omitted.

And my wife’s Acura will completely steer itself (with a much shorter list of features), but requires your hands on the wheel. It works without your hands, but it gets testy if you don’t move the wheel slightly every 15 seconds. It’s significantly nicer to drive long stretches with this system, but I can’t imagine doing it without your hands on the wheel. When the sensors fail (weather or poor lane markings), you nee…

Are you taking about LKAS? [1] Not trying to be rude, but that is nothing at all like driving with AutoPilot. [1] - https://youtu.be/iKWBdySgssE

You’re not being rude. It is the lane keep assist, and it is nothing like Tesla’s autopilot. But that was the point I was trying to make. It’s a very good, but limited, drivers aid. But because it’s likited, it won’t make (or rather let) you get complacent. This was the gist of the original article — the Tesla autopilot system is very risky when the system needs to hand over to the driver. The driver is likely to trust the system too much and therefore let down their guard and lose key context that is needed when the autopilot sees something that it can’t cope with. In those situations, a more limited system is likely safer because the driver is forced to be more engaged throughout the trip.

Re: Tesla has a self-driving strategy other companies abandoned years ago

#155

Post by ec109685 » Sat, Mar 09, 2019, 4:13 AM UTC

xeromal wrote: post #53

CPLX wrote: post #29

It's funny how much people have danced around all this, including this article, but what Musk has done with his statements on self-driving capabilities is called lying .

If someone has a delusional idea, is it still a lie? I don't think Musk 'lies' to earn more money. I think he is delusional for how long certain things take to do and at worst is trying to keep his companies afloat so he can continue with his ideas. I'd say his sins are less than the CEO of Enron for instance. It's hard to blame the guy when he successfully built a rocket company.

What do you call founders that promise their company will take over the world to VC’s and then burn out spectacularly?

Re: Tesla has a self-driving strategy other companies abandoned years ago

#156

Post by pests » Sat, Mar 09, 2019, 4:13 AM UTC

leesec wrote: post #117

close04 wrote: post #94

Earlier quoted context omitted.

What kind of traffic and road conditions does that trip involve and how much are you an observer vs. you hanging on to the wheel and paying attention to everything as if you were driving yourself (staying engaged) as you are required by the law and by common sense? And if you don't mind me asking, how would you quantify in real money the value added by the system?

The 90% I am talking about is a 12~ mile stretch of poorly maintained highway near Detroit usually in stop and go traffic and yes I am paying attention. I wouldn't quantify the value in real money. I would say the main benefit provides is that I have much less decision fatigue and am less stressed after my commute. I also don't care how fast I really get there anymore since I just optimize for less stress, and my tri…

I'm also in Detroit and my roommate owns a TM3. Ive drivin in it with him for hundreds of miles in autopilot and agree with you completely.

Re: Tesla has a self-driving strategy other companies abandoned years ago

#157

Post by ec109685 » Sat, Mar 09, 2019, 4:17 AM UTC

alkonaut wrote: post #44

syntaxing wrote: post #4

Not having LiDAR for anything over level 3 self-driving capabilities seems like a very bad idea...Computer vision right now just does not have the spatial awareness that you need for self driving capabilities. I wish Tesla would work on something similar to the Kinect v3 but for long range (above 20 m).

Regardless of whether cameras or lidar win out, just consider the fact that back up cameras already exist on tons of modern cars, and the solution to the environment issues of these cameras is one of: a) fold it out when you put the car in reverse, so it's always ice-free b) ignore this problem. It's covered in ice for months. This is a simple problem in comparison, and neither of the above solutions would work for a…

If the camera is broken, Tesla will obviously not enable auto driving. It has a steering wheel for a reason.

Re: Tesla has a self-driving strategy other companies abandoned years ago

#158

Post by SEJeff » Sat, Mar 09, 2019, 4:18 AM UTC

_ea1k wrote: post #35

peeters wrote: post #32

I've always found that half the issue is with the name: Autopilot. On an airplane, autopilot means you can take your focus off of both the physical and mental act of moving the plane through the air, and if it needs you to intercede there is pretty much no emergency in which you wouldn't have a few seconds to react. In a car, Tesla's Autopilot simply doesn't give you that. If you are reading and your autopilot exits,…

Autopilot in planes can be used, and often is used, in cases of VFR flight in which see-and-avoid is a requirement. Constant surveillance of the aircraft's surroundings is important. When it isn't done, midairs happen.

This should only be the case if you have lost your radio or have utterly incompetent air traffic control. It could also be the case if you're in class G airspace or in some BFE remote area without a tower, but meh. As stated elsewhere, autopilot on planes is pretty much not at all what most people think of as "auto-pilot". The pilots still are doing their instrument scan, and they're still maintaining situational awareness, but they know that the plane should maintain heading, altitude, and airspeed.

Source: Have private pilot's license and flew the Shadow 200 TUAV for the US Army. It is a VFR rated airframe that is flown 100% IFR, by nature of it being an ~11ft x 3.5ft x 2ft drone.

Re: Tesla has a self-driving strategy other companies abandoned years ago

#159

Post by ec109685 » Sat, Mar 09, 2019, 4:18 AM UTC

ableal wrote: post #3

"Self-driving cars also benefit from lidar sensors, and the best ones cost thousands—if not tens of thousands—of dollars each. That's too expensive for an upgrade to a customer-owned vehicle. But the economics are more viable for a driverless taxi service, since the self-driving system replaces an expensive human taxi driver." That's probably the crucial point. For now, lidar is needed for safe operation, and too exp…

What is your evidence that lidar is needed for safe operation?

Re: Tesla has a self-driving strategy other companies abandoned years ago

#160

Post by threeseed » Sat, Mar 09, 2019, 4:19 AM UTC

toomuchtodo wrote: post #131

danaliv wrote: post #107

Earlier quoted context omitted.

You're absolutely right, but the issue isn't what an aircraft autopilot actually is. It's what the word "autopilot" evokes in the minds of the non-pilot public. To them, autopilot means fully automated with no oversight. ("Planes fly themselves!" and other such nonsense.)

Please don’t force marketing to conform to the lowest common denominator. People have some responsibility to educate themselves about the products they’re using.

You are being ridiculous.

The definition of autopilot has been this way for decades. You could literally ask anyone of any age what it means and everyone would have the same answer.