No reasonable person would expect what is sold on Tesla's Autopilot page - including 250m forward radar, two types of automatic braking and collision warning - to erroneously half-take an off ramp and run into a concrete divider without ever braking. No matter what warnings the manual or car say about keeping your hands on the wheel.

Similarly, no one would expect it to not see a giant truck blocking the entire highway, pass under said truck decapitating the driver, and go for another 400 feet through two fences before crashing into a power pole - all without braking. No one would expect it to swerve into oncoming traffic because their car-in-front following is so atrocious that it mistakes an approaching car for what it should be following. No one would expect it to swerve dangerously when there's a sun flare or a lane disappears, trying to find a new one.

Yet this is the state of "autopilot" as extensively documented on the Tesla forums.

But it worked as designed.

The most fascinating data to me about autopilot is the above quote from the NHTSA report after the 2016 crash. I would love to know more about this data. What was the sample size? Is there updated data 2 years later? Does it show the same or different results?

If it continues to hold that Teslas that are capable of autopilot deploy airbags significantly less frequently than Teslas that are not capable of autopilot, then that is pretty good evidence to me that it works. This data seems powerful to me because it only looks at whether a car was capable of autopilot, not whether it was engaged. So there is no selection of certain types of miles that might be more or less likely to have an accident. Tesla drivers who own autopilot and tesla drivers who don't seem to be pretty equal groups. I can't think of any reason to think that one or the other would be better drivers. And airbag deployment is a pretty easily measured end point that is a good indicator of a significant crash.

Hopefully we will get a 2018 update on this data. This would be a lot more useful than Tooth repeating the details of the two fatal accidents in gory detail over and over and over again. The fact that this data has not been released makes me think that the difference seen in the 2016 report has gone away. I'm sure Tesla has access to this data and if it still showed a significant difference then a PR master like Musk would certainly be tweeting away about it.

Musk wouldn’t talk about this kind of data because he’s not selling “safer.” He’s just selling safe, having less airbag deployment suggests there are still accidents.

It’s kind of like Volvo wouldn’t advertise only xxx people died in a Volvo instead of xxx the year before due to some new safety feature.

Instead, they both just insist and talk about how they have the safest cars without going into raw data.

Why on Earth didn't they looked at the data on whether it was engaged? Surely Tesla has this.

I don't know the details on what Autopilot installation entailed, but collision avoidance and warning systems, not to mention automatic braking, save lives and tons of crashes, especially smaller ones. The seem to have appeared in 2015 on the Model S. They're on high end cars and one reason why late model cars are so much safer.

Autopilot could be significantly more dangerous, yet attenuated by collision avoidance systems. For example:
That's a huge drop in airbag deployments. This is what Tesla added to Model S in 2015, and in X. They're separate systems to "Autopilot". I imagine basic lane keeping and lane leaving warning systems reduce crash rates further. They do in other cars - many BMW models have zero deaths despite more cars on the road than Tesla.

By the way, these early systems were MobileEye systems. Tesla did in house after breaking up with them in 2016, and released shoddy in-house work soon after - work so bad that lead engineers quit after disagreements with Musk about releasing it while unsafe.

It would be nice to have hard data.

In both cases you are toast

TS,

I'm shocked that you can't keep track of your pathological lies. Much like your alleged "trading history", you have no idea what you've even claimed. It really is amazing how you know so little about business. Eager to hear about GM's batch production for their next vehicle while they sort our their production line!

Trolly,

Could you cite where Tesla even makes that claim?

The sales brochure posted earlier says "Fully self driving hardware." Yes, I realize there's fine print that qualifies it, but you can't look at that brochure and tell me they aren't over-selling what the car they sell actually does.

Okay. Glad you're actively walking back on that. You're quibbling over whether the car has LIDAR which may or may not be needed for SDC. TrollSayer thinks it isn't needed. Everyone else does. Elon doesn't.

Great to come to an agreement.

The heck are you even taking about? I said nothing about LIDAR or whatever.

He was referring to TS.

Tesla being down despite market ripping seems very bearish to me short term

That was the general "you" not referring to TS. (Maybe his English has improved from the last go which involved made up terminology and poor vocab.)

Trolly,

The conversation around that phrase is whether the Tesla actually does have the hardware or not. TS somehow thinks that cameras can solve the SDC problem despite using cameras with under 1% of the resolution of high end cameras, or the fact the processing on the cars is a fraction of a home computer. Clearly they couldn't pull the graphics card out of your phone along with the camera to get a major upgrade. Elon has a slightly less insane view that claims LIDAR isn't needed.

Everyone else thinks the Tesla does not have the hardware for level 4 SDC. So it is kinda bizarre you don't know why that comment in controversial, but are assuming everyone who reads the brochure suddenly is illiterate (but somehow can read well enough to pass a driving test).

I guess this really sums up the case of the Tesla bears, they lack any real familiarity with the subject so they look for nonsense to ramble about. (OMG ITS CALLED AUTOPILOT!!!!) Instead of actual issues. (Tesla has stopped giving model 3 production guidance.)

Elon mentioned 5k cars/week on Twitter (he was referring to July).

https://twitter.com/elonmusk/status/982848849361289222

https://twitter.com/elonmusk/status/982849024083484673

I'm shocked to see Elon engaging in some bush league propaganda.

So was I.

You guys are joking right?

I'm 100% serious, which is why I view this thread as an ongoing joke. It has become an ideological war between bulls and bears where the company is either worth 1 trillion or 0, when in reality they are at a significantly worse position than a few years ago and need to produce a pretty large amount of cars to just keep up with the company financing, but most people are idiots and think Elon's ramblings on Twitter are gospel. (How many hours a week is he spending at/on Tesla? 20-30?)

You string logic together in a way I don't understand

TS spamming the tesla motor club douchebag site (ok obv but the screen name made me laugh)

https://teslamotorsclub.com/tmc/thre...a-scam.112768/

The legend spreads. Solid poster, worthy of the name.

Those pics are insane


I laughed at "being treated like dirt, riding next to dirt"

lol, so now he's pretending to be in New York?

I wasn't, I was quite shocked to see him say that on Twitter.

A small startup recently released this drone with obstacle avoidance. It uses 12 cameras for obstacle avoidance and they fit all this tech into a drone where every gram makes a difference.

The drone can "see" front, back, left, right, up, and down. It can fly backwards avoiding obstacles while also predicting the path of the subject they are optically tracking, in this case, the runner. The unit can track up to 25mph at which point it starts to outrun it's obstacle avoiding sensors.

The unit sells for only $2.5K.



This is what they achieved in 3 years and less than 30M raised.

Of course things get more complicated at higher speeds and cars having the additional burden of not just driving, but following road rules and other moving obstacles.

Regardless, maybe this is interesting to people thinking about Tesla's autopilot.