Short-Term TSLA Price Movements - 2016

[WarpedOne]

This is vandalism, and there are 8 cameras to record the vandal.

One could fly in the robotic can of spray with a drone, very high up (thousands of feet) so the cameras wouldn't see it, directly over the car, lower onto the roof and carefully spray the cameras without being recorded.
Or the drone could climb out of road shafts if a car was parked over one and again disable the cameras.

You cannot say it is impossible, it is a real danger!

 

[Seesaw]

Here you can (or can't) see the new cameras

Can you spot the self-driving cameras on this Tesla Model S?

There is something oddly beautiful about how well they are integrated.

 

[CorneliusXX]

when you are using your vehicle for commercial purposes... you must get commercial insurance. if your car goes off by itself to pick someone up (this is 5 to 10 years away btw)... and runs over a nun because her black gown faded out into a shadow... who's at fault?... let's start with... not the nun... is it Tesla?... or you?... you're both making money off this... so could it be both?... how much will this insurance cost you?

why are we talking about this?... we were supposed to be talking about an affordable electric vehicle getting manufactured.

The management team outlined their goals in the SMP2. Anything related to those goals is within the current scope of the business. Even then the target is not static - as some items are achieved others can be included.

 

[NicoV]

There is something oddly beautiful about how well they are integrated.

It's as if the side badge was designed to house a camera from day 1. Very forward looking to take this into account in 2011/2012.

 

[Johan]

I think some bears made it up. Hopefully Jonas has a higher liklihood of a new stance either this or next week on the company.

OT: it's kinda fun listening to the Q2 ER again after having listened to the AP2 presser. Things fall into place, some gigles make sense, etc.

Neither AJ nor Morgan Stanley had said anything officially since October 4th. There is no new note to investors, I'm 100% confident that if there was an internal note to clients some of it would have been leaked and if there was a new official PT, even if it was unchanged in $$$, it would be made public. Re-iterating a price target is a new target - you're basically saying that x number of months ago you believed the stock would be worth y number of dollars one year later, now you're saying that you've changed your mind and still think it will be worth y number of dollars one year from now.

 

[Vitold]

Probably Brian Johnson's note started the downtrend in the afternoon. Just a guess.
Why Tesla’s Autonomous Car is Bad News For Its Stock

End of the article should say "as dictated by MobileEye PR department". This is obvious fud since we don't know how the system works and my bet is on a hybrid-not just a black box.

 

[erha]

For someone who really seems like they understand things: Am I correct in assuming Hardware 2 will learn the same features as AP1 on its own (Musk says will only take ~2-3 months for parity), and not use the mapping/learnings of AP1 which was based on Mobileye's system? Thanks.

No. I'm pretty sure that is not how it works. Tesla's autopilot is pretty rule based I think. And that is not a bad thing. AI can be rule based, it's just about how smart your rules are. To leave complete control of the car to a neural net is scary as you cannot be 100% certain of what the car will do in all situations even though it might perform well on tests after some training. A computer vision algorithm recognizing humans/faces/bikes/cars 99% of the time is a great algorithm, while an autonomous car is not great if it makes the right decision 99% of the time. You need to be certain of the control inputs provided by the system given the observed environment.

In throwing out MobileEye they have to develop their own computer vision system. This is the neural network they are training. I'm not at all an expert in this field, but I would assume that the neural networks job is to recognize and classify features in the image like humans, bikes etc. The car can then behave differently based on what types of features are in the environment.

Another interesting thing that I think is operating separately is the detection and safety marking of objects with the radar. This sounds a bit like a Simultaneous Localization and Mapping (SLAM) algorithm. Objects are mapped by all cars all the time, and if the driver drives by the object safely it is marked as a safe object to pass (e.g. driving under a bridge).

 

[NOLA_Mike]

Yes, the problem from a profitability/revenu side is that they can't recognise it unless they deliver the goods. Ie, until all of it actually works. And full autonomy is quite the hard binary thing. Either a car is autonomous, or it isn't. So it may be quite a while that they have to carry that deferral. Certainly way after the start of Model 3 deliveries, at which point Model S/X deferred revenue becomes a rounding error. But it does help with cash flow though.

Are you sure about that? Is that how they did it before?

The dual motor cars were introduced along with "Autopilot" in October of 2014. I purchased my car with Autopilot hardware in Dec 2014. They did not activate the Autopilot until Oct. 2015. Did they defer revenue on the option purchase in 2014/2015? I don't remember them doing that...

If they do, then TACC was turned on first in early 2015, do they recognize a part of the revenue at that time?

 

[davecolene0606]

I guess you've never heard about an invention called the Surveillance Camera. You can test how well it works by going to a jewelry store, pretend to steal something and see if the "it wasn't me" excuse will work.

Funny.
I guess you never heard of a condom wrapper that could cover the security camera lense

If a person threw up in that jewelry store you spoke of would they be charged?

Ok. So let's say they video tape you throwing up, having sex, and smoking in the car. How much of a fine would they have to pay for you to be happy with their use of the car?

Would a sick person get charged for sneezing and coughing in your car? For throwing up?

Or a kid coming from soccer practice with muddy cleats? Or a student with a pen sticking out of their back pocket that puts a hole in your seat? Would the camera catch that and how much does that cost them?[/QUOTE]

No, they would be cut off from using the network anymore, bad rating=no rides

 

[myusername]

Thank you Forbes for summing up all my ranting:

“To choose now you’re saying that today’s technology will be the solution five years from now,” said Nidhi Kalra, a senior information scientist at RAND Corp. and co-director of its Center for Decision Making Under Uncertainty. ”The package of sensors you might want five years from is probably going to be very different.”

Tesla's Self-Driving Hardware Gamble: Is 2016 Tech Good Enough For Next Decade?

 

[Seesaw]

Its all fascinating really, and shows how far Tesla is going and how much they are pushing the limits.

I still cannot imagine true autonomous driving without vehicle to vehicle communications - cars telling each other the path they are taking and the moves they are making.... but this in itself may be a simple OTA update away.

 

[Johan Gustafsson]

Thank you Forbes for summing up all my ranting:

“To choose now you’re saying that today’s technology will be the solution five years from now,” said Nidhi Kalra, a senior information scientist at RAND Corp. and co-director of its Center for Decision Making Under Uncertainty. ”The package of sensors you might want five years from is probably going to be very different.”

Tesla's Self-Driving Hardware Gamble: Is 2016 Tech Good Enough For Next Decade?

Give me a break......
Haha - Do you really think that there will be more advanced sensors in 5 years..... I am sure the with the speed Tesla is moving I am convinced that Tesla will put Lidar or other types of senosors if they are better/ have an affordable price. I would not be surprised if Tesla would change the ultrasonic sensors to Lidar in 2-4 years when they have come down i price.

 

[Seesaw]

Does anyone know details about the autonomous video. Did the car have to 'learn that route'? or could any address be entered and the car could find a way there?

 

[brantse]

No, they would be cut off from using the network anymore, bad rating=no rides

Not only would they be cut off from the network, but the network's insurance would pay to cover any damage to the owner's vehicle. Although this will never be for everybody, the general idea of it is that personally owning a very expensive asset that is only used 5% of the time is so economically inefficient that the margins will easily cover any of the "what if scenarios", all while generating a hefty profit.

 

[drinkerofkoolaid]

Thank you Forbes for summing up all my ranting:

“To choose now you’re saying that today’s technology will be the solution five years from now,” said Nidhi Kalra, a senior information scientist at RAND Corp. and co-director of its Center for Decision Making Under Uncertainty. ”The package of sensors you might want five years from is probably going to be very different.”

Tesla's Self-Driving Hardware Gamble: Is 2016 Tech Good Enough For Next Decade?

If an autonomous vehicle makes a mistake, the log and cameras will clearly show what went wrong. The computer will remember the mistake, calculate millions of simulations of the event to ensure the error doesn't happen again in an extremely short period of time, and the probability of the same error happening again is extremely low.

If a person makes a mistake, it isn't always 100% clear who was at fault, or precisely what went wrong. A person might learn from one error, but one error probably has very little impact on the probability of the same accident happening again. People aren't capable of running millions of simulations of what went wrong to significantly reduce the probability of the same error or a similar error from happening again.

 

[Tslynk67]

This is vandalism, and there are 8 cameras to record the vandal.

Sure, but they'll have sole disguise and casually cover the forward facing cameras while the car is stopped at a red light. The car is then immobilised and nothing can be done.

 

[myusername]

If an autonomous vehicle makes a mistake, the log and cameras will clearly show what went wrong. The computer will remember the mistake, calculate millions of simulations of the event to ensure the error doesn't happen again in an extremely short period of time, and the probability of the same error happening again is extremely low.

If a person makes a mistake, it isn't always 100% clear who was at fault, or precisely what went wrong. A person might learn from one error, but one error probably has very little impact on the probability of the same accident happening again. People aren't capable of running millions of simulations of what went wrong to significantly reduce the probability of the same error or a similar error from happening again.

but the probability of the exact same accident happening twice are exactly ZERO... even if it happened 10 minutes after the previous one in the exact same location with the exact same actors... the variables are enormous and start with position of the sun... cloud cover... wind speed... and the list goes on and on.

there is a bump in the road on my commute... it's on a curve on an interstate that is usually packed with cars... when hitting this bump, most vehicles usually shift to the left a bit as steering is momentarily (in the realm of 10ms) less effective. over time i've developed a strategy for this bump... one... drive slightly right into the bump so that when I shift left I end up straight... two... slow down if there's lots of cars that will hit it at the same time... three... if I see a person in a car next to me that is clearly not familiar with the bump I allow my car to drift to the left on the edge or slightly outside the yellow line since I'm frequently in the "fast" lane (another defensive tactic for this bump)...

now try to program a computer to learn that.

i'm not saying it can't be done... but I am saying that someone coming out and saying "AI box checked... now let go of the wheel!"... I will question for many years... I expect we will not be seeing level 4 or level 5 until 2025.

 

[Tslynk67]

Thank you Forbes for summing up all my ranting:

“To choose now you’re saying that today’s technology will be the solution five years from now,” said Nidhi Kalra, a senior information scientist at RAND Corp. and co-director of its Center for Decision Making Under Uncertainty. ”The package of sensors you might want five years from is probably going to be very different.”

Tesla's Self-Driving Hardware Gamble: Is 2016 Tech Good Enough For Next Decade?

Hi myusername

Firstly I just want to say that I appreciate your posts. I don't agree with you most of the time as I'm very bullish on TSLA, but I do like to consider your counter-opinions. I can perhaps appreciate that other members of the forum that have been here longer than I, may have heard it all before, hence their more negative reaction.

Secondly, and more to the point, of course the sensors in 5 years will be better, but Tesla aren't done yet, are they? I've observed since the debut of the Model S that it's in a state of continual improvement - much to purchasers dismay at times - and I fully expect the sensors to evolve likewise. On this basis, Nidhi's point is, well pointless.

 

[drinkerofkoolaid]

Sure, but they'll have sole disguise and casually cover the forward facing cameras while the car is stopped at a red light. The car is then immobilised and nothing can be done.

I can't tell if you're being sarcastic.

If a person were to attempt to steal a Tesla in Autonomous Mode, how would they prevent Tesla or the driver from knowing the car was broken into?

Just to be on the safe side. Tesla should add the following feature:

If a person manages to somehow deactivate Autonomous Mode, and make the vehicle work by bypassing every conceivable system, Tesla and the driver are immediately notified, the doors automatically close and lock, and the police are immediately called. The same should happen if their is any attempt to force changes to any software or hardware while the vehicle is in Autonomous mode.

Heck, this is a feature that could probably be implemented today, to some degree.

 

[techmaven]

IMO, the iPhone analogy is not quite correct. iPhone was ready, and there was no hurdle to use it if anyone wanted. But self driving cars raise safety concerns, not just for those using it but also for others on the roadways. So, everyone weighs in.

According to David Kelly, former NHTSA administrator, regulations won't be ready for 5 years. Then who will pay for self driving feature, when neither the software nor the regulations are ready?

Regulations may tie up Tesla's Autopilot goals

View attachment 199424

First, the original iPhone launch did not have the App Store... that was one year later. While the original launch was hailed as the a landmark moment at the time, it wasn't clear just how much until 3 years later and the collapsing trajectories of Nokia and Blackberry became readily apparent to even casual observers.

2nd, the beauty of Tesla's approach is flexibility. Switching on full autonomy mode is not likely to be a binary event. Instead, it is more likely a series of releases combined with a series of use cases and usage areas where fully autonomous is allowed. The problem with the approaches where it is a binary event is that full 100% autonomy in all cases could take quite a while.