Probably the learning will happen on the cloud.

Probably position, false radar detections, images(video), extracted lane markings, traffic sign positions, landmark positions etc.

They get data. Data is useful for machine learning.

Compared to Waymo they get data from more users. This will help them keep their maps up to date faster, give them more data from accidents and driver intervention and I assume they have a lot of more reasons for wanting data.

They use wifi for uploading this. Bandwidth may be a problem for some user but for plenty of their users bandwidth is no problem.

They have camera feed. It is unclear if they will try to learn the decision behaviour from human drivers. It is possible that they will just use that kind of data for validation and not for learning.

This has been a fun read. So where do people ITT actually think tesla's value is headed? Are the detractors suggesting they'll be virtually gone in the next 5 years? What are the supporters predicting?

Production of the Model 3 has started, here is #1:

Yeah, wondering that myself. TS is not up to speed on machine learning, which is going to be HUGE. But, not sure how far along other automakers are. They can definitely learn from cloud data with or without human drivers.

Sell off to Google

As a supporter I see 20k cars/month at the end of the year and about double that at the same time next year. I see about 2M cars/year (Model Y being the best selling of them) in 5 years with a 10x valuation. There are some unknowns and a high probability that something will go wrong, but also some probability of an in between 0x and 10x valuation.

Imo demand for Model 3 should remain pretty high as the car in many ways is a cheaper(TCO), better(0-100km/h, lower cog, more silent etc) and more environmentally friendly car than the best selling cars in many markets as of today. Demand should pick up as more and more people get to ride in their friends' Model 3s.

Disclaimer: I own some TSLA and a Model 3 reservation.

what do you mean by 10x valuation?
10x revenue? 10x earnings? 10x what it's at now (50b)?

Ya. Not saying present value today is 10x, just that if not anything unexpected happens, which something likely will, that's the valuation I see.


From TMC:

ponzi. gtfo with this musk propaganda fanboy clown.

confirmed Enron...

so a 500b $ company by 2022?

We know that demand for the Model S was so great they put it on sale. I'm sure the Model 3 will have the same demand that far outstrips supply they will lower the price!

Cadillac Super Cruise autopilot is coming in a couple of months.

Sounds a lot more competent and safe than what Tesla is doing.

I was an engineer working with maps for SDCs, my master thesis was about mapping. Did some presentation about what Tesla was doing and have been part of a team designing the map strategy for one of the major car manufacturers. That quote has a lot of inaccuracies, it was likely written by a journalist and not a quote from the company. From the text it doesn't seem as they are doing anything that TomTom, Here, Mobileye, Alphabet, Tesla etc are not already doing.

Fwiw, Tesla has a map and use GPS. So does pretty much every Navi system out there.

Tesla users are complaining about this speed adjustment, see for example 1min into the video and the rest of the video.

Heltok could u specify the inaccuracies?

Ok, lets do it sentence by sentence.

Tesla also relies heavily on sonar to detect vehicles and curbs. As for what they store in their maps it is not exactly known afaik, but I assume they store lane markings, but they don't seem to be able to use it well yet. GPS and digital mapping is very standard for navigation, using it for control is also pretty standard.

Sure. But TomTom and Here have HD-maps for with much more coverage than this. Tesla gathers data from their vehicles, so it might be more stochastic where they have maps. I think there are pretty large similarities between Teslas maps and Mobileye's maps.

Fwiw, here is how TomTom maps looks like, pretty neat right?

They store the HD map as a digital image attached to the center spline. I think including traffic signs, lane markings and HD content they were down to like ~50kB/km.

Google use a bitmap of the infrared reflectivity of the ground for positioning in the lane. But they also have a few different layers. As for Mobileye (and likely similarly Tesla) they are doing like this:
https://youtu.be/b_lBL2yhU5A?t=914

This is just an arbitrary number. We used to call this electronic horizon, I didn't like the term, but ok. As for Operational Horizon, which is used for control, you don't really need that long horizon.

Smooth ride was not our priority, not killing the driver was our. By being very centered and knowing exactly where in the map you are, you have a few extra seconds before you leave the lane if you go blind for some reason. A few seconds to lower the speed before a fontal collision. I agree that Teslas enhanced autopilot is not great at following lane markings today. The exact reason for this I don't know, but I assume they are working on improving this.

Autopilot does this already as the video showed.

Good! =)

Fwiw, I think with Karpathy as the head of the Autopilot project I assume we will see more DL in their maps. Soon humans might not be able to understand what kind of maps they use, only verify that they work in simulation and actual driving.


Seems like the journalist really likes to compare to Tesla Autopilot. But at least be accurate if you do that.

OK, well you obviously don't work for Tesla.
Yep, confirmed you don't work for Tesla. I posted a list of fatal mistakes that Tesla makes a couple of pages ago. They clearly don't do this.
What exactly is inaccurate except for the minor part about sonar? You seem upset that they're pointing out the ways in which Cruise is better than Tesla autopilot (which it obviously is - they're far ahead). Do you at least admit the possibility that Cruise mapping is superior? Tesla's use of mapping is horrible, we know that for a fact. Lane swerving, taking off ramps it shouldn't, relying completely on the camera and having it all go to a dangerous swerving mess if the camera is blinded, etc.

The sonar part is lol. It isn't in AP. That is quite a bit of pedantic neckbearding and poor reading comprehension to call that an "error".

No one mentioning Audi winning the Level 3 war (allegedly)?

ok a bit weird calling yourself a Musk cult member

I didn't really read that comment but it's absurd now that you quote it.

If Musk achieves 33% of what he claims, he is bankrupt. If Musk achieves 66% of what he claims, he is bankrupt.

Musk will need to achieve 100% of what he claims to equal the valuation of Ford in 10 years (about -20% from where they are now - -40% when you include Ford's dividends you're missing out on)

I agree with you for the most part. I don't, however, see Musk as a sinister figure or con man. He seems like an ambitious guy with relatively good intentions.

As a company Tesla is:
-in a shaky industry
-running on hype not results
-without a clear path to even achieve this bloated valuation imo

As a stock holder I am happy with results so far:


Path forward is:
Go from 0->5000 Model 3/week in December. Go from 5000->10000/week December 2018. Introduce Semi and make some semis. Introduce Model Y later this year, start production 2019, make 1M Model Y in 2020, 2M 2021. Get 20-30% gross margin on the cars. Start selling roofs in 2017, increase production over time. Sell stationary batteries, growth rate double that of vehicles.

According to my math if they manage this they should reach current valuation with good margins.

We can argue if they will be able to get there or not. But imo at least they have a pretty clear path in front of them. Do we have any similar plans for their competition? For VW, Toyota, Waymo etc?

guess those three will just keep on printing money and not diluting their shareholders. ¯\_(ツ)_/¯
but if you're happy, i guess everything's fine.

When do you see them making consistent positive earnings and what does their stock price look like at this point in time??

Thoughts on this? Seems bizarre, frankly. This is the kind of thing that scammy penny companies do: get big name unrelated people/former CEOs/politicians on their boards. How the **** do these appointments help them run the company better or make cars?

Interesting news but doesnt seem alarming to me. I think you often see members appointed with general experience. It's probably good that they have people less directly involved with Tesla on the board. It still doesn't make them worth anywhere near 320 a share though.