Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable

[JRP3]

So here's a weird one, electrical contractor claims he was fired from GF for reporting thefts of copper wire. Security actually called local law enforcement, arrests were made, but charges were dropped.
Ousted Gigafactory worker alleges Tesla management pushed him out for reporting theft

While on the job, Thompson noticed what he estimated to be copper wire worth millions of dollars being cut and loaded onto pallets.

The wire was then loaded onto trucks and hauled away, with union contractors saying the copper had been stolen, the complaint stated. It also stated non-union contractors hadn't reported being victims of theft.

Thompson said he reported the incident to management and was told to stop investigating.

"Somebody needs to make sure they are held accountable," said Thompson.

On June 6, according to the complaint, Thompson again saw people loading copper wire onto a truck and reported it to security, which called local law enforcement.

Deputies from the Storey County Sheriff’s Office arrived and arrested multiple individuals, the complaint states. They also required Thompson to fill out a witness statement, which he was reluctant to do for fear of reprisal.

The Storey County Sheriff’s Department confirmed the arrest of at least one person from that incident. Records from the Storey County District Attorney also show the charges were dismissed before they could go to trial. District Attorney Anne M. Langer didn't return a request for comment on Friday.

The morning after the incident, Thompson said he was told by “senior management,” he was no longer allowed on the property.

 

[Steve m]

The way lidar mapping works is that it subtracts the stored map from the current lidar measurement results. So any change compared to the map stands out. In fact it is very easy to detect a fridge or tree with a lidar and more reliable than cameras even without having the map stored.

Or a or a tire fragment or a tennis shoe or anyother object that randomly falls onto the road but isn't a significant hazard.

 

[S3XY]

Welp the FUD has Peak Stupid

The Insane Shortsightedness of Buying a Tesla

 

[Doggydogworld]

What makes you think margins will increase while they are ramping GF3?

We don't know enough about GF3 financial reporting to know how it will affect margins. It's possible there will be no net fixed depreciation and only ~20m/quarter direct labor at first. That would barely move the overall Auto gross margin needle.

 

[M3Rider]

So here's a weird one, electrical contractor claims he was fired from GF for reporting thefts of copper wire. Security actually called local law enforcement, arrests were made, but charges were dropped.
Ousted Gigafactory worker alleges Tesla management pushed him out for reporting theft

Sounds like the night shift management organizes the theft, tells him to shut up, and then fires. How were the charges dismissed then? Same management said nothing is missing?
Interesting what's going on there, since allegedly this is not a first instance. We probably not going to hear much about this.

 

[electracity]

So here's a weird one, electrical contractor claims he was fired from GF for reporting thefts of copper wire. Security actually called local law enforcement, arrests were made, but charges were dropped.
Ousted Gigafactory worker alleges Tesla management pushed him out for reporting theft

Obviously employees confused about Elon's wiring reduction program.

 

[StealthP3D]

At this stage it still looks relatively unlikely, yet possible, that Tesla delivers a feature complete unsupervised FSD 5-10x safer than humans within the next two years. They have made huge progress on the march of 9s on Navigate on Autopilot just with HW2. The real test will be just how much of a step change improvement they can achieve when they start making use of HW3 and the larger neural nets. In terms of regulation, I think Tesla will receive approval in at least some jurisdictions fairly soon after getting unsupervised safety to 5-10x.

Why do you think it needs to be 5-10X safer than a human to gain approval from regulators?

There are 1.2 million auto fatalities annually around the world. That's an astonishing 3,287 people each day. If autonomous driving was only 2X as safe it would save over 1600 lives every day! That's over 600,000 lives of real people, maybe your mother or child, every year! If regulators waited until it was 5-10X as safe, they would be guilty of homicide by gross negligence on a *massive* scale.

Autonomous driving should be approved as soon as it can help reduce the highway carnage that is currently happening.

 

[S3XY]

The above posts talk about wrong point.

Does it make sense for Tesla to help competitors from the financial standpoint? None whatsoever. If you want to maximize profits, kill them and get as much market share as you can.

But EM talked about the mission. Here the issue is that Tesla may not ramp up fast enough to make as much impact as early as possible as they would like. Just can't spend money fast enough and scale.

*** Snip ***

Wipers actually had some lower scores on the Bloomberg's survey. I also don't like Auto.
I think they are just neglected temporarily, b/c more value can be achieved elsewhere.

"Deep Rain" is coming.

 

[FactsBot]

Re, disengagement reports: most people mistakenly think that this means that "every time the driver takes the wheel, you have to issue a disengagement report. That's not how it works.

California’s self-driving car reports are imperfect, but they’re better than nothing

A disengagement report only needs to be filed if the human driver decides that the reason that they took control of the wheel was due to "safety" or a "failure". Driver takes control to avoid missing a turn? To not block an intersection? To not annoy other drivers? None of those things count as a disengagement. And those sorts of things are the vast majority of the reason why people take control over from Autopilot as well.

Also, while Waymo "tests" in a number of cities (including a winter proving grounds in Michigan), notice how they only have public operations in places that never get snow? Yeah, because this:

... will render LIDAR basically useless for detecting road obstacle detections. Forget about "falling snow", ridges of snow on the road are indistinguishable from tree braches or children laying across the road. And it's like that all winter in northern climates.

Doesn't just change the surface of the road with random ridges and troughs and blocks... it also changes the sides of your road.

Medians may widen and go windy. Shoulders frequently disappear, and may intrude into the road. Parking spaces may turn into huge snowpiles. Your landmarks on the edge of the road may be totally transformed.

It's not just about how to deal with driving on slippery roads and how to recognize changes in how slippery a surface may be, how to navigate to less slippery parts, etc etc. That's an entirely separate, and critical, issue. No, it's not about how falling or blowing snow blocks LIDAR. Again, that's yet another issue. No, even before you get to that point, when it comes to snow accumulation, to LIDAR, it's all an obstacle. It has no idea how to tell snow accumulation apart from anything else.

All lidar solutions have cameras too. Lidar just adds to the safety. Cameras struggle with low contrast in winter driving. This is a very difficult challenge and I expect it won't be solved in the next 5 years. WaveSense provides a solution on how to find lanes on a snow covered road.
WaveSense
which however involves mapping the ground first but it could work in cities like Boston

There are new filtering techniques that make lidar work in fog, rain or snow. MIT has shown it in lab, see the video below, but private companies are using it for a while.

 

[Doggydogworld]

Alrighty then, since nobody has stepped up with a estimate, I used Google Maps and my best guess for the end points of a segment of the back straight, measured at 1.29 km. Then in the Youtube video, that segment took 16 seconds for the Blue S Plaid Prototype.

Doing the math, that's approx. 295 kmh, or 180 mph. Yowsah. That'll work.

Taycan took a tick under 16 seconds for that same stretch. It started at 240 kph and drove most of it in the high 250s.

My Google Map shows 1.1 km from gantry to the tourist exit ramp, which the Model S Plaid reaches a fraction before the clip ends. In 16 seconds that'd be 248 kph or 154 mph. I get about 16.6 seconds, so a bit under 150 mph.

 

[EVNow]

WaveSense provides a solution on how to find lanes on a snow covered road.
WaveSense
which however involves mapping the ground first but it could work in cities like Boston

This was a specific question to Karpathy. He said it's actually fairly simple for NN. If you can annotate the lane on images, you can train.

 

[Doggydogworld]

Why do you think it needs to be 5-10X safer than a human to gain approval from regulators?

There are 1.2 million auto fatalities annually around the world. That's an astonishing 3,287 people each day. If autonomous driving was only 2X as safe it would save over 1600 lives every day! That's over 600,000 lives of real people, maybe your mother or child, every year! If regulators waited until it was 5-10X as safe, they would be guilty of homicide by gross negligence on a *massive* scale.

Autonomous driving should be approved as soon as it can help reduce the highway carnage that is currently happening.

Because people get 10x more upset (and likely to sue) over a loved one being accidentally killed by a corporation than by a person. Maybe 100x. Especially if an alert human would have easily prevented the wreck.

Lawyers aggressively seek out such cases, convincing victims and relatives their suffering deserves lottery-level compensation. Here in TX 46.3%* of all TV ads are lawyers saying "Hit by a company-owned car or truck? Contact Sleazebag & Feeraker to learn your rights!" If you're hit by an individual they have no interest.

Tesla (and Waymo, etc.) don't just need regulatory approval, they need laws to limit their liability. A 2x improvement won't get that. Alternatively they need to be so safe that serious injury and fatality accidents essentially never happen. Waymo seems to be aiming for the latter, while quietly lobbying for the former.

Meanwhile, Waymo invited the President of the Elon Musk fan club to take a ride without a safety driver.

---------
*Not an actual measurement, it just seems that way

 

[Mike Smith]

Tesla's brand alone has to be worth $15B or more

 

[EVNow]

Of course I could be wrong but this is why I am expecting continued margin improvement.

That's why we need to try to quantify the effect of each item.

 

[Antares Nebula]

Well obviously Tesla would sell a complete package including BMS, Batman and Robin ASIC chips and unified Powertrain software. That's the core of their product.

My point is that this would break their design principle of designing the best product by having total control over ALL hardware and software design. Specifically, they would have to commit resources to making sure their software (and hardware) worked within the various competitors' software stacks and with their various hardware (CPU, GPU, etc.), none of which they have control over.

And sure, they could in fact do this to accelerate their mission. But I think what Musk is thinking is "accelerate" by stoking the competition by coming out with great and in demand products, not by making stuff for them. He is not looking to take the Intel, Windows, or Android approach, but more the Apple approach. He is thinking along the lines of Steve Jobs, and adhering to the above design principle. Maybe I'm wrong. Let's see what happens.

 

[jerry33]

Why do you think it needs to be 5-10X safer than a human to gain approval from regulators?

There are 1.2 million auto fatalities annually around the world. That's an astonishing 3,287 people each day. If autonomous driving was only 2X as safe it would save over 1600 lives every day! That's over 600,000 lives of real people, maybe your mother or child, every year! If regulators waited until it was 5-10X as safe, they would be guilty of homicide by gross negligence on a *massive* scale.

Autonomous driving should be approved as soon as it can help reduce the highway carnage that is currently happening.

While it's not right, people will only consider the fatal accidents with FSD, not the lives saved. Similar to seatbelts, even though seatbelt laws have been in place for over 40 years and there is no question that you're safer being belted up, a large minority of people won't use them unless forced to.

 

[mongo]

So here's a weird one, electrical contractor claims he was fired from GF for reporting thefts of copper wire. Security actually called local law enforcement, arrests were made, but charges were dropped.
Ousted Gigafactory worker alleges Tesla management pushed him out for reporting theft

Welp, my quick math says one million dollars worth of 1/0 copper thhn wire weighs 134 tons.
6 AWG is 95 tons. 8 AWG is 93 tons at retail prices.
It would take over 3,000 2,500ft spools of 12AWG to reach one million dollars.

That's a lot of trucks/ pallets.

On the detail presented. Just because police arrest someone doesn't mean they did anything wrong. Conversely, having the charges dropped doesn't mean they didn't.

 

[j6Lpi429@3j]

The way lidar mapping works is that it subtracts the stored map from the current lidar measurement results. So any change compared to the map stands out. In fact it is very easy to detect a fridge or tree with a lidar and more reliable than cameras even without having the map stored.

Thats fab...but then...what? the car has to use a non-mapping based system to navigate its way around the fridge...which means if it can do that *without maps* then the maps are effectively optional anyway, rendering the whole mapping-based system pointless.

Maps cannot possibly keep up with real time changes that happen on roads, like accidents, roadworks, changes to the road, spillages etc. If waymo is dependent 100% on maps, its useless. If it can work without the maps, then the maps are useless. Either way, its definitely not an improvement on a vision-based non-mapped solution.

 

[KarenRei]

All lidar solutions have cameras too.

The issue is that if you've solved the problem for cameras, you've rendered LIDAR pointless. My point is that you must solve for cameras... and once you have, you've solved the problem.

Lidar just adds to the safety. Cameras struggle with low contrast in winter driving.

Our two cameras in our head handle just fine. And as noted above, LIDAR can't tell one iota whether what it's scanning is snow, slush, or a drunk passed out crossing the road. LIDAR sees "obstruction", but has no colour data (and insufficient point cloud density) to determine what exactly it is. Even if it did have more point cloud density, it'd just be endless pareidolia.

WaveSense provides a solution on how to find lanes on a snow covered road.

And is radar, not LIDAR.

There are new filtering techniques that make lidar work in fog, rain or snow.

Rain, snow, fog, etc will always be adding a tremendous amount of noise to your point cloud. Sure, you can try to dig out a signal out of the noise, and there's been a lot of research on that front because it's so important to LIDAR companies, but that's IMHO a waste of research money. It's always going to be a poor, noisy signal.

 

[KarenRei]

Welp, my quick math says one million dollars worth of 1/0 copper thhn wire weighs 134 tons.
6 AWG is 95 tons. 8 AWG is 93 tons at retail prices.
It would take over 3,000 2,500ft spools of 12AWG to reach one million dollars.

That's a lot of trucks/ pallets.

They were really stealthy when they pulled the train up to GF1 to haul the copper off