JRP3
Hyperactive Member
Curious as to which part of that comment @Artful Dodger finds funny? I like a good laugh as much as the next guy so please share.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable
- Thread starter AudubonB
- Start date
StealthP3D
Well-Known Member
Good analysis! And it's not just the risk of collision, this is how radar was previously used to smooth out speed control. For example, the immediate and direct measuring of speed of the car in front allowed AutoPilot to have really good speed control when in highway traffic. If Musk is saying they are no longer using it, even for this purpose, then it reflects just how good (sensitive)Tesla's computer vision intelligence has become. I say that because the sensitivity of the system to the react instantly to the car in front changing speed slightly was already better than a skilled driver would typically exhibit and that ability reduced the gap that would form in such situations and also the need to "yo-yo" in many common traffic situations where even a skilled human would tend to yo-yo a bit or allow a larger gap to form. If Tesla has stopped using radar for this purpose, I would assume the cameras can do just as good of a job or it would be an obvious step backward. This assumes Tesla has indeed stopped using radar for this purpose which is not entirely clear.
It would be interesting to measure the current draw of the radar through a driving cycle including the use of AP, FSD Beta, cruise control, etc. I'm sure Tesla would not waste the energy transmitting if it were not actively monitoring the radar data for some purposes. I'm guessing what we would learn is the radar is still used for many safety functions even if it is no longer used for primary FSD Beta behavior. I can see the radar data being migrated to it's own neural net that has some limited interaction with the primary neural net in extreme situations.
I'm confident Tesla would only make this change, whatever the specific change actually is, if the goal was to make the system even better and with less chance for injury/death. A common narrative in the media is that Elon plays fast and loose with safety, which endangers his customers, other motorists and his astronauts, crew members and residents around his launch and landing facilities. But I don't see that the data supports this false narrative and think most people here can see right through it for the disparaging FUD that it is.
willow_hiller
Well-Known Member
In case this is useful, we do know exactly what the current in-cabin-camera neural network outputs are:
BLINDED
DARK
EYES_CLOSED
EYES_DOWN
EYES_NOMINAL
EYES_UP
HEAD_DOWN
HEAD_TRUNC
LOOKING_LEFT
LOOKING_RIGHT
PHONE_USE
SUNGLASSES_EYES_LIKELY_NOMINAL
SUNGLASSES_LIKELY_EYES_DOWN
From What Does The Tesla Selfie Cam Capture?
As you said, dark environments, sunglasses, and tall drivers (hopefully what HEAD_TRUNC measures) are all limitations. But given the system is aware of its own limitations, it could be a hybrid driver-attentiveness monitoring system: where it relies on the camera when possible, and falls back to steering-wheel torque when not.
Buckminster
Well-Known Member
Buckminster
Well-Known Member
Artful Dodger
"Neko no me"
Good points. The obvious solution is 4 x wheel motors (eliminated 4 half-shafts, 2 differentials). But not like those clunky, heavy wheel motors that Protean/F-150 teased at SEMA in 2008 (and never produced). More like this:
- wheels integral with the motor as a unit:
- steel wheel takes on a dual role as tyre rim and SRPM rotor
- mass of brake rotor removed to compensate for integrated motor mass
- brake disk caliper replaced with stator (could be hubless, airless design)
- tyres are replaceable rubber tread inserts for million-mile service life
- friction-free regen/active braking is safe and practical for Robotaxi:
- robotaxi will NEVER drive with a cold battery pack (depot charging b4 departure)
- even large diesel-electric locomotives rely 100% on regen braking w/o a bty pack
- minimal mass of robotaxi and LFP tolerance to overcharging means no issues
- active breaking can use electrical power for deceleration in emergencies
- S3XY cars which start driving while cold-soaked will retain friction brakes
- this hypothetical Robotaxi is NOT the Model 2 coming in 2022/23, its the Model 1 in 2025/26
Last edited:
Johan
Ex got M3 in the divorce, waiting for EU Model Y!
Yeah, still remember their Superbowl commercial. If you look at what the actual message was, beside all the Will Ferrel and Norway stuff, it was basically: we will have good electric cars in the future. That is really telling, I mean you can go to investors for example and try to raise money for something you don't yet have but you are planning, but I think I've never seen a company advertise to consumers for a non-existent product. Oh the desperation.
Apparently, the NTSB is complaining to the NHTSA that they aren't regulating Tesla hard enough.
Wouldn't surprise me to learn that Musk doesn't want to make electric planes because he wants to avoid the NTSB at all costs.
A federal agency warns Tesla tests unfinished driverless tech on its users
In a letter to the NHTSA last month, NTSB chief Robert Sumwalt named Tesla 16 times while urging the sister agency to set stricter standards.www.cnbc.com
Right, not only the media reporting seems biased against Tesla here, there are also problems with the NTSB letter,
They single out Tesla and refer to three fatal accidents, tragedies that are probably well known to TMC-members. These three accidents are however two, three and five years old now, so the cars that actually have the FSD hardware and FSDBeta software have nothing to do with these accidents. Further, per the NTSB's own investigation these fatalities have several contributing causes: two under-rides where the truck driver failed to yield and one where a crash-attenuator had been missing for about ten days due to a prior (non-fatal) accident. There is a debate, whether Tesla's FSD increases issues of driver awareness, but regardless of that contributing to all three fatalities was also the (unlawful) missing awareness on the part of the Tesla driver.
Lastly, NTSB's letter does not mention Tesla's published statistics on how Teslas (whether Autopilot is activated or not) are involved in fewer accidents,
Tesla Vehicle Safety Report
PS. Copy-edited.
It occurred to me that the power saving from not using radar might allow for/justify overclocking FSD 3.0 HW.
StealthP3D
Well-Known Member
It's a fool's errand to try to coordinate share price movements with events such as these. Especially when they occur over a period of time and are not a sudden reaction upon the news. Even that is full of potential to misattribute. The only accurate way to provide reason to share price movements is to only attribute them to the general, ie. investor sentiment (or trading bots). Sometimes the people who run the trading bots don't even know why they do what they do. Because AI. We don't know why investor sentiment changed. There are multiple potential triggers, some of those triggers are even unknowable, and we don't know how they layer and interact. Trust me, it's MUCH more complicated and unknowable than you seem to think because humans are complicated, even in the aggregate. It could be that without the bitcoin investment prices would have been even lower. Then there is the complication of initial reaction vs. the reaction over time as the news has had a chance to sink in.
Assuming you know (as a fact) what drives investor sentiment at any given moment is counterproductive and even dangerous.
StealthP3D
Well-Known Member
I seriously doubt Tesla weighs power-saving that way. If they need to use more power to make it safer, they will. The differences in power consumption would be minor and it's not an "either-or" kind of decision.
Tesla Giga Shanghai is Fine-Tuning Manufacturing to Meet Demand Growth, According to Company Statement
Tesla plans to make changes to manufacturing processes at its Chinese plant to further expand production capacity at Giga Shanghai to meet an ever-growing demand. The changes relate to improvements in the production of various vehicle components.
www.tesmanian.com
Tesla plans to make changes to manufacturing processes at its Chinese plant to further expand production capacity at Giga Shanghai in an attempt to meet ever-growing demand. The changes relate to improvements in the production of various vehicle components.
...
According to Tesla's statement on the Shanghai government website, the changes to the Chinese factory will relate to the production of spare parts, including plans to improve the power system, chassis, and electric motor. The company also confirmed that these changes are taking place in response to growing demand.
That camera doesn't need to see through sunglasses. It only needs to see if the driver is looking at a phone or laptop or girlfriend or the ceiling because he's asleep. I suspect the camera is great at that. Good engineering is about optimization: don't spend money on more camera than is needed.
You seem eager to do verbal battle against overzealous Tesla fans, but one can get overzealous in that quest.
StealthP3D
Well-Known Member
The reason you've never seen a company advertise a non-existent product is because companies don't do that. They spend money to make money. What does that tell us? That they were trying to sell something else. Specifically, their gas cars (instead of their competitors electric cars). They are terrified of the mad rush to EV's when they don't have a volume solution.
What that ad was really saying was; Norwegians are different from Americans. You don't drive an EV because you are an American. A lot of Norwegians do that but not many Americans. You want to live like the American you are. It's OK to treat yourself to another gas car, the EV's will be even better in a few years. As an American it's OK to do this. It's normal. Buying an EV would only be normal if you lived in a foreign country.
Someone recently said I would never make it in marketing. That's because I understand it too well (and don't like what I see).
Since the dual-use ultrasonic sensors are not mentioned and can thus be assumed to remain in use by the FSD, I guess the motivation for removing the radar is cost - and possibly the risk of interference once many of these cars drive near each other.
Tesla like software solutions to a lot of problems. So maybe they will come up with a protocol through which trusted vehicles (i.e. other Teslas) on the road nearby can exchange information on the traffic conditions (assuming Internet is available). So when a Tesla slams on the brakes it could broadcast a warning, so for example another Tesla a few cars behind it can slow down. With a higher fraction of Teslas on the roads, this could to some extent compensate for the loss of the radar. Since this is basically a generalization of platooning, I would give it fair odds of happening.
Johan
Ex got M3 in the divorce, waiting for EU Model Y!
The reason you've never seen a company advertise a non-existent product is because companies don't do that. They spend money to make money. What does that tell us? That they were trying to sell something else. Specifically, their gas cars (instead of their competitors electric cars). They are terrified of the mad rush to EV's when they don't have a volume solution.
What that ad was really saying was; Norwegians are different from Americans. You don't drive an EV because you are an American. A lot of Norwegians do that but not many Americans. You want to live like the American you are. It's OK to treat yourself to another gas car, the EV's will be even better in a few years. As an American it's OK to do this. It's normal. Buying an EV would only be normal if you lived in a foreign country.
Someone recently said I would never make it in marketing. That's because I understand it too well (and don't like what I see).
That's a very good take on it. However it is not true that that was the commercial I watched.
It starts with Will Ferrel saying "Did you know that Norway sells more electric cars per capita than the US"? And then crushes the in front of him and says "Well I won't stand for it!".
Cut to him getting in a GM EV (the upcoming 2023 Cadillac Lyriq) and saying "With GM's new Ultium battery we're gonna crush those losers".
Cut to him going to his friend, saying "Norway is out EV-ing us" and "Grab an EV, meet me in Norway". Then he picks up another friend saying exactly the same thing.
Cut. "GM's Ultium battery is made for all types of vehicles, so soon everyone can drive an EV".
He arrives in Norway (he thinks, turns out he's in Sweden). His friends are in the 2022 GM Hummer EV, apparently they drove to Finland instead.
Cut to text:
Next screen:
So: They are advertising only the following: A 2022 concept car, a 2023 concept car, "30 new EVs by 2025" and the Ultium battery. Nothing anyone can get today.
petit_bateau
Active Member
When the batteries are fully charged regen braking needs somewhere to dump the energy into. If it gets dumped into the motors then the Curie limit is extremely rapidly reached on the neo magnets (which is terminal) and/or bad things happen to any other part of the motor due to the heat. In my experience just a few high speed braking events in some circumstances could cause those thermal limits to be reached (to be fair this would likely be more of an issue in some forms of highway driving). So @JRP3 is quite correct to point out that it is likely the friction brakes will likely be retained. I would however expect that one set of pads will last a few hundred thousand kilometres. I suppose it is possible they might be able to eliminate friction brakes provided the vehicle never exceeds a "one-time-max-regen-brake" speed and then goes into a "limp-until-cooled down" mode. I wonder what the regulators would say about that, or the FUDsters.
Tesla designed the hardware themselves, so they would never actually overclock it. Apart from the performance being designed into the solution at the actually used clock, there is also the "Thermal Design Power" i.e. the maximum amount of cooling the hardware needs, they would not want to go beyond that.
Under certain circumstances the hardware may very well be able to run with a reduced clock (to save power), but to go beyond what Tesla themselves designed the hardware to be reliably capable of would make little sense.
Knightshade
Well-Known Member
It can't do that if it can't see through sunglasses.
You can easily have your head facing forward while still having your eyes looking at things other than the road... (or heck, closed for that matter).
That's why "sunglasses" are the only states the camera software uses "likely" for- it can't tell for sure so It's guessing based on your head position.
Other companies using in-camera driver attention monitoring use additional hardware to see where your eyes are actually looking instead of guessing like that.
Your suspicion is not based in fact though. It's not great at it, because it lacks the hardware to be great at it.
If it was great at it it wouldn't have "dark" and "blinded" states, and wouldn't have "likely" qualifying anything it thinks about if you have sunglasses on- it'd have the HW to handle those conditions like other companies systems.
Again it's not a case of "tesla picked bad hardware"
it's a case of "Tesla never intended to use the hardware this way when picked and they're trying to tack on an unintended function so it's not great at it"
it certainly does a much better than nothing job though.
It's baffling folks feel the need to defend the hardware when it's still doing more than originally intended, even if it's not ideal at the job.
I don't think that's the case at all.
We know the camera wasn't intended for this use in the first place. Elon told us years ago what it was actually there for (watching robotaxi passengers)
The fact they're repurposing it for another job, and it can do a better-than-nothing job of it, is great.
But it's simply a fact it's less capable than the dedicated meant-to-do-it-from-design-phase driver monitoring camera systems other car makers are using.
Again I'm frankly baffled there even is an argument to have on that- it's pretty obvious from the specs and, ya know, how cameras and light work.
Now- will the limits of the interior camera for this purpose matter long-term?
Again- that's primarily a question of if regulators make it matter.... (or if Tesla gets to L4 really really fast, bypassing the need to monitor at all).
So far Tesla in the US anyway has gotten away with the steering wheel system for driver attention at L2.
There's currently no regulatory reason they'd be suddenly using the camera for anything- nothing's changed in the law right now.
So if they're using it now anyway there's 2 possible reasons:
1) They're seeing how limited or capable it is in real life situations, because they're concerned about what regulators might require in an L3 vehicle (or even an L2 one if the NTSB had its way) and want to know how "good enough" it really is in that context.
and/or
2) FSD is in the dangerous phase many have mentioned before, where it's so good that it gets too easy to trust it even when you still need to be paying attention to it, so they want a backup to the not-great-at-really-checking-engagement wheel torque system.
Pezpunk
Active Member
It has absolutely NOT been debunked. there is a huge amount of evidence that it is terrible for the environment on a scale that absolutely DWARFS the rest of the financial industry.
There is a tiny amount of speculation that "hey maybe it's not so bad, or won't be so bad someday?!?!"
These two things are not equal. stop believing things just because you want them to be true. Stop allowing weak arguments to muddy the water versus strong ones.
Last edited:
Similar threads
- Locked
- Locked
- Locked
