High res 3D maps don't work for FSD. World changes faster than the map. Just so you know -- that's a dead-end route. Maps can be assistive but the primary system must not rely on them and must assume that they may be wrong.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
- Status
Loved your whole post, and for others, his a real world example of this quote:
Bangladesh leads in clean energy use
It depends also how quickly the 3D maps are updated and can be transferred to the next vehicle traveling the same path.
I've found Waze to be very proficient. If you could automate it and with 10x the participation level I think they could cover and update all drivable surfaces quickly.
RobStark
Well-Known Member
Tesla will announce battery GF 2,3, and possibly 4 this year.
Because even with GF1 at full production they can't meet demand for S3XY plus BES.
Therefore continuing to build batteries at old factories in Japan makes sense. Because even at higher cost it still makes money and furthers Tesla's mission to decarbonize the economy.
They may very well put GF machines in the old factories. After GF1 starts to ramp up .
I don't know how else to explain it to you.
Sucks for that first guy who drives into the river when the bridge is washed out and dies, but the car updates the maps so the next guy who drives there is OK?
Basically, no, this doesn't work as a primary sysem.
Krugerrand
Meow
Small correction. Announcement will be for GF 3 and 4, possibly 5. We already know that Buffalo is GF 2.
Krugerrand
Meow
Somebody is always going to unfortunately die as *we learn*. That's a historical fact and no way around it. The idea is to keep the deaths to a minimum and make sure we don't make the same mistakes again.
I agree that we do not know. I am probably naive, but when Elon makes statements along the lines of "full self driving is way closer than most people think", and "It amazes me, and I am right in the middle of it." (Not exact quotes) I assume that he has the beta system installed in his personal car and is driving around with it every day. I assume that it works really well, but is still too buggy to release to the public.
On the other hand, after reading his biography, and the recent article on the development of the solar roof, I also acknowledge that it is also possible he is completely blowing smoke, and is just very confident that his team will come up with something soon.
On the detailed 3D mapping,I absolutely think that can be the primary system guiding the car, of coarse there will have to be additional layers of back up systems to deal with changes to the roads, traffic, pedestrians, etc.
Well, I realize I'm not going to convince some of you.
You're wrong, though. People won't accept a system which is worse than a human in any way, and they won't accept that first person going into the river in "self-driving" mode.
There's a reason Autopilot is what it is -- a driver assist system where the driver is supposed to be ready to take over in case of weirdness -- and there's a reason that such driver assist systems, constantly improving, but with the human always responsible, are what is going to happen.
Full "no human attention necessary" self-driving in the near future is a fantasy of people who haven't done their research. Attempting to do it with maps as the primary system is fatally flawed; you'll never manage it using maps, because maps are always behind reality. (Which is probably why Tesla is not doing that, and maps are entirely secondary for Tesla's system.)
It's become clear after getting some glimpses at the Tesla corporate culture that there are a lot of those 'did not do their research' people at Tesla with regard to any number of issues, including, sometimes, Elon Musk -- to be fair, he's an extremely busy man, he doesn't have time to research everything.
I didn't understand why they did some appallingly stupid *sugar* early on -- some of which they are still doing -- and it turns out they just didn't bother to do their research. Due diligence failures. It's a risk factor which I am substantially more aware of after visiting the factory. There's a certain sense in which the whole Tesla operation is unprofessional, and that creates certain serious risks, though I currently think those risks will not have a material effect on profitability even in the worst-case scenario.... until the day when all new cars are electric and they have to compete with operations which act professionally, which probably won't be until 2029 or so. They may fix these problems by then. They may not. But in investment terms it's a problem to worry about a decade from now.
You're wrong, though. People won't accept a system which is worse than a human in any way, and they won't accept that first person going into the river in "self-driving" mode.
There's a reason Autopilot is what it is -- a driver assist system where the driver is supposed to be ready to take over in case of weirdness -- and there's a reason that such driver assist systems, constantly improving, but with the human always responsible, are what is going to happen.
Full "no human attention necessary" self-driving in the near future is a fantasy of people who haven't done their research. Attempting to do it with maps as the primary system is fatally flawed; you'll never manage it using maps, because maps are always behind reality. (Which is probably why Tesla is not doing that, and maps are entirely secondary for Tesla's system.)
It's become clear after getting some glimpses at the Tesla corporate culture that there are a lot of those 'did not do their research' people at Tesla with regard to any number of issues, including, sometimes, Elon Musk -- to be fair, he's an extremely busy man, he doesn't have time to research everything.
I didn't understand why they did some appallingly stupid *sugar* early on -- some of which they are still doing -- and it turns out they just didn't bother to do their research. Due diligence failures. It's a risk factor which I am substantially more aware of after visiting the factory. There's a certain sense in which the whole Tesla operation is unprofessional, and that creates certain serious risks, though I currently think those risks will not have a material effect on profitability even in the worst-case scenario.... until the day when all new cars are electric and they have to compete with operations which act professionally, which probably won't be until 2029 or so. They may fix these problems by then. They may not. But in investment terms it's a problem to worry about a decade from now.
Last edited:
Air bag maker Takata to file for bankruptcy this month -sources
REUTERS 7:40 PM ET 6/15/2017
Symbol Last Price Change
HMC 27.89
F 11.24
TSLA 375.34 -5.32 (-1.4%)
TM 105.78
QUOTES AS OF 04:02:03 PM ET 06/15/2017
* Bankruptcy filing may come as early as next week - sources
* Takata has been working for months on deal with Key Safety
* Tokyo bourse suspends trading in Takata shares (Adds share suspension in Tokyo)
By Jessica DiNapoli and David Shepardson
NEW YORK/WASHINGTON, June 15 (Reuters) - Takata Corp , the Japanese company facing billions in liabilities stemming from its defective air bag inflators, is preparing to file for bankruptcy as early as next week as it works toward a deal for financial backing from U.S. auto parts maker Key Safety Systems Inc, sources said on Thursday.
Takata, one of the world's biggest automotive suppliers, has been working for months to complete a deal with Key Safety.
A person briefed on the matter told Reuters Key was expected to acquire Takata assets as part of a restructuring in bankruptcy.
The Nikkei business daily reported that a new company created under Key will purchase Takata operations for about 180 billion yen ($1.6 billion) and continue supplying air bags, seat belts and other products, leaving liabilities behind in a separate entity.
Takata declined to comment. Michigan-based Key, owned by Chinese supplier Ningbo Joyson, did not immediately respond to a request for comment.
The Tokyo Stock Exchange said on Friday it had suspended trading in shares of Takata following media reports of preparations for a bankruptcy filing. At its previous close, it had a market value of $360 million.
Also at stake is $850 million owed by Takata to major global automakers under a settlement agreed to earlier this year stemming from the automotive industry's largest ever safety recall.
Sources familiar with the matter, who asked for anonymity because they were not authorized to speak with the media, said a final deal with Key may not be reached before Takata files for bankruptcy. The company plans to begin proceedings in both the United States and Japan, the sources said.
Major global car manufacturers have expressed concern about the company filing for bankruptcy without a deal in place because of disruption it could cause to the production of replacement air bag inflators. More than 65 percent of 46.2 million recalled Takata air bag inflators in the United States have not been repaired.
A U.S. judge said earlier this year the costs of replacing all of the faulty Takata inflators could be $8 billion.
Takata inflators can explode with excessive force, unleashing metal shrapnel inside cars and trucks.
In January, Takata agreed to plead guilty to criminal wrongdoing and to pay $1 billion to resolve a U.S. Justice Department investigation into ruptures of its air bag inflators linked to at least 16 deaths worldwide. They also have been blamed for more than 180 injuries worldwide.
A federal grand jury in January indicted three former Takata executives for criminal wrongdoing in connection with the safety defect.
Takata has already paid a $25 million fine and $125 million to a victims' compensation fund, including for future incidents.
Takata has until early 2018 to pay the $850 million owed to automakers, or within five days of securing a financial backer. Then-U.S. Attorney Barbara McQuade said in January that if Takata filed for bankruptcy, the Justice Department would be a creditor in the restructuring.
Key was chosen by Takata earlier this year as the favored buyer for the company. Since then, Key has been working with Takata on a restructuring plan.
Inflator recalls began around 2008 and involve around 100 million inflators around the world used in vehicles made by 19 automakers, including Honda Motor Co(HMC), Ford Motor Co(F) , Volkswagen AG and Tesla Inc.(TSLA) Recalls are to continue through at least the end of 2019.
Last month, four automakers including Toyota Motor Corp(TM) agreed to a $553 million settlement to address class-action economic loss claims covering owners of nearly 16 million recalled vehicles with Takata inflators. (Reporting by Jessica DiNapoli in New York and David Shepardson in Washington. Additional reporting by Paul Lienert in Detroit and Taiga Uranaka in Tokyo; Editing by Tom Brown and Cynthia Osterman)
My model 3 reservation was fake. I always figured I would get a new model S or X. I bought an X and withdrew my reservation. So yeah, bearish. /s
Neroden, what did you see on the (very superficial) tour that led you to think that Tesla is risky due to unprofessional stuff/no due diligence failures?
The article had intriguing thesis.
It seemed to equate cell phones and EV's somewhat. Developing countries went straight to cell phones, why put up wires and expensive poles. And seemed to intimate that both EV's and an electrical grid could be similar. Micro and nano grids for electricity and EV's instead of ICE, with no need at all for fuel infrastructure
RobStark
Well-Known Member
Small correction. Announcement will be for GF 3 and 4, possibly 5. We already know that Buffalo is GF 2.
I said battery GF 2,3, and possibly 4.
Specifically to avoid correction
Except for trains and planes--nobody uses those because they use autopilot systems. Oh, and cancer diagnoses--Watson beats real doctors.
Autonomous driving will eventually be king. Especially because humans are also lazy and want instant gratification often, and an autonomous system that's reasonably safe will be used and abused by many people. Because of this group, it'll eventually become culturally.
The fact that AIs can beat humans at Go makes me confident that autonomous vehicles will 100% be better than humans and by incredibly large margins. Like driving, Go has far too many permutations and possible scenarios to use traditional computer problem solving methods--it makes decisions based on incomplete information but does so many times better than any human can. Go learned by playing itself millions upon millions of times. It's primarily a data problem--all you need is powerful enough machine learning (already exists in many areas of the world), and an absolutely incredible amount of data (not yet even close to finished).
This isn't about making 3d maps that dictate where a Tesla can drive--not in the long run. It's about making 3d maps that machine learning can interpret and understand trillions of times over. This machine learning is eventually what will tell a Tesla how to drive--the only 3d map the car needs is the one it's generating in real-time. Like humans, cars will only need to fall back to these stored maps in extraordinary conditions (i.e., how humans navigate roads after a blizzard today). They may still utilize some useful information, such as signs, intersections, lanes, navigation data--but they will primarily be using their own real-time 3d map to drive and will almost always use their real-time map to make decisions.
You very much misunderstand why Tesla needs these maps.
Strictly color. Attitude. Corporate culture. Stuff that's hard to pin down or describe; it's just the *attitude* coming off the employees at the tour and the annual meeting. An attitude combining naivete with wild optimism and hubris. This attitude has its upside, but it also has a downside, and I wasn't seeing hardly anyone with the attitude necessary to counteract that downside.
It explained the psychology which led to some long-standing idiotic decisions at Tesla which I have mentioned in other places (including pointlessly pirating software because they didn't bother to read the licenses of the software they're using, not bothering to look up the state registration & insurance laws before selling cars in each state, having weak software development protocols which led to serious regressions being deployed repeatedly, having no functioning system for escalating repeated customer complaints to the higher-ups, etc.) They're going to make the same sort of mistakes again. And again. They're also going to lose most of their court cases, because this sort of attitude is really unhelpful in a court case. It works better in the "court of public opinion" so they may make progress in legislatures.
I think the word you are looking for is they are too "Young."
I think you are getting way too worked up about demanding 2030 technological proficiency in the back half of 2017.
I don't know about everyone else but I'm perfectly content with proficient levels of automation 3 and 4 for the foreseeable future. Mainly because I know there is no regulatory actions that would slow that down.
Personally I'm pretty happy to go from
December 2016 - No cruise control LOL
March 2016 - Gimicky autosteer at what, 35-45MPH on the highway?
June 2016 - Drive 95% of a round trip from Orange County/Vegas at essentially Level 2 autonomy at 90mph.
I have no idea of Tesla vision is going to be Tesla Waze but I think the potential is going to be great - as long as you are not expecting to just sleep in the car, at least for the near future.
Except for the last sentence, the point is well taken. My understanding is informed by the original "Dave" presentation by Nvidia after 100 hours of machine learning. The machine learns from the driver's reactions to the external world. GPS is useful only to locate where that experience occurred, not to memorize the specific environment. What the computer remembers is that particular response to that particular stimulus. And learns to categorize those appropriate responses by a human driver. Maybe I'm way off here, but it never occurred to me that the point was to develop a 3D picture of the world, that may be the approach of Mobileye or of the lidar based efforts. The Nvidia idea in my simplistic understanding of the Dave paper is to teach the computer how to drive a car. As a 12 year old kid while reversing direction I almost left the road and backed into a ditch. I don't remember anything about that ditch in my mind, but I do remember to be careful about backing up.
Go to the success of skin cancer research. A bunch of photographs are taken. Researchers, playing driver, identify those which are actually found to be cancerous without specifying which signs, irregular color, irregular shape, etc., used by laypersons to ask questions at the next meeting with the skin doctor. The machine, like drivers, learns from experience not maps. One of my dermatologists got a biopsy done on something that turned out to be a pimple. The very same doctor looked at something I knew he had seen before and he said, "but its larger now." Turned out to be a cancer.
Last edited:
Yeah, I don't think there is any other way to do it with resources that exist. The Swift guy makes a big difference.
I think what you are describing is the difference between NBC Columbo (young Columbo ) and ABC Columbo (old Columbo). In that case, young was better. Look up who did the music. So it is not an age thing that you are talking about.
I have little first hand experience with Tesla staff. The folks who presented at SAE were good.
The folks who stood up at last year's shareholder meeting seemed good.
Go back and watch one of the ABC Columbos and look for pandering in the casting. I expect what you saw at this year's shareholder meeting was pandering in the casting as well.
I was not there. You were.
ValueAnalyst
Closed
This is what corruption looks like.
Tesla vs state of Michigan: disclosure of discussions with car dealer lobbyists would result in more lawsuits, state says
Tesla vs state of Michigan: disclosure of discussions with car dealer lobbyists would result in more lawsuits, state says
- Status
Similar threads
- Sticky
