Jul 27, 2016
msnow That could be. That's why I said late next year. I could see Elon pushing this with AP 2.0 to be concurrent with M3 release.�
Jul 27, 2016
sandpiper Suppliers don't "walk away" unless the customer sets their expectations so that the supplier can't comply and remain profitable. Tesla certainly does do that; they're quite ruthless.
It should be very interesting. Tesla, up to a very short time ago, clearly intended to keep working with Mobileye. That implies that they may be scrambling internally around how to continue development without taking steps back. I would argue that autonomy is not really central to what Tesla is. Yes it's great for them to be technologically ahead of the curve. But they are an electric car / eco-energy company - not an autonomous car company. There's a strong argument that vehicle autonomy systems is a separate business in itself You can't do everything; be careful about losing your focus.�
Jul 27, 2016
MP3Mike You mean besides that fact that in the SMP2 Elon stated that the Model 3 will be fully autonomous?�
Jul 27, 2016
Alketi SMP2?�
Jul 27, 2016
Krugerrand Secret Master Plan - Part Deux�
Jul 27, 2016
Cosmacelf That George hotz video in post 16 was very good. George used to strike me as an out of control arrogant dweeb, but maybe previous videos just caught him after he was up for 24 hours or something. In this one, he's well spoken, cogent, careful, skillfully avoids the traps set by the interviewer, etc. I'm impressed. I hope their company does well. Too bad their initial system will only be targeted at two car manufacturers.�
Jul 27, 2016
sillydriver It sounds like the whole vehicle-sharing part of SMP2 requires autonomy and perhaps the city bus replacements too. I think Tesla does consider autonomy to be necessary and thus central to what Tesla is, and likely a competitive differentiator too.�
Jul 27, 2016
182RG No, but a comparison was made to Google, which does offer free food to employees.
You seem to be incredibly pedantic.�
Jul 27, 2016
Chopr147 This forum has become incredibly pedantic in focusing on the cafeteria remark rather than the real meat of the article.
�
Jul 27, 2016
Tam Tesla Autopilot Program Director says need �tighter hardware/software integration� to move �quickly� in autonomous driving
"Tesla�s Autopilot Program Director, Sterling Anderson, is adding some color to the change of direction, saying that Tesla is looking to move quickly in autonomous driving and in order to do that, they need a �tight integration� of both hardware and software. Something they can do better if more of the program is developed in-house."
Mobileye speed is to have "Lateral Turn Across Path (LTAP) detection" by 2018 to enhance its current capability of Automatic Emergency Braking.
Elon Musk wants to have fully autonomous driving system (level 4) by the end of 2017.
The timeline for 2 companies are out of sync.
Now, you know why the divorce.�
Jul 27, 2016
RogerHScott Not to say "I told you so", but this seems to tip the balance fairly strongly in the whole "should I get a Model S now or wait until the
``imminent'' AP 2.0 update?" debate. There's a bigger "lesson" here: the future is always more uncertain than you think, so if there's
something that will make you happy today don't put it off indefinitely in the hopes of something that will make you only marginally
happier coming along.�
Jul 27, 2016
RogerHScott My thought exactly. This makes the already-unlikely prospect of significant improvements to AP based on current-gen hardware
even dimmer. Imagine how unprestigious working on the MBLY-based AP team will soon be.�
Jul 27, 2016
anticitizen13.7 I was surprised too. The Ashlee Vance article from last year portrayed him as arrogant and antisocial. I haven't watched the whole video, but Mr. Hotz seems more insightful and thoughtful than crazy.
Well, to be perfectly honest, I thought that the cafeteria comment was a cheap jab at Tesla. A company like Apple or Google, which is well established and has piles of cash, can afford to have a fancy cafeteria with free food. Tesla is still comparatively early in its development. Is a top of the line cafeteria with free food really where the resources should go?
On the other hand, I do see the validity of the argument that good on-site food that can be acquired easily will make things better for engineers working 70+ hours/week. I just don't know. I've worked at many different places, some with good food, most with average food, and some places with no food on site at all. It did not make a whole lot of difference to me.
Disclaimer: I am a picky eater. My diet consists largely of plants, seeds, and fruit. I rarely ate what was offered on site.�
Jul 27, 2016
RogerHScott You're kidding, right? How could this possibly not delay AP 2.0, and significantly?�
Jul 27, 2016
Cosmacelf Mobileye did say that Tesla would continue to get their EyeQ3 software improvements. Also, there are still algorithm improvements you can make on the legacy platform that can be transferred over to the new platform, so there will presumably still be AP improvement.�
Jul 27, 2016
Cosmacelf Hard to say. Mobileye EyeQ4 isn't expected in volume until 2018, 2 years from now. It is possible that Tesla could create a new AP platform even earlier than that. George Hotz's company says they're going to essentially replicate current Tesla AP functionality by the end of the year, that's about 1 year of development.
But these are very complicated systems, so who knows what difficulties lie ahead for all these efforts. All three systems (EyeQ4, Tesla AP 2.0, and George Hotz's Comma.ai) are still in development.�
Jul 27, 2016
RogerHScott Of course, there's a huge (dare I say "yooge"?) assumption built into that that anyone, let alone Tesla, can actually do it faster. Just
because you don't like reality doesn't mean there's an alternative. Everyone's spinning this as MBLY "not being up to the task" but
isn't it just as likely that they told TSLA very clearly what the limitations of their technology were and TSLA built something out of it that
MBLY never would have endorsed (due to its overreach)?�
Jul 27, 2016
dakh So essentially they want to ship Model 3 at least with hardware that will support full AP. Let's give them some slack on tidying up corner cases in software after they ship�
Jul 27, 2016
Chopr147 Then God said, "I give you every seed-bearing plant on the face of the whole earth and every tree that has fruit with seed in it. They will be yours for food.
Sorry, your comment sparked a memory�
Jul 27, 2016
MP3Mike Maybe AP 2.0 was never going to be a Mobile Eye based solution. EyeQ4 wasn't going to be ready in time, so Tesla went a different direction.
Elon has said that the Model 3 will be capable of full autonomy, i.e. AP 2.0, and that they will start making them in the Summer of 2017. That is when I expect that we will see AP 2.0. (Though maybe they will be able to put it in the S/X first, but I wouldn't bet on it.)�
Jul 27, 2016
dakh In a properly layered system it should be possible to swap out a lower-level component without modifying higher-level component. From the very incomplete information that we're exposed to it sounds like Tesla used Mobileye hardware and some of the lower level software layers but didn't use their whole stack. So if Tesla already has its own tech for using those chips to make sense of camera feed, it is a fairly well "boxed" problem to solve. Not saying it's trivial or easy and maybe there are some patent barriers to overcome. But Tesla is confident they can do it.�
Jul 27, 2016
sandpiper Well... the model 3 may be fully autonomous - but it certainly won't the the 2017, 2018, 2019 or 2020 model years. My bet is not before 2030 and probably quite a bit later. In any case, the whole market is pursing autonomy and when we do get there Tesla won't be the only one. Tesla's key differentiator is the electric part.�
Jul 27, 2016
Johan It seems you have not quite made yourself acquainted with mr. Elon Musk.�
Jul 27, 2016
Matias I'm pretty sure that not at launch.�
Jul 27, 2016
MP3Mike Maybe the software won't be ready or it won't be approved for full autonomous use yet, but all of the hardware necessary will be in every Model 3.�
Jul 27, 2016
RDoc Unfortunately, however, a lot of us have become quite well acquainted with Mr Elon Musk's company's software. Generally, on first release it's terrible, bordering on embarrassing, occasionally dangerous. Then it improves at glacial speed, never reaching the original promised functionality, even years later.�
Jul 27, 2016
MP3Mike Maybe the software won't be ready or it won't be approved for full autonomous use yet, but all of the hardware necessary will be in every Model 3.�
Jul 27, 2016
Matias How do you know this?�
Jul 27, 2016
MP3Mike I am going based on what Elon has said and his SMP2.�
Jul 27, 2016
Matias I've read SMP2 and it doesn't say, that Model 3 will have all the necessary hardware for fully autonomous vehicle at launch.�
Jul 27, 2016
Alketi The industry is moving WAY faster than a 2030 target. Elon believes it's a solved problem and will be ready 2 years from now, adding perhaps another year for regulation to catch up. Personally, I think that's a bit optimistic, but I would only add a year or two to his numbers.
In any event, yes, the Model 3 when shipping will absolutely NOT be fully autonomous, but the hope is that it will have the hardware that will eventually support it. Autonomy is a software problem. As we all know, software can be patched/upgraded in the field. )�
Jul 27, 2016
Johan I thought we we're discussing hardware here? In-house proprietary Tesla chips as opposed to Mobileye chips and cameras. In my experience Tesla's hardware ia excellent.�
Jul 27, 2016
dakh We don't know yet how the hardware is gonna look like. They're obviously designing something but is it complementary to say NVidia chips and a close collaboration announcement is about to be made, or full in-house built solution replacement, or something else, we'll just have to wait and see.�
Jul 27, 2016
sandpiper It's a mixed problem. Autonomy is only becoming possible because of dramatic improvements in parallel processing capability - which has enabled practical neural networks / deep learning. Yes there's software, but the hardware has to be capable of supporting the software. Full autodrive will also probably require some method of vehicle to vehicle and vehicle to ground station live communication. There has to be a standard developed for that, and you can't build hardware to meet a non-existent standard.
Elon tends to exaggerate / be over-optimistic. And even when he's doesn't people take his words and blow them up well beyond what he actually intended.
Autonomy is such a complex problem that it's simply not possible to deploy quickly. With over 3 trillion miles driven annually in North America, it's not good enough to get it right 9999/10000 times. It has to be much better. I'm also convinced that it will require infrastructure upgrades: better standards for road painting, lights/signs/signals, more accurate GPS and possibly some in-road active hardware. I suspect you'll first see auto-drive only lanes, then autodrive-only roads and eventually auto-driving will start to deploy into the mixed traffic areas.
I remain 100% convinced that we're going to see gradual improvements over the next 10-20 years with legislation slowly tracking behind. Full level 4 autonomy? 2030 to 2040.�
Jul 27, 2016
Tam I think quitting MBLY will speed up, not slow down the implementation of AP 2.0.
MBLY is just too slow for TSLA.
TSLA has a good track record that when suppliers are not up to the task, it would bring the job in house.
This year's annual shareholder meeting, Elon gave quite a few examples of that such as:
From the very early day, with the Roadster gearbox, they thought they would get faster implementation by letting the "experts" from outside suppliers to do it for them but it was in fact, much faster just to design it in house from a scratch and produced it onsite.�
Jul 27, 2016
scottm Plus, you could die at any time... so enjoy today!�
Jul 27, 2016
KZKZ I'm not so sure about that, especially when you look at how the Falcon Wing Doors have turned out. That work was brought in house.�
Jul 27, 2016
KZKZ everything I read also said it was Mobileye who did the dumping.�
Jul 27, 2016
plankeye Whatever... Their loss once Tesla starts selling half a million cars a year.�
Jul 27, 2016
3Victoria Do you have reason to believe a third party could have done better?�
Jul 27, 2016
Tam Originally, in February 2014, the job was sourced to a third party Hoerbiger to deliver the finished product to Tesla.
In May 2015, it came back with oil leaking from the hydraulic system so much that Tesla decided to bring it in-house.
Tesla decided to start it from a scratch and redesigned the system and ditched the third party's hydraulic system.
It then went with electronic mechanism instead.
And that was a very incredibly short time to turn a disaster around in just 4 months to deliver a first Model X in September 2015.
Yes, there have been problems but they are fixable because it's done in-house.
Those in-house problems are just like a walk in the park when you compare them with the finished product from a third party.�
Jul 27, 2016
Tam So who would you like to listen to?
Elon Musk or Mobileye?�
Jul 27, 2016
msnow Tesla because what ME said in that article made zero sense.�
Jul 27, 2016
RDoc Custom chips are run by software. Speaking as someone who's designed and build complex software system for decades, Tesla's software department is at best second rate.
I'm unaware of any Tesla custom chips, what do you have in mind?�
Jul 27, 2016
Stoneymonster Tesla's safety critical systems and drive control are certainly not "second-rate".�
Jul 27, 2016
trils0n Exactly. The infotainment is shoddy, but the software that actually runs the car has been rock solid since day 1.�
Jul 27, 2016
Krugerrand And yet Tesla does in fact offer free food stuffs to its employees.
Yeah, I like to get the facts straight.�
Jul 28, 2016
malcolm It's interesting to recall just how many "Never-gonna-happen" moments we've seen delivered in the course of the last five years.
Ludicrous mode. Increased range with AWD. The introduction of Autopilot. The Falcon wing doors.
Certainly the level of challenge and complexity with some of these has been more painfully apparent than with others and we can argue that most should have arrived sooner and/or in better shape, but it remains an astonishing tally of engineering achievements in a sector where most rivals are still trying to match the Roadster (*cough* Mission E *cough*).
MobilEye is very innovative compared with industry norms. But as several people have observed, when it comes to Elon, we're not talking normal (or rather, "just comfortably above-average")
And on top of that there's the engineering achievements of Spacex. Some of which, cross fertilize the work at Tesla.
Really envious of those of you with the chance to see a fraction of what the Oompa Loompas of Science have been getting up to at the Gigafactory tomorrow.�
Jul 28, 2016
Stoneymonster And the infotainment is still better than any of the cars I've owned before, so while I agree there's a *ton* of room for improvement there, you couldn't exchange it for anything else on the market for me right now. A lot of that is thanks to the screen.�
Jul 28, 2016
RDoc When we first got our car, the AP would consistently get way too close, way too fast, to stopped cars ahead, requiring me to always override it. That's much better now, but for them to release such a potentially dangerous system doesn't speak well for their software department. The car still will often overshoot lane makers, sometimes with other vehicles next to the car. This is also dangerous and has been complained about for months.
By "drive control", I assume you mean running the motor and things of that sort. These are orders of magnitude simpler than even the current flawed AP software system.�
Jul 28, 2016
RDoc My 5 year old aftermarket Sony head unit has far better software. For example, it knows how to use playlists and I've never had it crash when reading a USB drive.�
Jul 28, 2016
Stoneymonster I think the fact that you are using an aftermarket unit kinda supports my point�
Jul 28, 2016
RDoc Well, apart from dying, the Blaupunct that came in my 1991 NSX lacked a few features.�
Jul 28, 2016
J1mbo Tesla doesn't write the software for the safety systems. That comes from the OEM (Bosch or ME). Tesla adopted ME's development (learning) process to refine Autopilot, but the actual "safety critical" component of the software is from the OEM.
On the other hand, Tesla have spent over 10 years developing and refining their sublime drive control. At that rate we can expect the media player and navigation to be out of beta within about 5 years...�
Jul 28, 2016
Waiting4M3 You're aware of a bunch of chip design guys that Tesla hired from Apple(PA Semi) and AMD this year, right?�
Jul 28, 2016
MP3Mike Are you sure about that? I thought Tesla pretty much writes/maintains the software themselves linking all of the hardware they get from OEMs. Which is why Autopilot is so much better than any other car with a Mobile Eye system.�
Jul 28, 2016
J1mbo All other car manufacturers train the ME system (DNN) during the development cycle, using tools supplied by ME. The other manufacturer's system may be updated once a year as part of the annual service. Tesla's innovation seems to be that they have extended the ME training process to the whole fleet, and they leave it running all the time.
It will be interesting to see whether they are able to transition something as tailor-made as an ME DNN (taught to respond to millions of miles worth of radar + visual data, integrated and processed by ME's custom EyeQ3 chip) over to another hardware/software solution without compromising quality or safety.
My bet is that they won't: Tesla know full well how quickly this kind of system can improve once it is "out in the wild".�
Jul 28, 2016
Bruin1996 So I was talking to one of my Tesla-owning colleagues today about Autopilot and he believes that the discontinuation of the relationship with Mobileye will cause a delay in the release of the Autopilot updates in Tesla OS 8.0. What do you all think about this? Has anyone heard any news to support this?�
Jul 28, 2016
MP3Mike I highly doubt that since Tesla and Mobile Eye have said that they will continue to work together to support/improve the current solution Tesla is using. If it was going to delay anything it would be AP 2.0. (new hardware)�
Jul 29, 2016
Tam The Future of Tesla Autopilot - What Happens After Mobileye?
The break up was well planned in advance by Tesla as evidenced by the hiring frenzy in the following quotes:
"
- Jim Keller (legendary microprocessor architect and formerly of Apple and AMD) joined Tesla Autopilot as the VP of Autopilot Hardware Engineering.
- Peter Bannon (formerly of Apple) followed Jim to Tesla Autopilot.
- David Glasco (formerly worked at Intel, IBM and Nvidia) was Senior Director of Server SOC Architecture at AMD before joining Tesla.
- Thaddeus Fortenberry (formerly Cloud Server Architect at AMD) also joined Tesla Autopilot in the coming weeks.
- Debjit Das Sarma (former AMD fellow and CPU Lead Architect) also joined Tesla soon after.
- Keith Witek (formerly AMD Corporate Vice President of Strategy and Corporate Development) joined Tesla as the Director of Autopilot Enablement and Associate General Counsel.
- Junli Gu (Formerly Member Technical Staff, Machine Learning at AMD) joined Tesla as the Tech Lead for Machine Learning, Autopilot"
Speculation: 2 choices:
1) In-house chip design: just like some other components, in addition to using the products for itself, it then can supply them to others (powertrain/battery packs to Toyota & Mercedes in the past, Powerwall/PowerPacks for homes & utility companies...)
2) Switching to Nvidia that is flexible and adaptable to keep up with the fast paced Tesla.�
Jul 29, 2016
Frank Schwab I architect chips like this for a living, though in a different business. I've lived through pissing off every possible customer in the world because we couldn't immediately scale by a factor of 10 or 100, and couldn't deliver product on time. I've lived through having that attempt to scale completely disrupt the development of new products, causing delivery of new product to be years late.
It's not a stretch to assume that a similar thing could happen to ME.
It's not a stretch to assume that their current architecture has hit a wall, and can't be simply extended to meet the new customer requirements. In such a case, the next generation gets delayed while architecture starts over from a fresh sheet of paper.
Such a situation could certainly lead to what we see here - more-or-less a mutual parting of the ways. ME recognizes that they can't meet Tesla's new customer requirements, and Tesla recognizing the same. The two would simply be incapable of reaching agreement on delivery, and the relationship would be over.
It's also easy to imagine that, assuming a limited design capability, ME would have to make a decision on what they'd spend their design resources on. Should they bet their company on supporting a niche player in the market with a great name and great prospects, or should they bet their company on the entire rest of the market whose shipments two years from now will be 200 times that of the glamorous company?
Not all situations have Machiavellian maneuverings behind the scenes; sometimes your engineers are just tired.�
Jul 30, 2016
sillydriver My concern is that even if it's been six months since top chip architects were hired, the lag from that event to having software running on new silicon designs in shipping cars has got to be quite long. I wonder if there is an interim solution using Mobileye hardware to meet the Model 3 timeline.�
Jul 30, 2016
Odebek I think the timeline will work... basically 12-16 months for version 1.0. They don't have to worry so much about power, heat or size constraints the way you would for chipsets for mobile devices so that will really speed up the initial design process.�
Jul 30, 2016
sillydriver I've been meaning to ask - is that license plate meant to be binary for 42 (the answer to life, the universe and everything)?�
Jul 30, 2016
Odebek Yes, but Don't Panic�
Jul 31, 2016
ShockOnT I envy your wit. Shame here in Aus they want $440 per year for personalized plates....�
Jul 31, 2016
bhzmark The Future of Tesla Autopilot - What Happens After Mobileye? is a good article speculating on what is next.
Remember that in Dec George Hotz shared emails with EM suggesting that EM was open to finding a replacement for Mobileye. And then in March EM went to Mobileye in Israel reportedly to see what they offered for later iterations of AP. Elon Musk reportedly visited Mobileye to test tech for next gen Tesla Autopilot Now it seems that perhaps Tesla didn't like what they saw. or vice versa I suppose.�
Jul 31, 2016
22522 To do autonomous well you need to take advantage of the sensor array capabilities of a platform.
ME can't do that.
Tesla can.�
Jul 31, 2016
SR22pilot Interesting. Mine had a very nice Bose unit.�
Jul 31, 2016
Lex Well that's that for me, I'm totally comfortable on this topic now. Can't wait to see the in-house or Nvidia-partnered Autopilot 2.0 vision systems.
But I am wondering why folks seems focused on the Model 3 having the first rendition of AP 2.0. Elon said that the Model S (and presumably the X) will always get the best tech first as the flagship model(s). Though I agree that the hardware could be ready for all models by end of 2017.�
Jul 31, 2016
sillydriver I spoke above about meeting the model 3 timeline because I think it would be trouble for Tesla to introduce the M3 with the old system and replace it soon after. It's the M3 that has a public date attached (mid 17) that would upset people if it was missed. I agree that if the new AP is ready before, it would ship in the MS and MX first. I think we all hope that will be the case, but I worry that the bigger a technological discontinuity the new AP is, the longer it will take to get it out the door.�
Jul 31, 2016
Alketi If Tesla is designing their own processing solution in-house, then to my mind, that would seriously call into question an AP 2.0 hardware update by the end of this year (to presumably coincide with Musk's autopilot announcement that he alluded to at the All Thing's D conference).
Further conjecture -- Musk's tweets about ways to innovate with the current hardware (temporal smoothing, etc) also indicate that the current hardware may have a longer run than expected. (Of course, temporal smoothing could still apply to next-gen hardware as well.)
Another option for Tesla is to ship next-gen sensors before shipping the next-gen chipset and then update the processor assembly with a service visit , but that requires a common hardware interface, and that the old chipset would need to support new hardware, and that software would have to be re-written and tested, etc. All to deliver a temporary solution. It seems highly unlikely they'll go this route.
I had been thinking AP 2.0 hardware would arrive this year, but I'm now thinking that's probably too aggressive considering the ME news. I'm imagining that the deadline is really the Model 3 launch (next July) for AP 2.0 hardware with, at least, AP 1.0 functionality...�
Jul 31, 2016
Cosmacelf Yes, I was hoping for AP 2.0 by end of year. I now worry they'll get it by end of next year...�
Jul 31, 2016
Richg In reading through this thread, I find it strange that Apple is not considered as a potential supplier. There are large investments by Apple toward this exact area of technology. It addition with the investment in China's answer to Uber, it shows Tesla and Apple have similar objectives.
Being in close proximity, basically serving the same customer profile, and having like strategic visions ..... Why not an Apple / Tesla relationship?�
Jul 31, 2016
Alketi Apple is essentially a future competitor and currently developing their own self-driving EV. I think a partnership with nVidia would be more likely.
But, my guess is that the major players self-driving hardware players (MobileEye, nVidia, others) all seem to naturally be developing end-to-end solutions on a longer timeframe, solutions that also may not have the flexibility that Tesla wants, and that Tesla decided it's better to just bring this all in-house. The hardest part (by far) in self-driving is the software, which I think Tesla already significantly owns. The easier part is a hardware controller that takes inputs from the various sensors.�
Jul 31, 2016
AZ Desert Driver This AP2 news...I WAS hoping that AP2 was just around the corner and I could delay getting my MS by a month or so to capture the next iteration. Now I'm thinking the time to design a new chip, build it, test it, write software for it, test that, then deploy into the fleet.... might just take more than a month or so. Conclusion - buy the car of my reality now, and let the dreams mature for a future round.
Thanks guys for the sound analysis of the conditions. I had been speculating based on incomplete data. Now I think it is clear as to the length of the path.�
Jul 31, 2016
Tam 1) Apple's introduction of its map was disastrous. Apple is new in Autonomous Vehicle area and it is questionable that they can duplicate Tesla's commitment to solve the problems timely.
2) Apple's philosophy is incompatible with Tesla's:
Although both are for-profit companies, Apple goes out of its way to make sure it's profitable by using all the tricks in the books:
a) Monopoly any way you can from hardware to software while Tesla opens its patents and invites others to advance sustainable transport.
b) Human exploitation as evidenced by using low wage assembly workers in China while Tesla is willing to pay more for workers with rights as well as union rights (at least 50% unionized work force in Gigafactory.)
c) Although since 2014, Apple has made efforts to follow Tesla's footstep in avoiding using Conflict Minerals but Amnesty International still names Apple as one of violators.
d) Tax avoidance scheme by utilizing a variety of offshore structures.�
Jul 31, 2016
Stoneymonster Oh sure, you and @RDoc just casually NSX dropping like I won't be jealous.�
Jul 31, 2016
ShotgunF15E That's the same line of BS I was told about my MS. "All the hardware is already in the car."
Now we know the AP 1.0 HARDWARE in the Model S/X will NEVER be capable of full autonomous use. More false advertising...�
Jul 31, 2016
RDoc I stand corrected, it was a Bose unit. One of the door speakers failed, then a few months later the head unit failed. I replaced it with a Sony and the door speakers with some coaxial aftermarket speakers and couldn't believe how much better the sound was.�
Jul 31, 2016
Cosmacelf When did Tesla promise full autonomy with the current car hardware? Before and as AP came out, Elon specifically talked about freeway on ramp to off ramp autonomy. No more. Tesla, AFAIK, has never promised urban driving with the currently shipping cars.�
Jul 31, 2016
RDoc While I dearly love my NSX, I must say that once we got the Tesla, for the first time, the NSX felt like what it, of course actually is: A really great and classic antique.�
Jul 31, 2016
stopcrazypp That's what I remember too. On ramp to off ramp was the best case scenario for the current hardware. Full autonomy would take a lot more hardware.�
Jul 31, 2016
hybridbear I work in Conflict Minerals, and you can't claim that Tesla is more of a leader in Conflict Minerals than Apple is. In fact, Tesla included South Sudan as one of their countries of origin. South Sudan has no sources of conflict free 3TG minerals. The choice to list all listed countries as countries of origin shows a lack of understanding of the regulations on the part of Tesla.�
Jul 31, 2016
AZ Desert Driver I'm not clear what this means. Is everything from South Sudan off limits? Could, say the headliner, material be from a country that has conflict minerals, but not be a conflict material? What is 3TG minerals?
Not trying to be argumentative - please just educate me a tad on this important issue..�
Jul 31, 2016
sillydriver I was wondering the same thing, so i looked around and found this
Conflict resource - Wikipedia, the free encyclopedia�
Jul 31, 2016
SR22pilot I thought maybe cars in some other countries had different hardware. For its day I loved that car. Both of mine were very reliable and a blast to drive.�
Jul 31, 2016
Tam
Conflict Minerals: The Truth Underlying the Systemic Looting of Congo
The 3 Ts are minerals called:
1) Tin. It's used as solder.
2) Tungsten. Tungsten alloy vibrator helps your phone vibrate.
3) Tantalum. It's great as capacitors.
G: is for Gold. It's a great non-corrosive conductor.
Democratic Republic of the Congo has very rich ores with these.
The problem is the revenue from these mines only enriches the cruelty of war in the region, so they are called "conflict minerals."
If so, why don't we call oil as conflict resource that sustains perpetual wars such as ISIS?
Well, someone already set up an organization for anti-cruelty for Democratic Republic of the Congo with "conflict minerals" awareness so we'll need someone to do the same for oil too.�
Jul 31, 2016
DougH I think the chocolate that I buy at Whole Foods helps the animals of the Congo.�
Jul 31, 2016
Bet TSLA You know, random irrational Apple hatred sounds awfully similar to random irrational Tesla hatred.�
Jul 31, 2016
stopcrazypp The Conflict Minerals point is on iffy grounds, but the rest succinctly shows the major differences between the two companies.�
Jul 31, 2016
Tam There's no perfect company but I am just trying to point out the differences between Apple and Tesla.
Whether those differences are irrational or not, they are still differences.
I would not classify Tesla critics as irrational. They do have valid reasons such as:
All their lives have been comfortable with oil and all this Tesla talks would just threaten to disrupt their way of life...
Some would say that fear is irrational but I would say their concern is legitimate and is worthy of being addressed.�
Aug 1, 2016
ShotgunF15E When I bought my car. Granted, I guess I should have known better than to trust what a Tesla employee at a store was telling me at the time.....he also said my car would do the 1/4 mile in 10.9 seconds....�
Aug 1, 2016
Tam
It was understood that this would be a first version or beta and you can see very clearly that the camera is mono-lens. You need at least two separate cameras or a dual-lens stereoscopic canera for depth perception. That's a good indication that it won't stop with a mono-lens system.
Last year, an article explained its suitability for freeway driving.
Tesla�s Cars Now Drive Themselves, Kinda
"So will this sensor suite be enough to enable autonomous driving eventually? In a word, no. �This is not the full autonomy sensor suite,� says Musk...This software update won�t work everywhere you go. It�s really for highway and freeway driving..."
I assume that Tesla is just like a smart phone that you might want to reserve a budget to trade up every few years.�
Aug 1, 2016
Matias ^Anyway according to him Tesla's salesperson told him other story.�
Aug 1, 2016
Tam Agreed.
The lesson here is: Salespersons are a good start but you might have to cross check because they might just be too optimistic about a product and might overstate its capability.�
Aug 1, 2016
J1mbo While stereo vision makes it easier, depth of a perceivable object also be calculated using a mono lens.
Try wearing an eye patch and walking towards a brick wall. You won't hit the wall unintentially, because your brain has worked out how close it is using mono sight and relative velocity.�
Aug 1, 2016
sillydriver Yes, when it was introduced I recall that Elon said the goal of autopilot was thar the car would drive itself on the highway without intervention from on ramp to off ramp, which at the time seemed like an amazing thing. And that goal seemed to be almost immediately achieved. Then 7.1 introduced nags, which for me was a step back from the goal.�
Aug 1, 2016
AZ Desert Driver I can afford to toss away a $500 phone every three years for the newest and greatest. Not expecting much salvage value from an obsolete phone. But a $80,000 car better have some significant salvage value at the "end of its 5-year life". There should be folks who want a used car with lots of capability, just not the newest feature.�
Aug 1, 2016
jbcarioca As it happens I have only one functioning eye. Anybody thinking monocular vision can simulate multi sensor vision does not understand too well. I also possess an Airline Transport Pilot certificate. The examination to grant a SODA (statement of demonstrated ability) proved, to me anyway, that a large amount of ancillary information is required to compensate for the limitations implicit in monocular vision. I am confident that multiple sources will be cheaper and more reliable than will be monocular workarounds. Of course I am betting on my analogy...�
Aug 1, 2016
RDoc I don't see why the car's vision system has to produce 3D information. The radar unit is extremely good at depth perception.�
Aug 1, 2016
electracity The problem with radar is probably resolution. Is that a short fat person about to step into the street, or a mailbox?�
Aug 1, 2016
Tam Hopefully, Tesla will have more affordable cars, even cheaper than Model ? so that trading in for more whistles and bells won't be so bad on the family budget.�
Aug 1, 2016
Tam
I used to romanticize that a Radar would draw me a nice picture of a car just the same way as weather channel gives me a beautiful picture of radar based cloud/rain/snow image or nice icons of airplanes with their IDs at air traffic controller console...
That until Elon tweeted
"Radar tunes out what looks like an overhead road sign to avoid false braking events"
That's when I look around and realize that it recognizes a car by a pre-fed car signature, the complicate up and down waveform graph:
Radar Signatures of a Passenger Car
�
Aug 1, 2016
Tam To clarify, I thought Radar images should be like LIDAR before Elon's tweet. Since then, this is what I now understand:
LIDAR can map out a whole traffic block full of images and it is taught what those images mean such as the gesture of a bicyclist for a left turn signal below:
However, Automobile Radar seems not create an image and its shape. It can create a "signature" so that it could learn this signature below is a motorcycle but it has no idea what a motorcycle looks like (shape) or whether there's a hand signal or not:
�
![[?IMG]](https://lh3.googleusercontent.com/blogger_img_proxy/AEn0k_vBKYw--5lPhx8DJTArqzsXr-_IwmeP-85sl00QRv__4SVmmVBJeSU3kdjMUgC257E335vyBUHz-fX9knbngstipSTPTJ_HArZJMKE50yHb6NTg8EPku-0l47s3ynLUtL9J=s0-d)
Aug 1, 2016
J1mbo Sorry to hear that. When growing up, one of my best friends had a benign tumour which had blinded him in one eye and also made him completely deaf on that side. He was also a keen sportman, and one of the best cricketers I knew. When asked how he could tell where to hit a solid ball that was coming directly at him at very high speed, he put it down to "instinct" and "practice". I guess this is where the MobilEye IP comes in to play, replacing "instinct" with algorithmic prediction, and using fleet learning for "practice".�
Aug 1, 2016
Cyberax Radar that is used in Tesla has problems with detecting stationary objects. It can easily detect objects that are moving related to the background with good enough resolution, but not static objects.�
Aug 2, 2016
bhzmark Does it have a problem detecting them or merely distinguishing cars stopped in your lane or that you will otherwise drive around, from cars stopped/parked on the side of the lane? See the thread where people complain that Tesla detects cars parked in front of a lane that merges and reduces power. Something is detecting those parked cars.�
Aug 2, 2016
sillydriver I think the issue is that the whole world is sending back a mess of echos from objects that the doppler rader identifies as stationary, where the radar cannot reliably discriminate what these echos are from. So the system pays attention to those few echos that are not converging at the forward velocity of the car - those are the echos of moving vehicles.�
Aug 2, 2016
Alketi The camera is used, in some capacity, as a backup system for object identification to distinguish stationary cars from overhead roadsigns. However, the current generation camera is monochrome (RED channel only), so it only sees objects with enough contrast.�
Aug 2, 2016
electracity Why would it be red channel? Would not it just be unfiltered for color?�
Aug 2, 2016
electracity I'll throw an idea into why Tesla is not partnering with mobileye: Tesla Bus will be an implementation of Google Car.
It seems impossible to me that Tesla could beta a limited urban bus by 2020. Occam's razor leads me to Google Car. Tesla Motors and Tesla Semi will continue down the current Tesla development path. Tesla Bus will implement Google Car.�
Aug 2, 2016
Stoneymonster Four years is a long time in this business now. I disagree that it's impossible.�
Aug 2, 2016
electracity Why announce a low speed minibus? Why not just a minibus? Tesla's focus is highway tech. If they can do low speed in a complicated environment, they can do high speed more easily.
A dual development path, each starting in a different place, is likely to produce the best outcome.
Musk simply could not announce a Google deal until AP 2.0 is out, and the model 3 is in production.�
Aug 2, 2016
3Victoria @electricity this appears to be wild and unsubstantiated speculation on your part, ie FUD. Your other posts seem to be in the same vein.�
Aug 2, 2016
Tam May I remind 2 big differences between Tesla and Google:
1) Tesla believes in incremental improvements even when it's not perfect which means additional deaths and crashes are expected during those incremental rollouts. It believes even during a period of imperfections, deaths and crashes are much more reduced although not eliminated.
Google believes that drivers are not to be trusted with an imperfect product so it purposely withholds partial implementations and will only introduce a complete product. It is willing to continue to expect a traditional staggering number of deaths and crashes because those are not currently caused by its products.
2) Tesla does not believe in an addition of LIDAR because it is not necessary. Google believes that you need all the best sensors that you can get, so LIDAR is a must.�
Aug 2, 2016
electracity I'm not sure you understand the concept of FUD.
But I am sure you have no engineering or computer background.�
Aug 2, 2016
3Victoria Actually I have both.
Thank you.�
Aug 2, 2016
Alketi I've played with briefly with image recognition algorithms (pre neural network) and the first thing you generally do is throw away the color information, compress the histogram, and generate contrast for edge detection. I have no idea if this is what MobileEye is doing, but by having a monochrome camera you reduce the color information to a single channel (essentially giving you a tinted grayscale image) which I think would give you a headstart on that process. That's just a rough guess at what's happening.�
Aug 2, 2016
electracity "Tesla believes". Tesla had no choice. They started late, and leveraged mobileye technology. It will not be Tesla choice to determine how many people they get to kill testing autonomous driving. One reason to start on the slow speed end, like Google Car, is fewer customer decapitations.
It is not about LIDAR. It's a computer science problem.
Tesla might be doing the slow speed car without a major partner. But this situation would concern me. They have so much to do with the model 3 and batteries.�
Không có nhận xét nào:
Đăng nhận xét