That was in reference to releasing Navigate on Autopilot and the car being able to path plan it's way around cloverleaf interchanges where it can't see very far around the curve. The ability to path plan "blindly" around curves didn't exist in publicly released vehicle software before NoA was released. It existed before then for some time, being tested, though how long it was tested and how long it took to develop it before then is unknown to us. The same regarding cut-in detection. They don't just develop the feature and dump it on the public untested immediately after, he was referring to when we saw these features begin to be included in public builds.
-
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
Several people have in fact schooled Elon. I'd like to take a moment to recall the VC investor who begged him not to do franchise dealerships.
It is. You have to define what “the right” is in 3d space, in all possible conditions and what it means to “pass” in those same conditions. You need to identify high-level exceptions to it as well(car broken down on the left, for example, or if you’re exiting the freeway, or if the left lane is a carpool lane, or, in some cases, when the *right* lane is a carpool lane.
We subconsciously solve these problems pretty much instantly, but actually defining all the rules that govern them is extremely difficult. Worse, we need a way to inject these decisions into the middle of a neural network doing things we don’t understand.
They said that they were two years away in 2015 or 2016. I said they were wrong. I was right. Game, set, match.
Yeah, that's called "being an idiot" on your part. They are deluded, and so are you.
I have an awful lot of background knowledge here. I was not impressed by the technical detail of the presentation, and I did follow all of it. The Tesla guys are very smart, but they are too close to their work and are not seeing the next wave of problems they're going to hit -- and Musk is an incurable extreme optimist. I already know what they're going to hit. I'm waiting for them to notice it.
Once they notice that set of problems, it will take at least two years for them to resolve it (and that's optimistic). They haven't noticed it yet, therefore it will take more than two years.
Dan Detweiler
Active Member
Why wouldn't he if he, his team, all the people that work on FSD all say it's coming? The man speaks what he feels. He doesn't hold back. You know that. I know that. The whole world knows that. So why are you losing so much sleep screaming at the wind. You're going to give yourself a coronary.
Dan
I don't know if anyone saw my note last night, but if Google were to give every car owner (on earth, or in all new veh, or in a particular state...) a free camera with a cell connection, could they get their missing data and catch up to Tesla?
How else could someone level the data field without a car that includes data gathering?
The point that older Model S's still contribute to learning is a hint. Seems it doesn't need a NN?
I am not saying learning those rules by observing others, I am saying manually translate those road rules into software rules manually. I don't see how hard it is to specify the rule: "stop before entering the road".
UBS perma-bear "Colin Langan" is (unsurprisingly) doubling down on their LIDAR misunderstanding and FUD:
"UBS analyst, Colin Langan, reiterated a Sell rating and $200 price target on Tesla after attending the company's Autonomy Day. The highlights include a target date for he deployment of "feature complete self driving" capabilities, the launch of the Robotaxi in 2020 and featured the in-house designed TSLA chip has more power and uses less silicon than the prior gen NVDA chip."
'The analyst stated "The primary sensor is often vision, but most experts believe LiDAR will be needed since it provides more unique data, even if it's small, that makes the entire system safer. Regulators will likely want more sensor data to ensure the highest level of safety. TSLA may also need more nonsimulated testing. TSLA reported 0 AV miles driven in California in 2018 vs. GOOG which reported 1.2m".'
There's two false claims in these two short paragraphs already:
He is hiding behind the "most experts believe" weasel words and false appeal to authority, instead of using easily verified facts and logic. I absolutely do not want to read his full analyst report: garbage in, garbage out.
- He is misconstruing the fact that Tesla did not opt to report disengagements in California into a false claim that Tesla only performs "simulated testing". In reality Tesla got disengagement events from about a billion miles of Autopilot driving, a three orders of magnitude larger data stream than the Google case UBS cites ...
- They also don't realize that even a very cheap ~$5,000 LIDAR sensor will crowd out much more effective sensors, i.e. in Tesla's model LIDAR will actively reduce car safety and kill people.
In other, completely unrelated news, UBS is apparently positioning to IPO Waymo:
Alphabet's self-driving car business could book $114 billion in revenue in 2030, says UBS
"On the heels of Waymo’s commercial launch on Wednesday, investment bank UBS estimates that Alphabet’s self-driving car unit will reel in $114 billion in revenue in 2030."
Which I suspect explains why they absolutely have to "misunderstand" Tesla's plans.
I think the main problem analysts have here, is a misunderstanding of who the experts are. They're Tesla, and nobody else. All these other "experts" are clearly years if not decades behind the state of the art. It doesn't matter if most of the people in the field agree, if they're wrong and behind the times. Whether that's an intentional misunderstanding on the part of the analysts may vary by analyst - some may be following the herd.
You summarized it very nicely. Thank you.
This is the point the nay sayers out there are overlooking and so is @Bladerskb
The answer is clear: Because their system doesn't work there yet.
shlokavica22
Member
It's not just a camera though. Pictures are only part of the data they collect. You have velocity, speed, etc. Then you have the cars used as "a mule" to test scenarios and so on and so on.
My only guess is that they were trying to get people to stop putting money into Waymo and Cruise? (Which investors should stop putting money into)
Yes, you can write that rule for your car, but then you have to write another rule for the car that doesn't. There are just too many individual cases to write rules for, many of which aren't covered by any law (e.g. there is no law against a deer jumping out in front of the car--but it happens frequently).
Not exactly. It runs the network for some outputs that aren’t public yet and looks for conditions where that output turned out to be incorrect. When that happens, if saves a snapshot and uploads it to Tesla HQ for data ingestion. I believe one of the examples was for cut-in. They ran the cut-in detector in shadow mode and any time the person didn’t actually cut in, they uploaded video.
LOL. I can assure you they don't.
Actually, I think this is one of the most underrated qualities that the "spectrum" gave Musk.
It makes him think outside the box and fears and corporate cultures: if it's physically possible, then it's achievable.
The problem is when he encounters unknown unknowns, or simply he doesn't really know where real complexity lies:
he often misunderstands how people work, I would say (this is quite common, as you know).
Regarding FSD, I think he's overoptimistic with timelines, but also that he thinks that a simple 10x improvement in crash statistics will give Tesla the opportunity to use level 5 robotaxi.
I think this is a flaw in his reasoning: nothing guarantees us that regulation and media will be reasonable enough to take just that metric and say "yeah, you're right, we will accept to be killed by neural nets that no one in the world really understands". I fear people will be highly skeptic about this.
I mean, with all the Tesla in the world, we could have the first death in one or two days. Nobody will care about the thousands other deaths in those 2 days.
ZachF
Active Member
I really enjoy your posts, but sometimes you really need to tone down the "I know more than everyone about everything" aspect.
FWIW, I think Tesla has basically solved the vision problem, which is pretty amazing.
Just doesn't get you to robotaxis.
My interpretation is, they can't predict how soon this is going to become sufficiently safer than humans to declare victory. But they predict that they can do it within the limits of the hardware, tooling and general design they already have.
if that is so difficult, then the example from the karpathy presentation does not work at all. Karparthy specifically said they push rules to the cars, and the cars upload clips onces those rules trigger. One of the example is to look out for cars moving from right lane into your lane.
I am afraid you are complicates things. your rule does not deal with raw reality, your rule works on an environmental model produced by the NN.
Or because it wasn't a two hour test ride, only fifteen minutes. I'm not implying your wrong here, but it's another possibility, and perhaps likely given the time constraints.
Similar threads
- Locked
- Locked
