Tesla owners asking what happens if 'full self driving' isn't real
jalopnik.com
jalopnik.com
Autonomous cars are coming. They just will come first to specific areas where the cars are more experienced, like waymo does in Phoenix, and the cars themselves will be purpose built. As we are building more infrastructure, we could endeavor to do it in ways that make it easier for cars to drive autonomously, with things like grade separated or other protected bike lanes, clearer signage, and other accommodations that will make things easier and safer for human and ML drivers.
It’s crazy irresponsible that Tesla switched on what they did and marketed it as FSD, and it’ll set back self driving cars years when people start getting into collisions.
Will the definition of FSD be slowly reworded, or are they careful to never state the delivery date for it in contracts. Or will Tesla rely on traffic authorities not approving the technology, so they can say that they've delivered but government won't let customers use it.
SpaceX has been wildly successful despite Elon because of that adult leadership, with actual rocket engineers and doers in management. Additionally SpaceX has military contracts that bring with them actual real enforcement of penalties, like 5 years in prison just for making a single false statement.
We need more people like him in charge.
Especially Elon put all his money on the line a few times. And he is definitely not just an idea guy.
The Model T wasn't the first car, but it was the first widely accessible and usable one for the masses. Tesla's innovation is more comparable to that.
That said, Elon should have never been in charge of a public company. He's good at challenging the status quo and should keep doing that in new markets. He's a terrible person to be in charge of an established company (which tesla very much is now!)
The Tesla options are all extremely high performance cars. They're luxury oriented but the interiors left a bit to be desired. Either way the overlap between the market Tesla capture and the leaf captured are almost non-existent.
Hard pass.
This is a terrible idea. I hope it doesn't get anywhere beyond being a research project.
However Musk has much stronger persistence and likability. He can keep promise things and collect money for much longer and be forgiven when delivers much less of what he promised.
I bet Theranos could have still be a thing today if Musk was the CEO instead of Holmes. Instead of failing, they would have delivered slightly smaller machines that work with about the same amount of blood but with better UI and integrated fart jokes.
That said, I don't think that Musk is a fraud. I think that he is amazing in building financial hype machines, thus capable of concentrating wealth that can be used to work on far fetched ideas. Even if he fails to deliver on the high promises, he menages to change the landscape. After all, electric cars and commercial space flight are a thing now, aren't they?
Holmes should have delivered iOT blood testing machines of regular type when being funny on Twitter and could have been fine. When asked when the 300 test from 1 drop of blood is coming, she should have always answered "I'm confident that we will have it in 2 years". Even maybe she could have sold a more expensive machine that "has the hardware to do 1 drop blood test but will be enabled with a software update later".
Musk claimed that reusable rockets and electric cars were the future and people said those things wouldn't happen for 50 years. Now where are we? Saying that FSD is 50 years away seems like a risky bet given what musk has accomplished already.
What I am saying, she should have delivered what she could and promise to devliver better in the feature instead of failing to deliver on the promises.
She could have bought a small manufacturer with all that funding, slap a nice UI and iOT functions on top of it and deliver that. Continue to work on the original 1 drop of blood tests, maybe integrate it into the more expensive version of the machines of the manufacturer she bought, call it beta and use on own risk and the accurate version will be enabled through software update in the future.
Also, being funny on twitter could have been better strategy than imitating Steve Jobs.
Theranos was fraud from beginning to end. They literally tested samples on Siemens machines and pretended that their own machines did the analysis.
They were taking small amounts of blood, too small for the siemens machines and then diluting it down to increase the volume.
So not only were they passing it off as if it was their own machines, but they were providing inaccurate test results leading to incorrect diagnosis.
And I can't believe they got as far as they did with the scam because lots of experts were saying what she was claiming was not possible. Because a drop of blood is simply not large enough to get a statistical result for most things tested. It doesn't matter how amazing or sensitive your machinery is, you can't give a correct result if you simply don't have a large enough sample.
Also, Tesla too failed on cost and quality control issues until they stopped failing.
Yes, she apparently faked the tests but faking till making is nothing new in SV. With all that money, she could have bought a manufacturer or lease buy and rebrand from Siemens by adding something on the top of it. Many big names start by acquiring their parts from OEMs. She could have build on top of some less known European company that sucks at the software UX, for example and keep kicking the can down for the one one drop blood tests. When asked what is going on, give a presentation about the challanges and about the work you did and say that it's going to be ready in 2 years.
How so? The very basis of what Holmes promised with Theranos is impossible based on current knowledge. Musk promised something that at least had a basis in reality, and could be worked on refined (not the autonomous driving, but an electric car).
Holmes just promised something arbitrary and then hoped for a scientific breakthrough.
What he does is nothing new, electric cars or vertically landing rockets, reusable space vehicles are not new but he is the only one who could manage to collect the funding necessary to work on these things again and be the one who is doing it with the current technology(which could end up financially sound this time).
For example, hyperloop is proving to be total BS, no roots in reality. Just look what he promised and what he delivered. The trick to keep party going on, if the party is not over you don't fail. "Hyperloop delivered nothing" is a failure, "Hyperloop delivered something that is nothing that was promised but a car in a concrete tunnel, nice CGI and a model of what it could be" is work in progress.
If you can keep the party going on for long enough, ether the technology will catch up or people will forget the original promise and accept the current delivery. Musk could end up being Tube magnate if he can keep up the funding despite not delivering on the vacuum tube hypersonic travel stuff.
Yes, a lot of what he says is nonsense, like Holmes. But the difference is he also delivered big time on some of what he said. I assume many people are aware of his pie in the sky hopes for the future, but that is different from Holmes claiming she already had a technological breakthrough and succeeded in something all the other scientists with decades more experience knew wasn’t possible, and yet not providing any evidence.
Musk's biggest achievement is changing the Conventional Wisdom[1] of government policymakers in the west from "electric vehicles are a nice idea, but a toy for rich people" to "right, we're phasing out fossil fuel powered cars" in the space of a decade.
That is something I did not expect to see in my lifetime.
You will quibble that lots of other people did the technical work. But Musk had the public visibility.
Especially considering that a ton of these policies are about as old as the Tesla roadster. Must be a hell of a achievement to change policy retroactively.
The reason why we have an electronic car revolution now is because batteries have finally become cheap and dense enough. Maybe the largest manufacturer of EVs deserves some credit, but that's of course never been Tesla.
The world will not accept it when a self driving truck does this. Computers are supposed to be perfect, as far as the public is concerned. If a computer killed a bunch of people, there must be a flaw. The fleet must be grounded until the flaw is found and fixed.
How much better than a human driver does a self-driving truck need to be before people will accept and trust it? 3x? 10x? Infinitely better?
I tend to think the scale trends towards 'infinitely better'.
I believe that for self driving vehicles (at least commercial ones) to truly find adoption there's going to have to be the perception that there's still a human supervisor, even if this is a 'polite lie'. Perhaps it will take the form of lower paid and less skilled human 'co-pilots' who are simple there in the truck or taxi cab as 'sin eater' to take the blame when something goes wrong. Or perhaps one individual can monitor several vehicles remotely (in a totally ineffectual manner). At the end of the day though, I think that for the public to accept truly self driving cars in the near-ish future, there's still going to have to be a human they can blame if something goes wrong.
He needs the money to try and as such needs to present it as something that will happen soon to get that money. I don't think there's any other way to go about it. Who would spend on "we don't know if and don't know when"?
I think there's a lot more trial and error in human tech advancement then we want to admit. Especially for developments that take more than a few years. Failure is always an option, recent popular optimism "anything is possible" is wishful thinking. To answer if it is possible you can only try and see. And to really try you need to believe.
But let's focus on the salary package, it is a optimized for growth not for success, those are 2 very different things.
I am kinda conflicted about this, but when I am calm and analyze the situation from a neutral standpoint I always come up to the conclusion that it takes 2 parties to come up with an agreement . Musk managed to sell his way during the negotiation and arrive to that advantageous salary compensation which is basically crafted for him to achieve by basically keep doing what he's doing (growth and nonstop promotion), can't say anything other than "well played mr. Musk", good for him, he managed to get what he wanted out of the negotiation.
However, the lane keeping assist does not seem like a good idea to me. It sort of works, but I never know if it will lose track of the lane suddenly. When there is an exit or something that breaks the lines, I think of the Tesla crash I read about. When there is a curve in the road, it seems like there is a threshold beyond which it doesn't try to follow. So it amounts to an unpredictable resistance of the power steering, and doesn't reduce the driver's workload at all, in my opinion. Maybe it's supposed to be used on straight highways, but if it worked without attention for a while, that seems even more hazardous.
I wonder if that could be a direction to go in - future cars could let the driver steer, but mostly figure out going and stopping on their own.
However, rather than tried to crash, I think of it as an awareness aid. Blindly following it a bad idea.
I hate blind spot assist. It just makes me concerned that I'm missing something. I adjust my mirrors in accordance with SAE guidelines from 1995 that eliminates blind spots.
https://www.caranddriver.com/features/a15131074/how-to-adjus...
Still won't trust it though
The ACC was great, and very useful in both high-speed traffic and stop-and-go traffic.
But the lane-center tech was useless, regardless of the road conditions. Even with full sunlight and freshly-painted lines, it didn't work reliably.
When I was test driving a RAV4, the dealer told me not to use the lane-centering when in the slow lane, allegedly because people are coming and going a lot. But it was clear that the real problem was that it would keep you centered even as the line on the right veered off for an exit, which would drive you right into an overpass.
So sad that we live in an era of tech where basically everything is expected to severely degrade after 2 years and be essentially unusable after 5.
I'd be much more likely to buy a car that has pretty good driver assist tech if it were regularly updated. I don't want to have a car for 10 years if half the driver assist tech is useless when the car is brand new, and it will never be updated.
The first one by mobileye is actually pretty decent at automating the boring stuff. "traffic aware cruise control + autosteer". It was actually better than the next one for a little while until features caught up.
Some folks with the pre-autopilot tesla cars were able to retrofit a mobileye device on the dash somehow. I don't know if it still works to do that.
Relatedly, have you ever looked at the urban planning of Brasilia? Concepts similar to yours have been tried under the high modernism of the mid 20th century.
Obviously no. Most modern conveniences - electricity, planes, abundant food etc came partly from some form of big tech.
Self driving tech has a chance to fix the 1m+ annual road deaths. Not there yet but give it a while.
Sure some CEOs can the rent seeking etc but that's what competition and regulation are for.
Isn't the USA still waiting for that low price insulin shot?
Horse and buggy enthusiasts hate semis but people who shop at Amazon love them.
The same goes for people who want self driving cars and the infrastructure they will be supported by. If the world goes that way, it's because the people want it.
So we subsidize the cost of high end luxury items? Let's not. If the technology is ready, a faded stop sign shouldn't make any difference.
Most of America has no public transit, and cars are mandatory to transit any appreciable distance (Texas, Florida, most of California, for example). That is unlikely to change, and therefore it is reasonable to discuss efforts and resources into improving environment attributes for self driving vehicles of all manufacturers. Roads are already governed by safety and regulatory standards, there is no reason those standards setting bodies should not take input from developers of self driving vehicles.
> If the technology is ready, a faded stop sign shouldn't make any difference.
Sure, that stop sign (or rather, the roadway segment attributes necessary for the vehicle to infer that a stop is required) should be in OpenStreetMap anyway, which should be used as canonical digital ground truth for roadway intents by automakers, regulators, and roadway O&M alike.
That is pretty weak motivation, can you offer something stronger then "there is no reason" for taking requests from financial interests?
Why don't we invest that money in making current technology economical before we go asking single mothers driving clapped out Ford Taurus's to pay for it.
You conflate scaling Tesla up with scaling the technology up. You want everyone to invest in Tesla so that Tesla can fulfill the promises Tesla made to Tesla owners.
You misunderstand. All Tesla vehicles can use Superchargers. S and X owners paid for the network build out for the 3 and Y buyers (and future S and X owners, which there are less of naturally due to the 3 and Y cannibalizing higher end sales [Musk's statement from an earnings call indicates they continue to sell the S and X as "art pieces" and for "sentimental reasons", backed up by their most recent quarterly report]). Tesla has offered the network to other automakers in a cost sharing agreement, and publicly no one has agreed to that (privately, an automaker is testing interoperability). The point is that someone has to make the investment, those funds have to come from someone's pocket. The US government didn't cough up significant EV charger funds until this year as part of COVID relief [1] (and the goal is for 500k chargers by 2030 ಠ_ಠ). As of February 18, 2021, Tesla operates over 23,277 Superchargers in over 2,564 stations worldwide (an average of 9 chargers per station). Tesla's Supercharger network build was started in 2012 [2].
> You conflate scaling Tesla up with scaling the technology up. You want everyone to invest in Tesla so that Tesla can fulfill the promises Tesla made to Tesla owners.
No. I don't care if you invest in Tesla. I'm not invested in Tesla. It won't even matter if you invest in Tesla (institutional money and index funds are driving the show there, retail is just along for the ride). You're arguing that someone should come along and invest in self driving for everyone's mutual benefit, and that entity doesn't exist. Tesla likely won't even be held accountable for false marketing if they can't deliver on FSD other than a class action (for those who were savvy enough to opt out of arbitration) where purchasers receive a trivial payout, attorneys walk away with a chunk of change, but it's a slap on the wrist compared to revenue and enterprise value (which was able to grow faster than the potential penalty cost of Musk's tenuous promises). Not saying it's fair, nor right, just how it's always played out.
[1] https://www.whitehouse.gov/briefing-room/statements-releases... (Control-F "EV chargers")
[2] https://ir.tesla.com/press-release/tesla-motors-launches-rev... ("Tesla Motors Launches Revolutionary Supercharger Enabling Convenient Long Distance Driving", September 24th, 2012)
Almost exactly a year ago, the article below was published talking about this exact thing. Since companies tried restricting class action lawsuits, there are now instances of death by 1,000 arbitrations.
Judge William Alsup, who also handled Oracle vs. Google, is quoted:
"Your law firm and all the defense law firms have tried for 30 years to keep plaintiffs out of court, and so finally someone says, ‘OK, we’ll take you to arbitration,’ and suddenly it’s not in your interest anymore. Now you’re wiggling around, trying to find some way to squirm out of your agreement. There is a lot of poetic justice here."
The companies made this bed, and they don't seem to like laying in it.
https://www.nytimes.com/2020/04/06/business/arbitration-over...
The chargers should be open to all makes and models of EV. Single brand plugs and chargers is wasteful infrastructure.
Tesla switched to CCS in Europe. They should switch to CCS in the US as well. The next step is to open the chargers to all EVs like the other charging networks do.
You can fuel an ICE car at any fuel station so you should be able to charge an EV at any charging station. Not being able to do so makes EVs less practical than internal combustion vehicles.
Like we did for regular cars when they first became a thing?
This is a fundamental thing. Robot cars that can't interact with humans without killing them at a greater rate than humans do are not self driving.
Humans use road signs and lights and... We created separate, additional infrastructure for ourselves as drivers. Why wouldn't we add infra to "help" autonomous cars? Can't a car that senses RFIDs at every corner be considered autonomous, full self-driving? How is that different from sensing road markings via camera?
If the issue is just a terminology one, to distinguish "full self driving" versus "full self driving with fully self-contained reasoning" or whatever I must object - that's not how these terms get defined.
Tesla is arguably misusing the term "full" because it doesn't operate to the level we'd expect; not because of whatever resources it uses to function
If this is too long -
tl;dr: I think the absolute majority of people won't mind calling a car that requires _any_ additional infrastructure (including communication with other cars) a "full self-driving" car. As long as it drives itself 100% of the time without a human driver, it's full self driving.
If a robot can't deal with a human cyclist or someone crossing the street the that is not remotely self driving.
That is not the bar to pass here, as robot drivers have already surpassed it. L5 is way beyond that.
So, when you said that "Tesla cars" kill humans at a lower rate than "human drivers", could you provide a source for that? Or did you mean the regular claim by Musk about teslas being safer by 100% than an average driver, which means that an average tesla car while running autopilot in almost always highway conditions in almost always sunny conditions has lower fatalities than the average human driver, who will skew older and more accident prone, driving on average roads, in average conditions, with the average 10 year old car with less safety features.
Never in any of these statements have i seen any discussion of pedestrians deaths, but it makes sense right? How many pedestrians are on highways where autopilot is mostly used.
Trains can't pass each other, so you will need multiple lines with lots of trains, and they don't serve all transport needs unless you manage to get rail outside all shops for instance.
Roads outside of shops is a small minority of all roads, and we are currently undergoing a change in how city planning is handling commercial districts, with movements towards the edges or online.
There are however major problems with train networks, one being that lack of redundancy. With road maintenance you can shutdown a major highway and redirect traffic temporary to smaller roads with significant reduction in speed. With trains you either have the option to go around the maintenance, or you replace the train temporarily with road based travel.
A second big problem with train infrastructure is the lack of an obvious upgrade path. With roads you build a new highway when security, air pollution and noise become too severe for smaller towns around the old road, and the better road also adds higher capacity and faster allowed speeds for transportation. There are so many incentives to upgrade old roads, and with an increasing population it makes traffic heavier for old roads, speeding up demand for better and bigger roads that runs outside the towns.
For rails, there is mostly just an economic argument in favor of building more. The reduction in car travel is not significant enough compared to building a new road outside the limits of the smaller towns.
At least in the US, my major concern would not really be building it right the first time (though there are plenty of places where we don't even do that), but maintaining it over time.
Many jurisdictions can barely keep the paint legible on the roads, so autonomous cars are just going to have to deal with poor or deferred maintenance.
If an autonomous bus costs twice as much I think it still wins out. Maybe you still need a conductor to make sure people pay and don't act antisocially but such a person can probably be much cheaper/less skilled.
Why do you need to specify the "unionized" drivers? Do they cost more? Why? because they are paid fair(like paid extra if they have to work on Christmas or extra hours)? Or maybe they union negotiated to keep the salary in sync with inflation?
Las Vegas had autonomous buses for a while, but they're gone. Too slow and the routes were very limited for some reason.
I guess if you purchased say, 40 honda civics?
There’s a lot of other considerations to consider here as well.
Cars and vans would both take up more space for the same amount of people which means more traffic which leads to slower speeds and more noise and air pollution.
Cleaning all these cars and vans would require more workers than cleaning the equivalent number of busses either by capitol or by riders.
Cars and vans require more space to park.
There’d be far more maintenance for cars and vans than busses.
Busses are equipped with handicap ramps, bike racks, and seats to accomodate the differently abled, as well as payment processors. These features all drive up costs. Tech only solutions can prevent access to transportation for indigent and differently abled populations.
More cars and vans means more road wear.
And of course the lifetimes of the vehicles may differ adding in other costs.
In reality, they compromise. I live in a not-very large city. I used to rent an apartment in a suburb to the southeast just a few miles out, and the bus only made one trip in the morning to the city and one trip in the afternoon back, and if you didn't work the exact hours that you could catch the bus, like my job was 9-6 or 2-11, etc., you were out of luck.
Now, I live within city limits, in the southwest, but there is no bus route going north-south to where I work (or did before covid), because most of the traffic goes east to downtown and back, so that's how all the routes go. There aren't many routes or stops to the north and west.
When I was living in an apartment in a slightly different suburb, there were more frequent buses, but they were almost empty when I rode them. You can't assume that they're saving resources in any circumstance.
I'm saying that there's a tradeoff between investment and cost effectiveness relative to individual transportation.
A bus gets like 6 mpg. You have to have a certain average number of people on it per time or distance for it to be more efficient than a person in a car.
A network is more valuable the larger it is. But the more routes you have, the less traffic each one will have.
Being willing to invest is irrelevant, unless you think investing in something that makes people worse off is a no-brainer.
American governments tend to prioritize investment and spending for individual transportation over public transportation. The numbers get really crazy when you actually factor in the costs associated with 'free parking' - including requirements that new developments include a minimum amount of parking spots for each housing unit - and all the other subsidies car transportation enjoys.
[edit] > Being willing to invest is irrelevant, unless you think investing in something that makes people worse off is a no-brainer.
The thing is, public transportation isn't meant to be necessarily cost effective or even in all cases fuel efficient at the point of service, it levels the playing field. Poor folks currently have to buy a car to get to work. That's a huge investment in a depreciating asset at usurious interest rates. Public transportation reduces inequality and bolsters the economy indirectly, too.
[1] https://www.pewtrusts.org/-/media/assets/2014/09/ff-transpor...
The average number of people you need for simply the fuel is like, 5. But as far as all other factors, a lot are fixed costs, so number of trips is more important that average people.
Not to forget, again, electric buses exist and work well.
The ranking as I understand it is Waymo, Cruise, then everyone else.
https://www.youtube.com/watch?v=IxuU5L2MEII
Note: He regularly runs a Twitch livestream coding stuff in the space, recordings are on YouTube, very high signal low noise stuff there.
If you're interested in the state of fsd listen to this andrej karpathy podcast. Not Ford's PR department
https://twitter.com/pabbeel/status/1374450544475992064?lang=...
Waymo can scale as fast as they want to via partnerships like the one they have with Chrysler already. Other companies will build the vehicles and add the hardware, Waymo will handle the software. Not everyone has to do everything in-house like Tesla does.
https://news.ycombinator.com/item?id=22632588 -- comment a year ago on the end of Starsky Robotics (good blog post)
https://news.ycombinator.com/item?id=22633543 -- further down in the thread quoting Chris Urmson from March 2016, and also Bill Gurley from 2018
Sure things could have changed, but by all accounts they haven't, given the lack of deployment. They took 5 years to get to 90%, another 5 to get to 99%, and will take roughly the same amount of time to get to 99.9%, etc. And that doesn't mean they're done.
(I have been following this for at least 10 years, and I also worked on a college project in autonomous vehicles 21 years ago. It was not good, but the deployment and philosophical issues are identical to the ones back then IMO.)
You're absolutely correct, but there are far more people who want to make the argument that Tesla's ahead because they're "beta" on roads and "don't need LIDAR".
There's optimism, and there's reality-distortion field. I think a lot of people bought into the latter.
Waymo and Uber seem to be of the opinion that car ownership is going to disappear which continues to boggle my mind. The US absolutely needs car ownership for anyone not living downtown in a city, which is over 50% of the US population. People shouldn't be looking at them as examples.
The fixed costs of owning a car are expensive, but typical city/suburban usage ignoring work commute easily makes the costs worth it. The time expense of using public transit or the dollar expense of relying on lyft/uber are simply too high.
Self driving cars have some advantages:
They reduce the cost of sporadic non-peak usage. For some this will make the work commute on public transit a dollars+time net savings. This pressure will be especially high on 2+ car households, where the economics of a second car will be abysmal.
Self driving cars can act as a much more efficient feeder to public transit. At any given time, especially during peak usage, there are a bunch of people going from approximately where you are to approximately where you're going.
We aren't taking advantage of this today because of two problems: Coordination is hard and requires a critical mass. Any time stopped just kills average speed, greatly increasing time to get to a destination (this is why busses are soooo sloooow.)
Self driving cars can take people from their houses in suburban neighborhoods to coordination locations. Hop in a small bus/shuttle with a few others. Wait no more than a minute or two as cars arrive. Then the shuttle takes everyone to their destination (if it's a high density city where the destination is within a block), or the reverse of a feeder, where single person taxis are already there waiting to take you, with a pause as long as it takes to get out of a vehicle, walk a few feed, and get in another.
Public transport and bikes are not feasible in Houston or Los Angeles, due to low density and hostile roads.
Going back to your argument, I believe all of this would also be possible with cabs, from a user's perspective the only difference between an autonomous car and an Uber might be price (over the long term)
High speed rail between large agglomerations like SF/LA seem like obvious improvements,but seem impossible to build in the USA.
For many people a robot taxi is a viable alternative to car ownership, but for plenty of people owning a car is always going to be a better option.
> Why does the US need car ownership?
Cause freedom. You're talking about the U.S. Why does the US need gun ownership? Cause.
> wouldn't fleets that are constantly bringing in revenue be more efficient than everyone owning a car and driving maybe 2 hours a day?
I hear this again and again and I wonder how detached from reality people are. There are many reasons this won't happen. People have different standards of cleanliness. People have different standards of how much scratches they tolerate on a car. People view cars as status symbols. The list goes on.
People buy condos even though a rental would financially make more sense. But they want to own something they can call theirs and do with it whatever they want or keep it pristine to their liking.
Is it other people detached from reality? It seems unrealistic to assume 100% of people will not want to use such a service.
> There are many reasons this won't happen.
The very fact taxis and Uber exist means there is a demand for such a service.
When I lived ~10 miles outside of a town, I would have been fine not having a car if I could reliably summon cheap transport. In a small town, all the same. The amount of waiting matters, but not that much.
While I own a Tesla model 3 with FSD I am more impressed by what is shown on youtube by beta testers than by anything my car can do. Simply put, Telsa only recently changed their entire FSD suite to include maintaining an active awareness of all labeled objects which in itself is greatly improved as my car clearly does not see everything.
Think of it this way, current versions work on a very short moment of time so and don't have labels for everything it can see which leads to missing parked vehicles (as in how did that survive this long) and more. My favorite example is my car rendering cones around a parked cable/phone truck but not the damn truck! It also had fun with a garbage truck doing a great job rendering the guy and cans and with the truck not always there but immediately real once it moved.
As for buying FSD, don't. Regular autopilot provides the driver assistance people can actually use on long trips. Which is, keeps you safe if you get distracted or worse. Plus there is still an issue with trade ins to Tesla, pretty much they were assigning ZERO dollars for FSD. So until Tesla provides a transfer fee or puts FSD at the account level I am never buying another one.
FSD needs to be at the account level. It would actually make Tesla more money as they could then sell a subscription for those who only need it for trips.
Directly relevant to the OP:
> 714 days since Elon Musk said there will be a million fully autonomous Tesla robotaxis in a year.
> "I am selling almost all physical possessions. Will own no house. [...] Don’t need the cash. Devoting myself to Mars and Earth. Possession just weigh you down."
> "Based on current trends, probably close to zero new cases in US too by end of April" (Posted Mar 2020)
Unbelievable, these are just the ones from a year ago. People that take him seriously, could you explain why you do?
His predictions on Crypto and BTC now seem even more ridiculous and pump'n'dump-y.
I'm not an Elon fanboy, it's just a little bizarre to me that people mark this guy up as a crazy liar who shouldn't be taken seriously, meanwhile many of his bold claims have actually resulted in advancing tech and science in staggering ways. Like, people were saying the same thing about his claim to land first stage boosters, and here we are.
Side note, apparently he is still actively selling off all his possessions.
F9 is also incredible and has made the impossible possible... and Starlink is one instance of that, which also exists and is usable.
You can't really highlight his failures without also giving credit where credit is due to his (and his organizations') successes.
I'm really curious how he gets such smart and talented people to work at his companies.
Meanwhile, Falcon 9 and Falcon Heavy have made numerous successful and commercial launches and landings.
Also consider that it took NASA about 9 years to fully develop the Space Shuttle from ground up and they didn't lose a single vehicle during testing - 50 years ago!
This is not accurate: reusable orbital first stages is a SpaceX first, and the cost savings it yielded enabled Starlink, which is another SpaceX first (high speed, low-latency, low-altitude LEO satellite internet).
Of course, so has every famous idiot.
That describes every user on this website...
Elon Musk is demonstrably very good at some things - in particular running companies. I can't think of any other person who has successfully started multiple billion dollar companies. Which isn't to say that they don't exist, but... (see next paragraph)
Elon Musk is also much more willing to speak his mind than most successful entrepreneurs (which is why you hear about him). Sometimes he is rather stupid (see paragraph 1), sometimes he is not. Often I think his statements (and to a large extent probably his actual opinions as well) are biased towards what is good for his companies. The information you can get from when he is not being stupid or overly biased is probably worth putting up with the stupid.
Finally he's just involved in a lot of cool technology... and however else you feel about him you have to admire him remaining that involved when he has more than enough money to fuck off and retire.
Maybe one way to drive home the point is to just answer a lot of yes/no questions about him:
Do I think he's overselling autopilot, yes. Do I think he's wrong about bitcoin, yes. Do I think he's wrong about how to run companies, I mean probably also yes, but on average less wrong than basically anyone else talking about it, it's a complex topic but he is good at it. Do I think he's wrong about how to build rockets? Like the answer to companies. Do I think he's wrong about how to build factories? The same again. Do I think he intends to put people on mars? Yes. Do I think he will succeed? I put it at above 50% probability, which is far higher than I think of anyone else. Do I think we will get his self sustaining city? Almost certainly not. And so on and so on.
Its still a good skill picking winner in a hyper competitive environment, but he's closer to an average trust fund kid than a jobs or a page.
Steve Jobs did, I believe there are a couple others too. I agree though, the list is short.
Also they should also update the post about the 500k Teslas/year. 185K in Q121 places Tesla right inside 500k/year territory.
A while back, I was like "This Elon Musk guy is really an example of what we need more of in the world".
It makes me cringe.
This was the days of solar rooftiles, Boring company, going to mars soon, the hyperloop, battery technologies, self driving vehicles.
All of it was complete horseshit.
The worst thing of all is that, I can hold my hands up and be laughed at but since then there are armies of Elon fans that don't recognise it.
In for a penny, in for a pound.
EDIT: Even on hackernews, there are people with this fetish. "It isn't his fault, the public is to blame for thinking that fully automated self driving cars, meant fully automated self driving cars". Ugh.
The difference is he has been involved with three large successes - ebay, telsa, and spacex. This success/failure ratio is leagues better than just about anyone else. And he did this in three very different industries.
So his self driving car effort will likely fail, or at best be very late.
But we do need more people willing to take chances like this who can also execute as this is rare.
The thing people need to realize is that people can start hitting misses at any time.
Elon Musk is perhaps a con man (in the fully expounded sense of the term "confidence man"), but particularly one who believes his own bull.
If he wasn't like that, he might not have done some of the worthwhile things he's done. I definitely wouldn't take anything he says at face value, but I think there's value in him being like that.
Consider Toyota. 25 years ago their upper management was capable of fostering the development of hybrids. Now they seem to be unable to change direction.
VW is doing a better job. Probably because Germany threw some of the previous management in prison.
Im a fan of what he was. Not a fan of what he has become.
Like:
> "Building 100 Starships/year gets to 1000 in 10 years or 100 megatons/year or maybe around 100k people per Earth-Mars orbital sync"
That's still the plan, and they're actively working toward that goal, and no deadline was set.
From what I can tell, the ones that are jokes are the ones he said a few times but made no progress towards. Like the self driving ones. And the ones where he throws paying customers his product killed under the bus.
I don't find that kind of joke funny.
Good question. For me personally it's always been obvious and I've yet to be wrong, other than the time he said he would drive cross country on autopilot by a certain date. He's very obviously been wrong on Autopilot/Full Self Driving dates. I read every tweet he writes and I have a good sense of his personality having watched most interviews with him since the early 2010s when I originally started following him via SpaceX.
> Like the self driving ones.
I don't think the self driving ones have been jokes. He just really underestimated the difficulty of it. He has many times tempered his well known self driving tweets with less talked about tweets talking about 9s of reliability and how hard it is. People naturally only see his most wild claims reposted the most so assume most of what he says is the wild claims when that's generally a tiny minority.
And they have been making progress towards it. The system as of a few years ago compared to the system today is dramatically different. At current rates they're probably 2 years or so away from a system that matches Waymo's system, but works almost everywhere in the US, not just one city.
> And the ones where he throws paying customers his product killed under the bus.
To be honest... I'm on the side of Tesla on these. I'm not a fan of blaming a company when people misuse a product that's gated behind warning labels that say to not use it improperly. So when people die and he says "people shouldn't have been using the product this way" he's right, they accepted the risk when they turned the feature on past the warnings.
And not only is the marketing fraudulent, I also found the manner in which they let 'beta test' software loose on streets completely irresponsible.[1]. I have no idea why that didn't have consequences either.
But in any case, one would simply sue the manufacturer here, not its weird CEO personally (this might lead to the _shareholders_ suing the CEO, of course).
The problem here is obvious, if the automatous system is incapable of handling 10% (or less) of some niche scenarios on the street due to some intelligence limitation/blind spot when contrasted with that of the average driver, then automatous driving will never get approved.
And given that the current machine learning is nothing more than overhyped statistical algorithms with fast processors and lots of data. Then, we can say it is highly probable that what Elon Musk promising is....vaporware.
And knowing Elon Musk from twitter and his gambling tendency with sci-fi trends, we can also be sure he knew that, he used the AI hype to sell his cars, a great marketing guy, I actually like really him, I'll side with him over many politicians, but he will surely blame it on the regulators afterward.
But I for one will be happy if it works well in highways, or interstate drive, I think it would be great progress.
I don't think we'll ever get to fully autonomous travel in my lifetime, most likely on the interstates and in cities, but as soon as you leave the cities, there is way too much unpredictability for AI to handle IMO
I'm also not sure how those autonomous system are planning to handle scenarios that would require ethical judgements.
Say we're driving at the country side, on a beautiful mountain on a sunny day and the sea is beside use. Music is fully blasted, car is driving itself, and we are on a way street surrounded by a cliff and mountain rocks on the sides. Our autonomous driver have two choices, move forward or slow down, any right/left turns will result in the injury/death of the passengers.
All of the sudden, our car noticed an animal crossing the street. Now, what should the system do? keep going and kill animal? Let assume that we programmed it to slow down in order to minimize the impact knowing that it will end up killing the animal, but that is tolerable loss, the system computed. Now, what if it was a family crossing? Should the car just slow down or attempt to turn? Again, who would decide this? the engineers? lawyers ? how will this be handled at the court? should the car be programmed to protect it's body at all cost, a self centred car? would regulators allow a self-centred car to sell in the market? or should it be altruistic car? what if it made a poor judgment out of good intentions etc etc.
I don't think engineers were ever presented with challenges like that. Historically, engineers were building machines to be used in a controlled environment under the supervision of their users, and if it was autonomous, then it would have very specific goal.
Perhaps this would be the start of robotics ethics and law. But I don't think societies today are ready for this undertaken, I'll be very curious to learn how those challenges are being addressed and would appreciate any references.
Full self driving isn't necessarily safe. What you want is full self stopping. Full self stopping has the potential to dramatically reduce road fatalities.
Rudimentary versions of FSS are fitted to many vehicle models (e.g Toyota Camry) already. Edit: such things as autonomous emergency braking, rear cross traffic alert (and reversing auto-brake), blind spot warning and lane change refusal if there is a vehicle in the blindspot, pedestrian avoidance -- all these and more are available today and all are better than nothing when it comes to reducing accidents.
Tech to avoid most road accidents is there, proven and operational for over a hundred years: trams, trains, metro's, buses* etc.
We often make the tradeoff to risk our lives, because of the (percieved and real) downsides of choosing the safer option.
This is not meant as a praise for PT, but to put down another datapoint on why people deliberately choose the more risky option. Both in (political) investment and on a daily base.
*while certainly no to 100% safe, public transport, and slower transport like bikes and walking, is almost always safer than private vehicles, and where not (e.g. buses), many accidents happen because of the shared space with those private vehicles.
I was trying to say that traffic is dangerous, which your experiences seem to confirm. And that we, as society, somehow, don't push back against, but simply embrace and acknowledge. Every morning that we get into the car to work, again. And again.
[1] https://www.cdc.gov/nchs/fastats/leading-causes-of-death.htm
FSD is perhaps the ultimate solution but there are common-sense road fixes that can be applied, a la Oslo, Norway which achieved zero annual traffic fatalities:
https://thecityfix.com/blog/how-oslo-achieved-zero-pedestria...
FYI here is the article for the death of my friend. The philosophical differences between DOT and safety advocates are quite infuriating:
https://pamplinmedia.com/scc/103-news/199653-buttressing-saf...
[Don Hamilton, Oregon DOT public information officer] "There's nothing that engineering can do to stop drivers from harming themselves, and there's no way we can engineer from an irresponsible behavior."
[Roger Averbeck, Chairman Southwest Neighborhoods Inc. Transportation Committee] "It's a very wide roadway (with a) wide cross section, very wide traffic lanes… ong distances in between traffic signals and safe crossing locations… and particularly in this… historic section, the land use is set way back away from the road with parking lots between the businesses and sidewalks. …This creates this wide, open environment that is definitely a factor in people’s behavior. People do not drive that fast… through the middle of Multnomah Village or even on Capitol Highway through Hillsdale in the middle of the night. It's the environment that people are in that encourages speeding."
I wonder which worldview the HN crowd will more closely identify with?
The article says he was driving “80 to 100 mph in the 9000 block of Southwest Barbur Boulevard” in a Prius. This road has multiple stop lights, pedestrian crosswalks, and a speed limit of 40 mph. He crashed into a bus stop that could have had pedestrians.
Assuming it was a new 2013 Prius, “going from 50 to 75mph takes a slow 8.3 seconds” (from an article about that car in 2013). So he was driving at least double the speed limit down a commercial corridor with a top speed of 40 mph, for an extended period of time (at least 20 seconds over the speed limit to hit 80 mph), with a car that was not designed to be driven that speed. He had no respect for the law, his safety, or the safety of others.
You could say we need better laws (e.g. full self driving required) but he didn’t care about the laws then, and there will be outliers in the future won’t care about the FSD law and will override their cars to get on the road. For the next 30 years there will be millions of non-FSD cars mixed in with FSD, and then there will be classic cars.
One could argue the road needs engineering changes to make the road so it’s impossible to drive faster than 40 mph, but how many ambulances or police vehicles use that road and need to go faster in true emergencies with the added safety of sirens and lights? And right now the road is engineered for a top speed for 40 mph. So the engineers put a speed limit of 40 mph. The vast majority of drivers on that road follow the speed limit (or close enough) to make 40 mph acceptable.
Life doesn’t come with bumpers. Part of it is figuring out which things have lasting consequences and which things are reversible and recognizing the very real and often precipitous difference. Along the way we learn from others what the difference is, but it doesn’t mean we should necessarily make changes to the environment based on others’ mistakes. We can’t fill in the Grand Canyon because someone fell in. We can put up signs (speed limits) and guard rails in high risk areas, but can’t always re-engineer the environment. And even if it is re-engineered, someone will come along and exceed the engineering design of that environment.
Maybe increased policing on that corridor would make sense to enforce the existing laws.
This is Monday morning quarterbacking it 8 years later. I’m sorry for your loss. I’m sure he was a good guy. Sometimes people make mistakes and no matter how much one would like to say “if only it was designed differently” or “maybe we could prevent this in the future for someone else”, there’s a point that people need to be free to make the decisions and deal with the consequences. It doesn’t make it any less sad, disappointing, or frustrating to deal with.
My lease is up in a bit and I’m considering a Model Y or a Mach-e. Something could come out soon that’s a contender, but I can’t imagine what.
Can’t imagine opting in for FSD in 2021.
The actually usable stuff - namely autopilot - is available without FSD. All this said, the car is great, even though FSD is a con.
I've been thinking about FSD for a while but I'm less and less likely to pony up $10k for it. I would like to get several other features that are bundled with it though. Things like Autopark, Auto Lane Change, (Smart) Summon, and Traffic & Stop Sign Control[0] would be super valuable to me and if they were available separately I would definitely be more inclined to purchase them. Maybe that's why they are not un-bundled though; no one would purchase FSD otherwise.
For me, “foreseeable” also includes FSD actually becoming a real software package we can download for a fee. I expect that package along with Tesla Network would be worth somewhere around $US30k/year for the first couple of years until no more taxis are driven by humans.
I also expect that either FSD/TN will become available as a subscription option where I can send my car to take part in Tesla Network while I am not using it, or I will be able to finance the purchase of a new FSD capable vehicle and the car will earn me enough money to cover the cost of ownership and participation in TN ie: it will cover the cost of commercial registration and insurance, tyres, maintenance, and payments on the principal + interest loan used to purchase the vehicle. I don’t expect it to be printing money for me, but I do expect a better ROI than having my money in a term deposit or superannuation (somewhat like 401(k)) account.
IMHO once any vehicle hits the commercial market with fully autonomous capability, the concept of private ownership of passenger vehicles will be dead within ten years.
I fully expect this Model 3 to be the last car I buy for my own use. I anticipate robot taxis being reality within a decade, even for limited scenarios such as inside specific geofences covering the “safe” roads inside a city’s metropolitan limits, and the well maintained highways linking those geofences regions.
In the meantime I am enjoying the comfort of a car designed for long distance travel with pleasant seats, quiet interior (quieter than my Jazz), and reasonably attractive styling.
I like the feature but in its current state is not worth the money.
Unlike many many others here, I think self driving is closer than we think. Mostly because humans are horrific drivers, not because robots are great. Even a pretty bad robot is better than plenty of people on the road. If robots drove badly and made mistakes but didn’t, say, ever run a red light or pass on the right, they would probably be safer. Pragmatism over idealism. I think Tesla is certainly liable for their marketing claims - and I don’t mean this as a defense of Tesla whatsoever - but are we all going to pretend we believe marketing? McDonalds burgers don’t look like they do on TV and AI required careful supervision - that’s life.
If the rumors come true and they offer it as a monthly subscription service, I might buy it for a few months at a time for summer road trips. Navigate on Autopilot would make a long drive like that more comfortable.
But otherwise I made the decision on the feature set it has now, and I can’t justify spending 20% of the car’s price on that.
If they brought out the “enhanced autopilot” feature they offered in the EU for roughly equivalent to USD $4000 I’d jump on that in an instant.
The best thing is to be objective. Tesla's FSD implementation based on their videos suffers from one major fault: object detection. Why ? Because they rely on low resolution cameras and even the best algorithms we have today are very far from perfect. Without detecting objects 100% of the time you simply can't do full service driving.
That's why companies which have LiDAR i.e. Cruise and Waymo are so far ahead despite not having as much training data to work with nor the head start that Tesla had.
So my opinion would be by all means buy a Tesla but the facts are that FSD is not done and there is a good chance it will never be done without a hardware upgrade.
Here's one example of NoA trying to run me into a wall. https://www.youtube.com/watch?v=VRSbJ6uxGck
Dont get me started up on autopilot not detecting stalled cars in car pool lane.
1) wow this is not ready for prime time at all
2) wow I'm astonished what they've accomplished so far without lidar
Both of which are true, based on multiple views of FSD beta so far.
If it is 90% accurate (and probably 99% accurate for time driven vs a human), that is pretty bad/unacceptable from a statistics, safety, and human comparison
But 99% accuracy may show a path to five-nines or whatever is needed to replace a human. I honestly don't know if it does. I can't sit here and say "no way" since it did identify bicycles and pedestrians and lanes and curbs and lines.
Edit: I waited in line in 2016 to preorder a Model 3. And have been driving one for 3+ years.
On the other hand, I've seen what old age does to a person and their ability to perform complex maneuvers. As I mature, I might be more willing to have the car help me from time to time. Things like keeping in the lane while turning, or parking properly, or even just pulling into a garage, will be more and more difficult and it would be nice to have some help.
There's 1.3m Teslas sold - how many with a $10,000 FSD upgrade? That's got to be single-digit billions of revenue, but that's all I'd expect them to be liable for.
Somewhere there's a liability account paying out to individual owners who have asked for refunds. If it turned into a class action, I'm sure Tesla are ready to pay customers back and it will have been pretty cheap finance for them.
And aside from the small print does anyone really think they were guaranteed to make working L5 technology? I'm an optimist but I think it's always been uncertain as to whether they could succeed.
Ranging from 8 to 9/10 depending on the site.
My bigger concern is charging infrastructure. Tesla is just so far ahead in that area that it will be difficult for others to catch up in the short term. Especially if they don't get their act together and concentrate on it.
Not really. There are many CCS charging sites worldwide. The fastest CCS chargers are 350 kW (with some at 400 kW), which is faster than Tesla's fastest chargers.
It's too bad that Tesla hasn't yet switched to CCS plugs in North America like they have in Europe. If Tesla switched to CCS then drivers of new Teslas would have access to many more charging sites with no need for adapters.
Unfortunately he bought into deep learning hype where AI is always going to be "almost there" (that is until funding runs out).
Which would be fine if he's a private citizen and AI enthusiast, but putting this brainless FSD monstrosity on public roads is irresponsible and going to kill people. Check out FSD videos on youtube, this is scary stuff.
And it's not about moving fast and breaking things, tweaking the models for endless "edge cases" won't help with the infinitely long tail. The whole deep learning approach is fundamentally flawed. This "AI" has no idea what it's doing.