Hacker News Reader

Comment by quickthrowman | Hacker News Reader

Anyone hawking FSD is a stock promoter (liar). Uber/Tesla/Lyft. Elon Musk claims Tesla FSD will be L5 by end of 2021, but they recently filed a document with the CA DMV stating Tesla FSD is L2 automated.

Which one do you think is more realistic, staying L2 by EOY 2021, or magically leaping forward to L5 FSD in 8 months?

The only company working on self-driving that I believe when they issue press releases is Waymo, because Google isn’t trying to juice their stock price all the damn time, and they have operable robotaxis in AZ. I don’t think Waymo claims L5 capability either.

View on HN

It's those edge cases that make me think real FSD (vehicles with no steering wheels) is a decade-timescale problem.

Figure out 99.9% of driving, but otherwise take a family off a bridge when the sun is blinding the camera? Still need a steering wheel.

I used to live in Mountain View where Waymo tests their cars so I would observe them quite often. I sat a railroad crossing with one, thinking about what could go wrong. What if the signal didn't work, but you could hear the train coming and see it way off to the right or left if you looked. Are Waymo cars listening for trains? Or are they looking at railroad guards only? I suppose they could program the train schedule in, but what if the train system changes the schedule? That could be adjusted for, but how much would I trust this system? Not enough to take a nap behind the wheel.

In in Whitby, Ontario, Canada, a dad and his son were killed driving a car across a railway crossing with no gates. The dad (driver) was on his phone at the time and it is believed the phone blocked his view of the train and he was talking to much to hear the train.

The person he was talking to said they were talking like normal and suddenly the phone cut off.

Computers probably can not be worse that the people already on the road.

Even if you're right, that's the same argument for airplane safety. The problem is being in control versus putting your hands in the hands of an unknown third party.

The psychological, ethical and legal implications are completely different. If tomorrow I drive a car and run a kid over then I'll be in trouble and you'll probably never hear about it. If tomorrow I get in my Tesla self-driving car and it runs a kid over then you'll hear about it everywhere and Tesla's responsibility will be invoked. Because whose else?

The bar for self driving cars is not being as good (or bad) as a human driver, they need to be orders of magnitude safer in all situations. They need to have airplane industry numbers, not Average Joe drunk driving numbers.

If the bar is "better than someone talking on cell phone while crossing a railroad crossing without looking," I'll stick to manual driving.

Self driving needs to be better than me, not better than average.

Hell, it needs to be better than you with a "train is coming, I'm going to slow down" driver assistance.

If you tell me that no such driver assistance exists, just apply it to scenarios where it exists, like lane keep assists and automatic emergency assists

It would be interesting to put a black box in a car tied to the driver. Anyone who shows patterns of driving worse than average gets a self driving car.

If you are better than average you can keep your dumb car.

Does the average driver even look both ways at the railroad crossing? I do every single time. Such a low-cost preventative measure that could very well end up saving my life one day.

At least where I learned to drive, not only was stopping to look both ways at railroad crossings taught in driving classes at school, it was part of the written test to get a permit/license, as well.

How many people on Hacker News think self-driving cars will be a better driver than themselves in the near future?

Probably me. Waymo may already be in most situations.

I drive about once or twice a year, always in a rental, and consequently never really know how well my car handles or its dimensions. That, plus generally being rusty, tends to make me fairly nervous.

Yet, nevertheless, I am in charge of a 2000kg block of metal hurtling around pedestrians and cyclists at 70mph.

Scares the shit out of me, to be honest.

What are your circumstances where you simply couldn't get a driver? If I could at all afford it, I would in that case same as I do in very remote 4WD territory.

I don't own a car, so I only ever really drive on holiday... or in extreme situations.

The last time I drove was taking my rabbit to the vet at 2AM on X-mas day, where the vet said due to Covid I would have to wait outside. Not really practical when the temperature is below zero, so I used Zipcar!

Besides that, mainly if I'm going somewhere for a week with poor public transport and high distances to cover (impractical to cycle/walk).

Honestly, I don't think self-driving cars will /ever/ be a better driver than I am. But that also hinges on what you mean by "better". There may reach a point in time when self-driving cars have a lower statistical likelihood of being in an accident than I am, but I doubt that they will ever do that while getting me to my destination more quickly than I do.

Both of those things relate to the fact that I have significantly better situational awareness and driver training than the vast majority of people, including significant time on track racing cars. To match my situational awareness it will not be enough for self-driving cars to rely on their own sensors, but they will need to communicate with one another and act in concert, otherwise my reading of traffic patterns and looking far ahead in the distance will always beat it out for the quickest path through traffic, or the safest response.

To me, where I see a benefit for myself personally, is that self-driving cars don't necessarily need to be better than me, they just need to be better than most, and turn driving from an active to a passive activity. It means I can be engaged in other things. But, that's also the rub. I /like/ driving, so I don't want to exist in a world where I am /forced/ to have a self-driving car to use the roads, as I still want the ability to go out and enjoy myself from time to time. But for a daily commute, turning driving into a passive activity would make me a happier and more productive person (other people are horrendous drivers and it frustrates me observing their unsafe and careless behaviors).

I'm sure almost nobody even outside HN thinks that self driving cars will be better than themselves. Everybody thinks they're above average drivers.

That's a psychological problem that self-driving cars will have to overcome. They need to be so obviously better than any human driver for people to actually consider them. A handful of fatal accidents that people think they could have avoided as a driver and they won't want to get on board.

People are bad at risk assessment, but that's a fact of life. As a result, self driving has to be better than what people _perceive_ as their risk of being in an accident while driving, not what the actual risk may be.

Not a problem if you only sell to the bottom 50% of drivers.

Only a small minority of drivers consider themselves to be in the bottom 50% of drivers, so you might have trouble with that strategy.

People are great at some things that computers suck and computers are great at some things that human suck.

Machines doing horrible mistakes that no conscious person would ever do is problematic because we don’t have reliable error correction methods for that kind of mistakes.

A FSD car will never accidentally press acceleration pedal whey trying to press the brakes or loose control when trying to read an SMS. Instead it will mistake a bird for a train, hit and run someone and it would be like “something slowed me down, are my batteries degrading?”

How do you deal with a driver that fails to understand what’s going on?

A human being can drive a million miles without a serious accident. 2018 saw 1.1 deaths per 100 million vehicle miles driven[1]. Human beings are extremely good at driving safely overall.

[1] https://en.wikipedia.org/wiki/Motor_vehicle_fatality_rate_in...

Yet driving deaths are the leading cause of death in older children and young adults.

It's not that we're bad at driving, I suppose, it's that we do an awful lot of it and it's quite a risky activity - especially if you're not 100% engaged.

That being said, if the US cared about road safety, there's quite a lot they could do to improve it already. Many techniques have been used successfully in other countries to significantly reduce road deaths, for example, a nationwide vehicle safety inspection standard and lowering speeds on urban roads.

The fact that these aren't being done leads me to think the concerns about road safety are actually rather irrelevant for the country as a whole. When self driving cars exist and are convenient, people probably will switch to them, as long as the accident rate is vaguely comparable.

> Yet driving deaths are the leading cause of death in older children and young adults.

All vehicle-related deaths in the US average around 32,000 annually, of which about 50% of those are alcohol and/or drug related, leaving the true accidental vehicular death number around 15,000 annually.

The problem we're trying to solve with FSD just doesn't exist at the levels FSD proponents would have you believe. Humans are incredibly safe drivers when you account for the number of people driving, and the number of miles driven per year.

I’m not sure you can remove any driving deaths as alcohol or drug related. A hypothetical FSD system would be a solution to those deaths too, and as long as you are driving on the same road as everyone else, abstaining from them yourself still puts you at risk.

And if everyone drove modern high-end cars with the best assistive driving systems, especially as those improve to something approaching FSD on highways (or even achieve on certain highways in certain conditions) over the coming few years, it would be better still.

Yeah, I think people conflate seeing people drive erratically or doing dumb things in cars with an inability to operate a car in a way that ultimately prevents deaths. Humans are incredibly good at that in the bigger picture for reasons that are currently impossible to replicate with a computer.

Criticize human driving all you want, but even the worst of us can typically manage to get from point a to point b... Somehow.

One of the eye-opening things about the wikipedia statistics is that advent and universality of mobile phones at worst caused a very small uptick in fatalities. With all the bullshit you see with on the road, somehow people manage to mostly avoid death doing it. My guess is people can pull attention back to the road at moments it's really needed.

For sure. Well, humans don't really need to make great decisions individually to be protected, ultimately, by most people making an effort not to kill or be killed – to some degree at least. It's a very strange and fascinating thing about us.

I mean, that industrial agriculture eventually results in food on tables across the globe. So many systems at work. Run by the same people using their smart phones in their cars, haha. We manage to collectively make things happen.

Easy to criticize, but hard to argue against the results?

> A human being can drive a million miles without a serious accident

A million miles sounds like a lot, but it isn't. The average American drives 12,000 miles annually. Assuming a driving career of 52 years (from age 18 to 70), the average American will drive 624k miles. So the lifetime chances of getting into a serious accident are not exactly small.

(I don't know how to calculate the probability here, actually. Is 624k/1M the lifetime probability of getting into a serious accident?)

Just to comment on the math (not the actual numbers):

624k/1M is the expected number of serious accidents.

The probability of getting into at least one serious accident is 1 - (999999/1000000)^624000 = 46% (Every mile, you have a 99,9999% chance of no serious accident. You want that 624k times in a row).

Neat, thanks! So nearly a 1 in 2 chance of getting into a serious accident in a lifetime. Not great odds IMO.

I believe the mileage should be higher because most of us spend time as passengers in cars too.

> Computers probably can not be worse that the people already on the road.

And yet there is no evidence that computers are better drivers than people are.

I do have to ask, are we using the same computers? I've been using them for decades, and they're consistently buggy, error prone and straight-up factually wrong a lot of the time.

Self driving cars would have to kill over 1.35 million people each year to be as bad as human drivers[1].

But there is no reason we should be happy if a self driving technology "only" kills 1.2 million people in a year. That number is absurd and should not be considered acceptable. I think in the semi-distant future we are going to look at manual operation of a motor vehicle as a dangerous party trick, something only to be attempted by professionals or in limited circumstances like pulling into a field to park or some other low speed maneuvering.

[1] https://www.asirt.org/safe-travel/road-safety-facts/

We should be delighted if we have a self driving technology that only kills 1.2 million people a year.

1. That's an incredible amount of saved time, people now get that time back that they use to have to spend driving. We would have eliminated the largest suck of human time on the planet (truck driving). Etc. The main benefit of self driving is not safety.

2. We have a working baseline that we can improve upon to drive that number down, and since computer programs don't have the "new drivers need to learn from scratch" problem, those improvements will stick around approximately forever.

> semi-distant future we are going to look at manual operation of a motor vehicle as a dangerous party trick

totally agree but I'm skeptical of current tech to get there in the near time frame being talked about. It's all in how you define semi-distant

Hell, I would happy if self driving cars only kills 1.5 (slightly more) million people each car.

> Computers probably can not be worse that the people already on the road.

Computers are better than some drivers, but they're worse than others. If the goal is safety, computerized assistance is almost certainly better than self driving. Keeping the human driver engaged and doing their best, but having the computer supervising works a lot better than having the human supervising; it's hard for a human to remain alert enough to intervene quickly enough when needed. I don't know if there are good studies yet, but I expect automatic emergency braking to reduce severity of a lot of collissions. Cross traffic warnings probably eliminates a lot of minor (and some major) collisions. There are systems now to detect wakefulness; if those are combined with something to safely pull over and stop, that could reduce a lot of tragic collisions as well.

> If the goal is safety

If the goal was safety, a computer assisting a competent driver would be the best. If the goal is private profits for a few individuals, then a computer that doesn't take a paycheck would be best.

Unclear. There's some level of computer driving competence where humans will simply zone out.

that's not quite right.

Computers currently are far worse than the good drivers out there. It is not clear if they ever could be better than a trained and cautious driver.

For example you can easily avoid the situation above by not talking on the phone while driving across rails.

if the AI abruptly crosses the median how are you going to avoid that?

Now look at AstraZeneca craze. Vaccines save lives, thrombosis cases are extremely rare. Yet the guy on street is talking to someone on the phone: “Do not take AZ! I repeat: do not take AZ!!!”. Technology has to make just one mistake to lose layman’s confidence.

>I suppose they could program the train schedule in, but what if the train system changes the schedule?

I laughed at this, thinking about how the German trains are usually ±30 mins and often off by hours.

The train schedule would probably not fare better than a coin flip, except in Switzerland or Japan.

> I suppose they could program the train schedule in, but what if the train system changes the schedule?

Just check the train locations instead :)

https://rata.digitraffic.fi/api/v1/train-locations/latest/

Theoretically it should be able to use side-facing lidar and cameras to sense that a big truck-like thing is crossing.

Why not radar?

Because radars lacks resolution. Why my Tesla phantom brakes due to low hanging signs or slow cars in the other lane. They tune it down to ignore these anomalies and it has issues running into disabled cars and emergency vehicles.

You don't need resolution to detect train about to ram into you.

Similarly, lidar resolution is a great overkill, and overcomplication over a moderatly sophisticated mm wave radar, which is on top of that will be much more durable, and reliable.

In addition to obeying all railroad crossing signals, I'm certain the Waymo Driver has the concept of trains built in. The long-range sensors on the vehicle would project the probable path of the train and determine if it was safe to proceed along the car's planned route.

I'm not certain the Waymo Driver has the concept of trains built in, but what if they did? How good is it? I applied at Waymo and found out they don't do well in the rain. What if they train is splashing up water in front of it? I'd like to know how it deals with falling rocks, which is a problem in California. Would they tell me or lie about self-driving, like some well known spoke persons do? What are the hackable weaknesses of Waymo? Have they tried to have visual and computer hackers fool the car and send it the wrong way? It just seems to me it's going to take awhile, maybe decades in my amatuer opinion.

Unless the train was obscured from view by foliage at the height and angle that the lidar rests at, or if a truck pulled up next to the car, or... Well yeah, the only reason people can handle that situation reasonably well is that they have a large range of sensory input, most of that input has symbolic representation in the mind, and there's a healthy fear of dying that causes us to notice when the ground is shaking and reassess whether the train signal is working correctly.

To be honest, this seems like one of the easier edge cases to handle. Before crossing tracks, check if there's anything coming down the tracks.

Much easier than all of the weird edge cases related to the behavior of other drivers or pedestrians.

In England at least, if you're crossing HSR by foot, you pick up a handset and call train operations to cross. Probably overly conservative but obviously a simple thing for a car to implement along with the usual safety measures (though those would presumably involve gates or not at grade anyway).

This seems like the edgiest of all possible edge case arguments against self-driving cars, considering that a meat-driven car crashed through a working barrier and into a perfectly obvious train in the Bay Area just this morning.

Obviously meat-based drivers have flaws. The question is if the FSD robots at least retain all the existing capabilities of meat-based drivers. Trading one set of deficiencies for another raises the question of which set is preferable.

The reality is that for a long time we will combine both sets of capabilities and use "self driving" tech to enhance human driver capabilities.

In that case self-driving first needs to be able to avoid the relatively simple case of not of smashing through a barrier and the human driver can use their wetware to figure out how to handle railroad crossings which are diverse and complicated.

I agree that it’s an edge case but there are always completely bonkers people, out of the world on drugs or just suicidal.

I don’t want to risk getting figuratively in a car with one if I turn on FSD.

This is where you're wrong because I'm obviously a far better driver than that person who drove in front of a train this morning and it would never happen to me. I trust myself more than I trust other drivers. I trust myself more than I trust Waymo. I'd be happy to have Waymo prove me wrong.

The edge case that I always go to on why truly autonomous vehicles are decades+ away is winter driving. Lane markers and road signs disappear. Slush and salt spray constantly obscure sensors.

Winter driving as a human driver requires an entirely different approach and Waymo hangs out in the Arizona desert where there is basically never any inclement weather.

That, and construction, detours and city driving. In the latter, driving contexts can change without a second's notice, without any signs, markings or signals denoting the change. You learn what those contexts are from experience and understanding of what are ultimately complex social situations, and a shared understanding between you and the people around you.

Simultaneous location and mapping (SLAM) was solved a while back. Autonomous vehicles can locate themselves using any detectable landmark, not just lane markings and road signs.

don't even need to go to bad environmental setting. Let's talk about other countries. I haven't seen a lot of self-driving cars in Rome or Mumbai yet, there's a lot of places where traffic rules are treated more like suggestions than actual rules. American style neat wide highways and grid-like streets are not how the places look where most cars are being driven right now already.

Mumbai literally has bumper to bumper traffic, I think any self driving car's sensors would just go crazy and give up.

You can simulate poor weather with sensors by either 1. simulating it or 2. driving at night.

Digital data instead of signs seems like the obvious solution.

IMHO FSD cars (along with drones, secure programming languages/OSes, and distributed manufacturing) are a post-war technology. The economics don't make sense while we're at peace and locked into the Pax Americana economy. However, once folks start trying to kill each other, these technologies will give a very large survival advantage to those factions that adopt them. Imagine how useful not having to risk human lives on supply lines would be. And risk expectations get reset when things go from "nobody ever dies unnaturally" to "people are actively trying to kill you", which eliminates the biggest barrier to FSD adoption.

I think we'll get such a war within the next decade, so we may see FSD vehicles sooner than expected, just not in the way we want.

In a hot industrialized war (one that isn't wildly asymmetric), a supply line has so many soldiers' lives depending on it that having the transports manned or not is insignificant. But I think I know the general concept you are getting at, I'd call it the wartime risk economy: when a large percentage of your combat aircraft sorties don't return, it becomes reasonable to operate the engines much closer to the limits of reliable operation than in peace time. The extra performance can save more pilots than the occasional engine failures caused by the emergency operation mode cost.

Just like reliability SLAs, I suspect every extra nine requires an exponential increase in investment. My armchair-know-nothing assessment is we're maybe at 90% coverage* right now, with maybe 10x all investment up until this point needed to reach 99%; then 10x that to reach 99.9%. Even at 99.9%, though, a failure scenario still threatens injuring or killing 0.1% of drivers (350,000 motorists for US alone).

* Coverage as a percentage of scenario's occurrence over the total duration of driving. For example, over 90% of a long-distance trip will be spent on a highway following traffic patterns within a lane with the occasional lane change.

We regularize traffic more than your estimate though. Like 99.99% of the 300 mile drive I took yesterday was sitting in a lane.

The hardest part was that Google hadn't mapped a service drive, so it thought the adjacent service drive was the best route (which would be addressed pretty fast if you were trying to deploy self driving service in that area).

I don't think we get to level 5 very soon, but level 4 cars will have the ability to go lots of places pretty soon. If I overestimate driveway distances from yesterday, it's like 99.966% lane miles. There was some construction, but it was already well marked for human drivers.

Would not they detect the gates where down ? or are there still that many crossings without gates in the USA

Where I live in the Southwest US there are many crossings without gates. I'd say very few of the ones outside of well populated areas have them.

Gates are only a suggestion - signals/gates malfunction so often you can find compilation videos of them on youtube.

The place I talked about had no gates, after the accident they added gates.

That shouldn't be the standard. It probably will be, but it shouldn't be.

Humans are terrible drivers. If a self-driving car got into half as many accidents as the average human, it would save millions of lives. And kill people, to be sure, but fewer on net.

I also think you could make a reasonable argument that all cars should be banned, right now, based on how many people they kill, but since I don't think that's gonna happen...

You are correct on the aggregate, but that doesn't necessarily work when applied to each individual driver.

I'm a pretty attentive and cautious driver. In the 20 years I have been driving I've been in one accident and that was because another driver was attempting to make an illegal left turn, came across two lanes of traffic, and t-boned me. So if self driving vehicles are only doing better than the worst human drivers, I'm going to be pretty hesitant to turn over control. I'd be in favor of that other driver that hit me using an autonomous vehicle though.

Everyone thinks they're an above-average driver. Half of them must be wrong, so the reliability of any individual's self-assessment is about equal to chance. And even if you're one of the genuinely good drivers—your record certainly sounds excellent—our laws should account for the fact that most drivers can't judge their own skill.

...and they might be right.

It doesn't seem impossible for there to be a long tail of driving (in)ability: most people drive pretty well, with a small fraction that are distractible or reckless enough to account for most accidents.

Just wait until self-driving "personality" becomes part of a car company's brand identity. I suspect that we will see no lack cars that are designed to be terrible drivers. Of course they will call it "confident", "assertive", perhaps for some brands even "masculine". Cars could become even more civilian tanks thank that already are. I desperately hope that I'm wrong.

To be fair, humans are literally the best species on earth at driving cars

I think you’re right. For the AI we have today, it’s the equivalent of 1990. My guess is that as sensor arrays improve, wireless networks improve and quantum laptops are available in 2040 or so.

Think about it — most software is deployed cloud first these days, but one of the most complex computing tasks we have is relying on some black box computer.

> It's those edge cases that make me think real FSD (vehicles with no steering wheels) is a decade-timescale problem.

I'd be willing to bet it's a quarter century or longer problem. (Longbets, anyone?)

FSD is absolutely achievable, but the task is much bigger than some proponents give it credit.

Or you need a high-bandwidth coverage-everywhere comms system and a robust infrastructure for remote teleop, like what Starsky was building:

https://medium.com/starsky-robotics-blog/the-end-of-starsky-...

Then the steering-wheel-less car at least has the ability to call for help when it's only 98% sure of what to do instead of 99.99% sure. But obviously this kind of model only makes sense in a fleet context, not as an extension to something you own, so it requires greater shifts, at least for personal automobiles.

Most remote assistance is not live driving. A remote human adds labels and maybe even marks out a safe path for the vehicle so that vehicle has enough information to proceed.

Starsky's was. Did you see the picture of their console?

Driving in all conditions requires the computing power of a human brain. When we have computers that powerful, we can talk about L5.

If that logic were valid, then Garry Kasparov would have beaten Deep Blue in 1997.

chess doesn't require the processing of rapidly changing visual, inertial, tactile and auditory information while at the same time performing fine motor-coordination.

> but otherwise take a family off a bridge when the sun is blinding the camera?

I love these wildly over-the-top exaggerations.

When was the last bridge you saw without a barrier to prevent going off the edge?

What makes you think a vehicle vision system will handle "blinded by sun" any worse than humans already do?

Remember it's projecting and predicting the road ahead, even around corners and in the dark - so being blinded by the sun isn't going to cause it to swerve wildly off course and off the bridge - it can continue to use the data it had before being blinded (just as you do).

Also remember it has eight cameras it uses for this. The 16 year old new driver texting and talking to friends coming towards you at 60mph has two.

> When was the last bridge you saw without a barrier to prevent going off the edge?

That's a disingenuous question. Many bridges have wooden guards that will prevent you from going off if you make a small mistake, but not a large one.

> What makes you think a vehicle vision system will handle "blinded by sun" any worse than humans already do?

Humans can move their head, and block the sun with a hand, hat, or sunshade in the car. Humans have two eyes so if one is obstructed, the other may still get good vision.

> so being blinded by the sun isn't going to cause it to swerve wildly off course and off the bridge - it can continue to use the data it had before being blinded (just as you do).

Unless it's an incredibly well-trained AI, it may mistake lens flare for oncoming traffic, or a pedestrian. Car AI is not at the point where it has common sense to assume that lens flare is incorrect information.

"Blinded by the sun" is an understandable failure-state when you design a driving system that's reliant on cameras. From https://arstechnica.com/cars/2019/05/feds-autopilot-was-acti...: "Theoretically, it should be possible to detect the side of a truck using cameras. But it's not always easy. In some lighting conditions, for example, the side of a trailer might be hard to distinguish from the sky behind it."

How many of the 100 daily fatalities on the roads in the US do you think are caused by drivers not "seeing clearly"?

The goal of an AI driver isn't no crashes. The goal is less crashes than human drivers.

Idk if it's that simple, what if you reduce highway deaths by 100 but rural roads go up by 10. If I'm primarily a rural driver that's not enticing.

OK then, let's not put it into full scale production until all types of driving are an order of magnitude (10x) less.

>The 16 year old new driver texting and talking to friends coming towards you at 60mph has two.

My two analog eyes see this perfectly well, why didn't the AI?

https://www.thedrive.com/news/33789/autopilot-blamed-for-tes...

I think there's a lot unfair criticism of human drivers in this thread. I don't think we're at the point where we can call machines better than humans when it comes to these tasks.

The human driver's two eyes still have a wider dynamic range than any current affordable video camera. This makes a huge difference in difficult lighting situations like looking into the sun.

>What makes you think a vehicle vision system will handle "blinded by sun" any worse than humans already do?

Self driving software tends to have very poor object permanence.

It is mostly a hardware problem. Tesla's can't do FSD with current hardware because there aren't redundant cameras in each direction. One camera gets mud on it and your car runs itself into oncoming traffic. (It does that anyway right now, but even if software bugs are fixed this is still intractable).

A real solution has to be capable of cleaning all cameras (probably faster than a windshield wiper would), because the distortions caused by even normal rain are hellish to train an AI to handle.

Tesla would be lucky if the extent of its challenges were having enough cameras pointing in the right direction. Tesla handicapped itself by only developing its driving systems with cameras and not lidar.

I don't disagree. But since adding more cameras to a car that already has cameras is a much more intuitive extension for Tesla to select, I used that as an example.

Right, because people have LIDAR installed already.

100% agree but i wonder about the standards applied.

Where I live the streets are tight and most drivers mediocre at best, unaware that cyclists might have right of way and about 1/20 doesn't seem to know the difference between different kind of light settings in the car. At least half the cars on my street have visible scratches/dents.

For me that is the standard to be beaten, not perfection. And the car could still give signal and ask for a human to take over in some cases. Self-drivinf cars for me could also be much slower, no need to speed when you can read a book or play games

So, essentially, you live in Italy where traffic signs and lights are nothing more than street art and some extra frills on the side of the road.

Silly generalization, please avoid.

It's a generalisation but not silly.

It’s not a generalization.

I loathe to think about more complicated regulation, but I think Elon is pushing the boundaries of disclosure. That, and that we have a lot of new populism in stocks, makes me think the SEC maybe needs an update.

I kind of like Elon being able to say 'whatever', but on the other hand, it's not his money now, it's crossed the threshold into public financing, so statements like L2 v. L5 are 'material' and saying the wrong thing is a 'lie' and 'bad'.

Again with the paradox is that he's going to be hosting SNL which is kind of fun to see, on the other hand, it's going to be another occasion to hustle a stock or some kind of crypto which is distasteful.

Waymo has given up on ever getting L5 capabilities. They have so much invested they can't just shutter it, and keep it going in hopes it attracts an interesting partnership or purchaser.

They did the same thing with Skybox imaging/Terra Bella.

> Waymo has given up on ever getting L5 capabilities.

Except Waymo has never tried to claim L5 capabilities. They are and have always been a strictly L4 technology and they have said multiple times that L5 is not possible to achieve.

I dunno, I've seen the Domino's [Nuro](https://www.nuro.ai/) car roaming Houston and it seems to be doing okay and I haven't heard of any issues... If it's good enough for my pizza, it's good enough for me -I'm mostly pizza anyway! :P

I think it's remote-controlled though?

No, fully autonomous! https://www.nuro.ai/technology

You are probably right about 2021, but I think it is very likely that when L5 does come, it will arrive suddenly and unexpectedly. The same as AlphaGo did. Yes, I am aware that AlphaGo was the result of years of effort. Obviously there is an element of “how do you define ‘arrive’“ here.

I don't think Tesla will have L5 in 2021, but the FSD videos on youtube are pretty impressive no ?

They're impressive when they're on a straight road. But I think it's very concerning just how noisy and low resolution the data the cameras are working with is. You can watch some of the videos and the car won't recognise other traffic until it's literally within 20 meters of the car. The 3D positions of the cars are jumping around constantly. It will get the speed of traffic wrong and try to turn straight in front of someone that would cause an accident. Yes it's Beta but it's still a huge huge huge way behind what Elon claimed it would be by now. And he's doubled down on the current hardware being all that's required.

I am 100% unconvinced that Tesla can get anywhere with their current system. I don't see how their low resolution cameras can get the necessary information for Level 5 autonomy. It almost feels like a reckless brute force approach to the problem. "Just let AI figure it out". Every autonomous vehicle company is going to be using some sort of machine learning but they're going to be feeding in huge amounts of data. Waymo for example is using multiple LIDAR scanners to build up an accurate 3D model of the world surrounding the car. That's what you need. Not what is effectively guesswork by an AI.

We still don't have truly reliable face detection even after decades of research and we're supposed to believe that a car can reliably drive itself on shitty low resolution cameras alone.

I agree with most of this. I would be happy of Tesla (or anyone else) could just achieve L5 on highways so I could read a book on long drives on say the I90 in NY or I95 going to Florida.

That seems like a problem that could be solved with their current sensors.

Yes, they are but also a lot are sped up. The videos that are raw show the car still struggling to see around corners since the cameras are just not in a good enough spot. They really should have put cameras on the very front corners.

But TBH I'm not really THAT impressed it can take corners and follow more lines. It still doesn't handle very important edge cases, and the people testing and uploading videos are naturally biased to show how good it is to push their referral codes.

I'm not anti-Tesla, I love my M3, but we need to be realistic about the future of what these particular cars can do. They're never going to be L5 with the current sensor suite - and they're certainly never going to be robo taxies. Who really wants that anyways? Last thing I want is some drunk bros to destroy my car and have to deal with Tesla support.

I agree, I would not loan my car out to be a robo taxi. As it currently stands, the people that can afford Teslas probably would not take the risk of damage to their cars for the extra income.

They are impressive for what they are, yes, because it's a hard problem.

But it's non stop interventions...

Do you have a link to Elon's FSD claim?

He definitely uses ambiguity to his benefit (eg "soon", "by fall/winter/spring/summer" (in which year?)), but I haven't heard anything about Tesla being L5 by the end of 2021.

This feels like bad faith. Musk has been making these claims every year since 2016. His coverage of these claims on HN is regular and thorough.

I went ahead and Googled one for you from 2020:

Tesla will be able to make its vehicles completely autonomous by the end of this year, founder Elon Musk has said.

It was already "very close" to achieving the basic requirements of this "level-five" autonomy, which requires no driver input, he said.

- July 2020

https://www.bbc.com/news/technology-53349313

Do you mean this comment?

I remain confident that we will have the basic functionality for level five autonomy complete this year.

Basic functionality for level five isn't level five.

I mean, off the top of my head he claimed 1 million robotaxis on the road in 2020, and that the car would be able to drive LA to New York by itself, charging and all, in 2018.

Who are you trying to fool?

That's not what I asked about in my initial post.

Who are you trying to fool?

"Basic functionality for level five isn't level five."

I'm surprised you say this totally seriously. For me this means that level 5 will be this year, and then there might be extra new cool features, which are not necessary.

Basic functionality for level 5 sounds like both necessary and sufficient to me

I view part of basic functionality of level five as "Can the car drive on all roads in at least some circumstances?"

That's still not sufficient, because it needs to do that in all circumstances (vehicles, people, animals, stuff, weather, etc), but even if it can handle all the circumstances (also basic functionality), if it can't at least all roads, then it's not level five.

Right now, it's level two, and it's IMO going to be level two for a while. People can ask about level five, but I take any statements about Tesla's current autopilot and/or beta as statements about the level two system.

People can assume that those statements are about some future level five system, but I don't think that's an accurate assumption unless someone from Tesla says autopilot and/or the beta is now level three, four, etc...

If that is your understanding of the phrase "basic functionality for level 5", can GM announce that they are installing two cameras in each car and now they have basic functionality for level 5?

If basic functionality does not need to be sufficient, then even the tiniest necessary thing is basic functionality? Like installing a camera on each new car? In that case, Tesla already has basic functionality for level 5 in terms of both hardware and software.

Having hardware present could be the basic equipment required for level whatever functionality. Assuming the hardware had sufficient coverage/functionality.

Tesla does have basic hardware needed for level two functionality, maybe more. They could have basic software functionality needed for level two or more with the beta, but someone would need some kind of audit to confirm that.

> Do you have a link to Elon's FSD claim?

Google is universally accessible to everyone. Please don't be that guy who corners himself into a blind spot.

Musk claimed coast to coast self-driving trip by end of 2017.

https://www.theverge.com/2016/10/19/13341100/tesla-self-driv...

I've searched, but I haven't found any comment from Musk about level five being ready by 2021.

edit - The closest is about the basic functionality for it being ready by the end of the year (see above).

end of last year.

2020 Q4 Earnings call, 27th Jan 2021 (according to https://www.fool.com/earnings/call-transcripts/2021/01/27/te... )

Musk: So -- and this is -- basically I'm highly confident the car will drive itself for the reliability in excess of a human this year. [...]

<later>

Director of Investor Relations: [...] The next question is, why are you confident Tesla will achieve Level 5 autonomy in 2021? [...]

Musk: I guess, I'm confident based on my understanding of the technical roadmap and the progress that we're making between each beta iteration.

Saying he's confident about each beta iteration, and the FSD beta is currently level two, is not the same as saying Tesla will hit level five by the end of the year.

He is explicitly asked by the person feeding him question why he is confident that they will reach level five this year. He is confirming that he is confident they will reach level five this year. Not confident that they will improve, confident they will reach level five.

He was asked that question, but he didn't say he was confident they would reach level five this year.

He did refer to the current FSD beta, which is a level two system, and expressed confidence in the technical roadmap and progress they are making between each beta iteration. Then he talked about the level of reliability they would need, capabilities of the system, and challenges they have.

I guess, I'm confident based on my understanding of the technical roadmap and the progress that we're making between each beta iteration. Yes. As I'm saying, it's not remarkable at all for the car to completely drive you from one location to another through a series of complex intersections. It's now about just improving the corner case reliability and getting it to 99.9999% reliable with respect to an accident.

Basically, we need to get it to better than human bio factor at least 100% or 200%. And the business happening rapidly because we've got so much training data with all the cars in the field. And the software is improving dramatically. The -- we also write the software for labeling.

And I'll say it's quite challenging. We're moving everything toward video labeling. So all video labeling for video inference and so there are still a few of new met that need to be upgraded to video training inference. And really, as we transition to each net to video, the performances become exceptional.

So this is like a hot thing. The video -- the labeling software that we work for, video labeling, making that better has a huge effect on the efficiency of labeling. And then, of course, the Holy Grail is auto labeling. So we're -- we put a lot of work into having the labeling tool to be more efficient when we used, as well as enabling auto labeling where we can.Dojo is a training supercomputer.

We believe it will be -- we think it may be the best neural net training computer in the world by possibly an order of magnitude. So it is a whole thing in and of itself. And this is offer potentially as a service. So some of the others need neural net training, we're not trying to keep it to ourselves.

So I think there could be a whole line of business in and of itself. And then, of course, for training vast amounts video data and getting the reliability from 100% to 200% better than average human to 2,000% better than average human. So that will be very helpful in that regard.

https://www.fool.com/earnings/call-transcripts/2021/01/27/te...

If your response to "why are you confident about X" is "I am confident because...", you can't claim you didn't say you were confident about it.

That's not the situation though.

The question is "why are you confident about X", and the answer is "I am confident because we're making progress in these ways with Y".

If someone says by "fall/winter/spring/summer" and they don't mean the immediate upcoming instance of that season, they are lying. That is not being cleverly ambiguous, it is pure dishonesty.

They could also be wrong.

I think the assumption that it's a specifically lie else really highlights where these ideas are coming from.

He made the claim during the Q4 2020 earnings call for Tesla in January 2021: https://www.fool.com/earnings/call-transcripts/2021/01/27/te...

An investor asks “Why are you confident in L5 self-driving before the end of 2021?” and Elon goes on to explain why he is confident that Tesla will achieve L5 self-driving by the end of 2021.

> Do you have a link to Elon's FSD claim?

Musk said he's 'confident' reaching l5 autonomy by end of 2021 in the Q4 earnings call, which is on youtube.

Even if I don't want to, I can't help but keep up with his marketing-claims through the interests I have. My take, being aware of him but not focused, is that beta-FSD has already been released! ... I mean ... what? Maybe I've got it wrong.

It's been released to a small number of people, but... beta-FSD isn't level five.

The only specific statement I've read from him is that the basic functionality for level five will be available this year.

What's interesting to me is that people seem to attribute what news articles synthesize about his statements as statements made by him.

https://www.bbc.com/news/technology-53349313

>news articles synthesize about his statements as statements made by him

Musk promised the coast to coast drive in 2017:

https://www.theverge.com/2016/10/19/13341100/tesla-self-driv...

Then admitted he couldn't do it in 2017:

https://www.autonews.com/article/20180208/MOBILITY/180209770...

This isn't the media delivering a dishonest commentary. These are his words.

I'm referring to my initial post in this thread.

But yeah, we can have both.

Those were his words in that situation, and he was called out for it. In this situation, the media is delivering dishonest commentary, and trying to call him out for their commentary.

He made the claim in the Q4 2020 Tesla earnings call in Jan 2021, here is a transcript: https://www.fool.com/earnings/call-transcripts/2021/01/27/te...

But you see how he/they should be clearer about it?

To me, FSD is ... FSD ... what's FSD if it isn't level 5?

Sorry, to Elon fans (and I appreciate Tesla), but it's nonsense and people are fools to put up with it, let alone pay $10k or whatever it is for the "opportunity" to have it later.

They could, but that's the nature of sales, and language to some degree. I drive on a parkway and park on a driveway.

As for the specifics, it's what's described on the Tesla website.

Reply on news.ycombinator.com