Hacker News Reader

Comment by thepasswordis | Hacker News Reader

It seems like there is a huge disconnect between people who know about FSD from using it and people who know about FSD from things they read on the internet.

I have FSD, I use it every single day. I love it. If every car on the road had this the road would be a substantially safer place.

View on HN

FSD is terrible. I have it, it's dangerous. If I had to go 5 miles through city streets I'd sooner trust a human after 3 shots of tequila before I'd trust FSD every time.

Disagree. I've had the beta for a year and drive in NYC. It has been mostly flawless considering all scenarios it has to handle. It does make the occasional mistake but (1) it's a beta and (2) I'm responsible for monitoring and correcting it. Agree that those that expected more should be entitled to a refund if they so desire.

> It does make the occasional mistake

Yeah, that's what I mean by "dangerous".

Ok but don't you have to weigh that against the potential of avoiding accidents by reacting quicker then you as a human can?

I have defiantly seen cases where FSD Beta stopped and the driver didn't understand instantly understand why.

Or simpler cases where the car follows a lane while the driver wasn't paying attention (adjusting the radio or whatever). Those can easily lead to an accident but are unlikely to if you are in FSD.

How do you make that calculation?

The calculation is easy, FSD does the wrong thing so often that there's no question a human driver is safer.

The question is not if FSD makes mistakes, its if FSD+a human driver monitoring is worse then a human driver.

All the available scientific evidence suggests these systems diminish driver focus and attention which definitely makes them less safe since these systems can't be operated safely without an attentive driver.

I have seen some evidence bases on some limited lane keeping systems that suggest this. But not on systems that are as advanced as FSD Beta for inner-city with the same amount of forcing the drives attention systems.

Consider the possibility that you might both be right. Two people may have vastly different experiences with FSD, depending on the routes they take and other variables.

Also some people may have a very low tolerance for the car not driving like they would. It does make you uncomfortable .

Do you have the latest beta?

How did it improve in the last 12 months?

That other drivers are beta testers for software controlling a large mass of metal and combustibles hurtling down asphalt at _ludicrous_ speeds is something you'd write into a dystopia story.

The fact of the fiction is terrifying.

It’s a choice, you are in control all the time.

https://www.sae.org/publications/technical-papers/content/20...

"This study investigated the driving characteristics of drivers when the system changes from autonomous driving to manual driving in the case of low driver alertness.

...

In the results, a significant difference was observed in the reaction time and the brake pedal force for brake pedal operation in the low alertness state, compared to the normal alertness state of the driver, and there were instances when the driver was unable to perform an adequate avoidance operation."

The other drivers on the road, and pedestrians, aren't getting to make that choice.

You can re-take control, which may take seconds depending on the mental and physical state of the driver. A lot can happen in a second.

Not sure if it's the latest, and nowadays I rarely use it, and when I do it's just to show off to curious friends, but I live in an east coast metro and I've never been able to use FSD for more than 20 minutes without it making a dangerous mistake - 90% of the time it's been a mistake making a left turn, turning into the wrong lane or oncoming traffic after making the left, not correctly respecting the traffic lines or the left turn lights and other very scary situations.

So it’s good enough that you let it control and become less attentive?

I was thinking that you feel in what kind of conditions it behave correctly (highway), low traffic and then use it in progressively more difficult situations as it prove to you it can handle them consistently.

I know it’s not ready yet. I heard as saw so many positive reviews (but still with corner cases and strange behaviours)

I've not had any FSD issues on the highway besides phantom braking, it keeps lanes correctly, safely adjusts to traffic speeds and it merges into offramp lane without issue (though it still makes me nervous), once it gets into city streets though...

Musk has the latest internal release that will fix every problem at the end of the month, every month.

It seems there’s a huge disconnect between people who know how to drive and are confident in their skills, and those who are scared to drive and want to offload it to Musk.

I’m actually afraid to drive behind a Tesla and either keep extra distance or change lanes if possible. I still have more faith in humans to not randomly brake them in beta FSD.

It’s one thing to put your own life in the hands of this beta model, and it’s another to endanger the life and property of others.

Yeah it's not that I'm scared of my own driving, I mean that if other people had it it would be safer. Things which have happened in the last 3 days of moderate driving around a city:

1) A car gets impatient at some traffic turning right on a green light in a construction zone, and comes into my lane meaning into oncoming traffic. We have to swerve out of the way and slam on the brakes to avoid them.

2) A guy gets impatient behind me slowing down over a speedbump, tailgates me a few inches behind and then passes me by cutting into oncoming traffic in a no passing zone, then cuts me off, again with a few inches to spare, and speeds through the neighborhood where we eventually meet at the stoplight at the end of the road.

3) Every single day people run red lights. Almost every light where there are people turning left at a green there are 4-5 cars that go through the red light after it turns.

My safety score on my tesla is 99. I am an extremely safe driver. I wish FSD was more common because so many people are terrible drivers.

Your Tesla tells you you're a safe driver. Well, geez, I'm convinced that the $55,000 (maybe+?) you spent tells you things you like to hear. I mean, one of the criteria is "Late Night Driving". Why not driving between the hours of 4pm and 6pm? When there are a lot more cars on the road, which is just statistically less safe.

https://www.tesla.com/support/safety-score#version-1.2

And looking at the rest of the criteria, they're ok, but hardly comprehensive. This is like, the bare minimum. It doesn't measure what I would call "doing stupid shit". Like crossing several lanes of traffic to get in a turn lane. Forcibly merging where you shouldn't. Nearly stopping in the middle of traffic before merging into a turn lane at the last minute. Straddling lanes because you're not sure if you really want to change lanes or not. Making a left turn from a major artery onto a side street at a place that is not protected by a light. Coming to a near complete stop for every speed bump, then rushing ahead to the next.

And a host of other things that demonstrates the person does not consider other people on the road at all.

Here's an entire article about how to game the safety score:

https://cleantechnica.com/2021/10/14/three-quick-tips-for-a-...

One of the tips is to "drive around the neighborhood when traffic is light".

And the car doesn't ding autopilot for behaviors it would knock a human for. Because the assumption is that the car would know better I guess. But then why isn't the safety score simply a deviation from what autopilot would do in a situation? If autopilot would brake hard to avoid a collision, shouldn't you?

The Tesla doesn't just tell me I'm a safe driver, it also gives me a cheaper insurance rate due to the safe driving. The safety score is related to my monthly insurance premium.

So maybe it's to stroke my ego? But they're also putting their proverbial money where their mouth is.

The insurance you get from Tesla, which has a vested interest in its own Safety Score. It's not giving a cheaper insurance rate "due to the safe driving", it's giving a cheaper rate due to having a better score on the metrics it decided.

You see how that's circular, right. It does not mean you are a safe driver.

…what? No I am not following that. The safety score is a representation of the metrics used to determine the insurance rate. This is not in any way circular. The insurance rate and the safety score are representative of the same thing: safe driving.

https://www.tesla.com/support/safety-score#version-1.2

They're both from Tesla. There's no real proof that they're representative of safe driving. And you kind of want it to flow both ways: "the safety score represents good driving, so I get cheaper insurance" and "I get cheaper insurance so the safety score must represent good driving".

You're trying to use each of these things to validate the other so you can then claim it's something entirely else.

I think it's telling that autopilot is allowed to drive in a manner that would otherwise negatively impact your safety score. Either autopilot is unsafe or the safety score doesn't measure actual safe driving.

Man I think you're really confused about what the "safety score is". Here it is stated a simpler way:

Tesla tracks driving style and gives you an insurance rate based on those driving habits.

My insurance rate is about $90, which is the lowest the rate goes, because the things it tracks (following distance, for one) corresponds to a lower liability risk for the insurance underwriter, which in this case, is tesla.

The monthly price is directly correlated to driving habits. You seem to be getting confused about "safety score".

I'm genuinely confused as to what about this doesn't make sense to you.

Tesla is the underwriter. They give a monthly premium based on driving habits.

I'm not confused.

Tesla is the insurer. Tesla is also the manufacturer of the insured item.

You are using as proof of the claim, the fact that the group making the claim says they're right. This is big "We've investigated ourselves and found nothing wrong" energy.

You are talking to someone who is excited about… The Future (tm). And will defend their purchase and choices to the death. So I doubt there is any reason to continue. The fact they even gave us their Tesla given “safety rating” like it was something to give a shit about says a lot.

We are all glad your car is giving you head pats and game badges.

Also, the Tesla driving score reminds me of the Windows “Experience Index” score. Which was complete and utter toss. But made people feel good about their systems.

> I’m actually afraid to drive behind a Tesla and either keep extra distance or change lanes if possible.

Irrespective of the Teslas or FSD:

If you're afraid of an accident due to the vehicle in front of you braking regardless of the circumstances, then you're following too closely. It doesn't matter if the braking event is anticipated or unexpected. If you're not confident in your ability to avoid an accident if the vehicle in front of you slams on their brakes then you are following to closely.

> If you're afraid of an accident due to the vehicle in front of you braking regardless of the circumstances, then you're following too closely.

I know what you're saying, but that is not what I'd meant. Also, I don't follow too closely.

The difference here is that almost always a driver will brake depending on the happenings in front of them. So, if you pay attention to not only the car in front, but in front of them and in the neighboring lanes and so on, you can sense and also detect patterns in how a particular driver is driving. There are many skittish drivers who brake every 1 second and some don't, and so on. Basically based on your driving experience you can predict a little.

The problem here is that this stupid POC FSD will brake randomly or change lane randomly or whatever, so there is no way you can predict, and hence my concern and issue with it. I just prefer to change the lane, but that's not always an option.

> Basically based on your driving experience you can predict a little.

Yes, we all do that AND you're using that predictability to take liberties in safety such as following too closely. FSD's unpredictability exposes the vulnerability in your driving process and makes you feel uneasy.

You (and everyone else) can follow too closely AND FSD can be an unsafe steaming pile of crap. It's not an either or situation.

If you drive like everyone is trying to kill you, which they definitely are, then whether it's FSD or another idiot makes no difference to you.

This is actually how I tell my kids to drive lol.

“Drive like everyone else is an impatient brain dead dipshit intent on killing you”

“Also, check your mirrors constantly. “

I have it (subscription, just wanted to try it out). I won't use it again. It confidently attempted to drive me into a rather solid object with very little warning. If every other car on the road had this, I would sell all my cars and forbid family members from getting anywhere near the road.

It's nice on the freeway though.

To me it's weird you can voice an opinion like "I won't use it" and then say it's nice on the highway, as if you've given it enough of a trial there to endorse it when you think otherwise it's going to be killing people.

How is it weird that FSD might work well on extremely simple, boring stretches of road but fall apart as complexity increases?

Not that weird, it's just weird to recommend it after you've seen it almost kill you. I doubt he's put in the requisite time/miles/effort/observation/whatever to know that it does great on the highway but not elsewhere.

There's solid objects on highways, too.

Because the Tesla fans will cling to anything to further their arguments.

Pretty sure almost crashing into a solid object seems good enough for giving it a failing score.

I definitely agree, in which case I wouldn't find myself endorsing it for any use. Just because I drove X amount of miles without a problem on the highway doesn't mean it's any less likely to drive into a solid object, why would he recommend it?

I dunno man, according to another poster, I think FSD would probably give itself a safe driver rating of at least 85 for that.

"Works on my machine!"

In fact, there is a huge disconnect between people who design similar systems (Vision, safety, self driving, robotics, simulations) and those who use them. The criticism of Tesla is coming from experts, not jealous competitors.

The problem with systems where 99% is a failing grade, is that 99% of the time, they work great. The other 1% you die. No one is against FSD or safer tech. They are against Tesla's haphazard 2d vision-first FSD.

Wanna hear about this new fangled tech that avoids 100% of accidents, has fully-coordinated swarm robotics between all the cars, never swerves and can operate in all weather + lighting conditions ? It's called a street car.

The words 'road' and 'safety' should never be uttered in the the same sentence.

Yep, I used to be in this field back in 2018, and everyone was extremely dismissive of the rest of the industries skepticism

GREAT comment.

I think a large issue is people are really bad at risk assessment.

Also, why the hell did they try doing camera only? Fucking dumb as hell. I would have trusted something like FSD a lot more if they had also put in very advanced radar etc. trying on fucking /cameras/ only for newer models should be a criminal offense.

It's well known at this point.

When the model S was at its sales peak from 2013-2020, Lidar was still pretty expensive. So he put cameras in the original Teslas.

When Karpathy took over in 2018-ish, Tesla realized that all their advantages (dataset size, quality of researchers) was in the 2D realm. Waymo had been doing lidar for longer, seemed to have better 3d engineers & could keep upgrading lidar devices as they went.

Tesla's insistence on 2D cameras has to do with stubborn desperation more so than a willingness to play catch up. Elon needs 2D cameras to 'win' because even as early as 2019, he thought that it was too late to start playing catch up in 3D.

> The other 1% you die

Um... no one died. I understand your point is hyperbole, but you're deploying it in service to what you seem to claim is serious criticism coming from experts. Which experts are claiming FSD beta killed someone? They aren't. So... maybe the "experts" aren't saying what you think they're saying?

> Wanna hear about this new fangled tech that avoids 100% of accidents, has fully-coordinated swarm robotics between all the cars, never swerves and can operate in all weather + lighting conditions ? It's called a street car.

And this is just laughably wrong. Street cars (because they operate on streets) get into accidents every day, often with horrifying results (because they weigh 20+ tons). They are surely safer than routine passenger cars, but by the standards you set yourself they "aren't safe" and thus must be banned, right?

As advanced cruise control? It's great. As what it's marketed as? It's fraud.

I have the same experience. I think if there was more transparency in the data around FSD related accidents the conversation would be different. The last evidence we have is Tesla stating that there have been no fatalities in 60 million miles of driving under FSD. Pretty good so far.

There is no such evidence beyond unsupported proclamations by the vendor who, by the way, has compulsively misrepresented the capabilities of the product for years. The only evidence available that is not completely tainted by a complete conflict of interest is basically Youtube videos and self reported information [1] by investors and super fans that on average show critical driving errors every few minutes and even those are tainted by self interest to over-represent the product.

Tesla’s statements should only be believed if they stop deliberately hiding the data from the CA DMV by declaring that FSD does not technically count as a autonomous driving system and is thus not subject to the mandatory reporting requirements for autonomous systems in development. If they did that then there could actually be a sound independent third party statements about their systems. Until that time their claims should be as trusted as Ford’s were on the Pinto.

[1] https://www.teslafsdtracker.com/

Absolutely need more data, but if a public company makes untruthful claims about the past they are in big trouble. They are mostly free to make claims about the future without much consequence.

This is why startups are allowed to project massive sales next year but Elizabeth Holmes is headed to jail.

If it turns out that FSD has had a fatal crash and Tesla lied about it Musk is headed to jail too.

That’s cute that you think that.

We don’t live in a “just world”. There’s a reason a lower tier person like Holmes gets some prison and psychos like, say, Trump or Musk barely get a slap on the wrist.

Yes if we forget that the FSD may conveniently disconnect right before an accident

For Tesla's collection of data on accidents with Autopilot technologies, they considers Autopilot to have been enabled during the crash if it was on several seconds beforehand. These systems may hand control back to the driver in crash scenarios moments before impact, but that doesn't mean the data will report that the system was not enabled for the incident.

Yeah, let's now hear the stats on "fatal accidents within 30s of FSD being enabled".

You really just can't trust Tesla at all, about anything. There's no integrity there. They won't even be honest with the name of their headline feature!

Regardless of whether or not that's true, that's also not the impressive number you think it is. Human-driven cars have a fatality rate of about 1 per 100 million miles travelled, and that's including drunk drivers, inattentive drivers, and all environments, not just the ones people are comfortable turning on these features in.

Yes. The data collection continues. But this initial report indicates that FSD is probably not massively worse that a human.

This has been my experience as well - I use autopilot a lot and find it works great

Ditto. Most fun I've had in a vehicle in my whole life. A robot drives me around every day, and the worst misfeatures are that it's too timid at stop signs and left turns, makes poor lane choices sometimes (oops, can't turn, will go around), occasionally zipper merges like a jerk (yes, technically you can use that lane, but if you try you'll get honked at -- car doesn't care).

But the company is run by an asshole who people love to hate, so... everything it does is Maximally Wrong. There's no space for reasoned discourse like "FSD isn't finished but it's very safe as deployed and amazingly fun". It's all about murder and death and moral absolutism. The internet is a terrible place, but I guess we knew that already. At least in the meantime we have a fun car to play with.

> FSD isn't finished

"Full self-driving" has not been delivered after years worth of lies. It isn't about "murder and death and moral absolutism". It's about fraud.

Your vehicle is a dead end. It will not be upgraded to the latest hardware. It will not attain "full self-driving".

This is a "me or your lying eyes" argument. It drives my kids to school. Why is it so important to you that I not enjoy the product I purchased and love? You realize you aren't going to win an argument like that, right? You think I'm going to just wake up and realize that it can't do what I see it do?

It doesn't matter how you rationalise it to yourself or how irresponsible you are in using it. Your car does not do full self-driving. It is stranded on Hardware 3. It will not get Hardware 4 unless a lawsuit enforces it: https://electrek.co/2023/02/15/tesla-self-driving-hw4-comput...

Tesla has lied about full self-driving for years: https://www.tesla.com/blog/all-tesla-cars-being-produced-now...

Tesla has admitted to false advertising around full self-driving: https://electrek.co/2022/12/12/tesla-ordered-upgrade-self-dr...

Tesla has admitted that they have failed to deliver full self-driving. Tesla is now scrambling around trying to avoid being done for fraud: https://electrek.co/2022/12/07/tesla-self-driving-claims-fai...

If even Tesla can admit it, find a way to admit it to yourself.

> Your car does not do full self-driving.

It literally drives itself. You're hiding behind a semantic argument about what "full" means, and how autonomy is deployed. It's the "Money Printer Go Brrr...." meme.

You don't seem to be able to understand that the product as deployed has value to its purchasers, and that's leading you to some really emotionally dark places. I mean, it shouldn't be this upsetting to you that I like my car, and yet it clearly is. Just... maybe just don't? Let it go. You know these cars are safe as deployed. Let the Tesla nuts have their fun.

> You're hiding behind a semantic argument about what "full" means

No, I'm "hiding" behind the truth. I'm "hiding" behind the benchmark Tesla very publically sets for itself. Every year. Year after year:

https://jalopnik.com/elon-musk-promises-full-self-driving-ne...

> that's leading you to some really emotionally dark places

Okay, this is just sad now. It's like a kind of Stockholm syndrome. Cult members go one of two ways when confronted with realities which contradict their beliefs: they either finally recognise the silliness of their beliefs or they become even more ardent believers.

It's better to try to find a way to be objective. You shouldn't let people con you and you certainly shouldn't propagate the lie.

I have a feeling the differences between experiences is heavily influenced by where you are located. There is no way they are adequately able to train across all of the locations FSD is expected to be able to work.

Australian here. We drive on the "wrong" side of the road. I have no idea if FSD even accommodates this different scenario. And if it did, and being from Melbourne, having diabolical yet somewhat logical hook turns (turning from the furthest left lane to turn right) I will bet would not even fall in FSD scope.

https://en.wikipedia.org/wiki/Hook_turn?wprov=sfla1

Because of this, if Tesla solves it they are going to have an almost insurmountable moat.

It really depends where you drive, it behaves super well in some conditions, and awfully in others. Hence the wide disparity in experience.

Exactly.

I think a lot of "people" online are paid shills.

HN should severely downrank any "opinion" posts from anon accounts.

I have a Tesla with FSD and it's incredible. Though it drives way worse than me, it drives pretty darn good, and will save a ton of lives and change the world by freeing up a lot of time for people to do more important things.

Is your post supposed to be read as genuine or sarcastic? It looks to me like a strawman of a Tesla owner stereotype, and I don’t want to accidentally eat the onion here by taking it at face value.

Your post appears to be saying that Tesla’s FSD is way worse than human drivers, not safer than them, but we should still welcome it for the convenience. Again, this reads like a strawman. I would like to think the average Tesla owner would not endorse that statement.

The goal isn’t to save time at the cost of lives, it’s to save both, non?

> is way worse than human drivers

No, I said it's way worse than me. Probably around performance of the average human driver.

I am training to be an astronaut, so on the advanced side.

A lot of people say they’re significantly better than average driver. It’s such a stereotypical thing to say, that it probably has negative truth value on average, since it reflects broad personal confidence more than actual driving skill.

You still sound like you’re trying to present the weakest possible straw man for Tesla critics to attack. Stop this bad faith attempt to make Tesla look bad; you’re dragging down the discourse. Nobody is going to take your bait and feed you the expected counter arguments to the nonsense you expressed up-thread.

I suspect that the NHSTA know about FSD from all of the above and more.

Even if Tesla’s current implementation were objectively safer than the average human driver, it would still represent a net negative on safety because of the negative impact the glaring flaws will have on peoples’ confidence in self driving tech.

Like Wozniak? https://www.techspot.com/news/97563-steve-wozniak-slams-dish...

I've never used it but in my book "Full Self-Driving Beta" is an oxymoron

Until your car decides to randomly slam on its brakes and cause a pile up.

I'm reminded of a thread on HN where someone did a cross country trip and their car only phantom-braked a couple of times, and called that a huge success.

Ha, my new Model 3 phantom braked the first time on the trip home from the service center, which was only 10 miles. I doubt it had even gone over 10 miles on the odometer at that moment.

How were you able to put it into autopilot? Typically, the autopilot is calibrating itself the first 60 or so miles of driving, and it does not let you engage it.

Good question! It didn't argue one bit, literally drove it off the lot at the service center and hit the freeway and turned AP on (it's my second Model 3, so I already have such habits).

It's entirely possible it had more miles than 5 on the odometer when I picked it up. Officially the paperwork says 15, the associate who gave me the car said it was actually less, and the odometer isn't very prominent so I never even looked. Maybe it had 50 miles and was a reject ;). I should go check TeslaFi, since I reenabled that the day I bought the car. I confess that the only way I ever know how many miles my car has is when I see it on TeslaFi.

My 2022 Model Y didn't require any calibration to use autopilot after initially picking it up. IIRC there were 7 miles on the odometer at the time.

If my car slams on its brakes, and this causes you to hit me, then you are exactly the type of person I wish was using FSD.

You're following too close to me for the conditions, you're distracted, or some combination of both.

My car might have to emergency brake for any number of reasons. This should never cause a pileup, and if it does it's your fault, not mine.

That’s.. dumb.

I’m the type of person who keeps a lot of car lengths between me and the car ahead like you are /supposed to/ but I’ve also had someone panic stop in from of me and, due to road conditions, I cannot stop in time and have hit them. Not hard. But did all the same.

FSD and similar aren’t magic bullets that would prevent that. People like you who seem to think they are are also part of the problem.

Would a completely automated roadway be fucking awesome for safety and productivity and relaxation? Yes. Is it going to happen anytime soon? Doubtful.

Blows my mind how many people are totally willing to risk their own safety and everyone else’s using known buggy as fuck software. On the road. At 70mph. In a 2 ton vehicle.

… but it hasn’t been an issue for you! You say. Ok.

But I also watch dumb asses lane hop with barely enough room to pass likes it’s a video game and then act surprised when they wreck. So I guess I’m giving people too much credit.

If someone brake checks and you hit them, you're just as much of a problem. Learn defensive driving and stop driving so close to another driver. Keep a safe distance.

Tailgating pisses me off. I don’t do it.

But this backlash to my comment about hard braking and wrecks is kind of hilarious.

Basically, my point is that Tesla has made egregiously bullshit claims about FSD and autopilot. They have given them misleading names. And people who want to “live in the fuuuuture” have eaten it up and are using the Teslas like they are actually capable of driving themselves 100% of the time. In all conditions.

If anything, it has made people who rely on it seemingly worse, inattentive drivers.

So while I’m glad y’all enjoy your purchased vehicle, don’t pretend all gripes and complaints lobbed at them are invalid or someone else’s fault. That’s asinine and childish.

Because only a FSD car causes piles up... your right, I NEVER saw 90 car pileups on the news before FSD. /s

I'm pointing this out because he said (in italics) "substantially".

It doesn't really matter, we have multiple incidents of FSD causing accidents due to outright mistakes, not even "a human would have messed up here too" situations.

Do we? Becuase most of the news reports about claims blamed on Autopilot or FSD turn out after investigation not to be cases.

In one famous case the policy claimed that the driver was '100% not in the driver seat'. This caused a huge media storm and anti-Tesla wave.

Just recently the full report came out and stated that Tesla AP was not used at all and the drive was driving normally.

https://electrek.co/2023/02/09/tesla-cleared-highly-publiciz...

There are quite few such cases, this one maybe being the one that caused the most media attention.

So I tend to discredit all such reports unless its several months after and a full investigation report has been done.

Do you have links to verified reports of FSD causing such crashes.

It seems to me that phantom breaking could lead to such issues, but I have not yet seen a real report that claims this happened.

Humans also mess up in situations that aren't "a human would have messed up here too".

Sure, but say I consider myself a diligent driver and I've never caused an accident... what does FSD have to offer me? Yet another random opportunity for a car to fail me. Why would I surrender control for that?

Reply on news.ycombinator.com