I Tried Tesla FSD
werd.io
werd.io
I like the way the author writes this. Interesting to hear unfiltered thoughts from a first-time user of the software.
Meh. Believe what you want; read what you want; fear what you want. Whether it "works" or not is a question susceptible to factual analysis, not vocabulary.
HN talks a lot about “we need FDA to preemptively block chemicals that seem like they’d be dangerous”. This is the logical extension of that.
* horseless carriages won't avoid running into things whereas horses can see and instinctively won't run into a wall
* the greater speed of horseless carriages in the 1890s without modern safety infrastructure like signs and traffic lights does greatly increase the risk of injury
I'm sure in time the technology will evolve and there won't be a new car on the road that wouldn't support FSD - just like it did with "horseless carriages". It will take some more time though and not all companies will be equally advanced. From what I've heard (from other sources too) I wouldn't want to be driven by a Tesla FSD at the moment.
> * horseless carriages won't avoid running into things whereas horses can see and instinctively won't run into a wall
Off topic, but interesting - if you think about it, we actually already had mostly-FSD in the 1890s. It just ate a whole lot of grass.
A secondary implied point is that new technologies which may seem dangerous can still be safe and useful when used properly, such as driving horseless carriages carefully in the 1890s or using FSD (Supervised) with proper supervision as directed.
On the other hand, horseless carriages also don't spook.
I've yet to have my car buck because it didn't like how something looked or sounded.
But maybe FSD will bring back that "fun"!
I live in a city with an unusually high number of highway miles per capita. I have to drive 24 miles each way to and from work. AI that can complete this task for me, safely, is pretty cool, but still largely hypothetical and not available to most people. What I'd really like is public transit and an honest national conversation about urban sprawl. My metro area is more-or-less three Manhattans placed side-by-side, but with 2.5 million in population. That's insane. Another automotive technology isn't going to solve that.
The solution to traffic is the same as it’s always been: increase throughput on fewer vehicles. In other words, you should take the bus or the train or increase carpooling or really anything that gets fewer vehicles out of the transportation network. Most importantly the solution is to be less individualistic and more focused on the problem as a collective. Self driving cars are the opposite of that solution.
If everyone already has cars, it takes a lot to get people out of cars --- if someone already has a car, taking the bus to go somewhere slower than driving is completely out of the question. Conversely, if everyone doesn't have cars, it takes a lot to get people to drive cars due to the upfront cost of car ownership.
I think more transit-oriented city development is definitely the way to go. But nibbling away at the problem is going to take a long time. I think robotaxis are definitely going to play a part in this: by being compatible with existing car-dependent infrastructure and providing a similar speed and flexibility as privately owned cars, existing car owners are more likely to give up personal car ownership. Once people stop owning cars, they would be more amenable to development that gets rid of parking spots, improves walkability, increases density, implements mass transit, and so on.
As for buses, they have many of the same drawbacks as personal cars (gets stuck in traffic) while having more disadvantages like slower speed, making them an unattractive option for people who already own cars. Buses are incredibly good for already dense cities where people aren't already car-dependent, however, and some of these drawbacks can also be overcome with dedicated lanes and other tricks.
In my opinion, the king of urban transportation is undoubtedly the grade separated subway or elevated train, which, if implemented properly, can have a speed advantage over cars in some cases since they are immune to traffic and weather conditions despite needing to have many stops. Of course, this is also expensive to build and requires high density to be viable.
Hence they would increase traffic and slow down commutes. Maybe we’re fine with that, but we should at least be honest with ourselves.
My opinion: if you look at what the average American pays to have their own vehicle (fuel, maintenance, depreciation, insurance, financing), I think you can make the argument people are literally willing to pay hundreds of dollars a month instead of sitting next to people on public transit.
Would this be the case if public transit was better? Not sure which is obviously your point. I guess I'm trying to argue "the demand for not-public-transit is there, so maybe that's why public transit isn't improved upon"
I know it's popular with left wing Youtubers to say the car lobby did it.
But it didn't: people wanted cars and governments built roads.
Why does everything have to be a goddamn conspiracy?
The US had a vigorous middle class when the Interstate Highway System was implemented. A family of four could own a home in the suburbs on one income, and it was fine to spend 15 minutes driving from said suburban home to the office. None of that is true any longer.
Americans who have never left this country (the vast majority) have no idea how competently ran government services can look like.
We also have a party in the US that actively adds more regulatory pressure IFF the government is the main stakeholder in a project. For example, environmental protections are significantly stricter (arguably, to a fault) if the project is being led by the government.
Also, for example, the government is not really allowed to negotiate Medicaid medication prices. Something private insurance does not need to deal with.
I also believe FSD cars that actually can go on any dirt road are still at least ten years away, maybe decades.
Then we still need to have a park of shared FSD vehicles big enough to become a reliable mean of transportation. This is also going to take a couple of decades. I won’t see this for sure.
Is it 100% perfect? Of course not. But this article is the ultimate in low effort. A person who felt that the car might hit another one but probably not. Yes, real value there.
What’s low effort in sharing your personal experience? It’s very clearly written as an opinion.
Any reader would be reasonable to assume that there are other opinions out there, and that this opinion may be in the minority.
That said, I’m happy to see a dissenting opinion in the comments here! Would be curious to see others thoughts in the sudden appearance of a free trial.
I find it extremely biased and lacking in substance.
I get it though... Elon does deserve a lot of hate so I guess the HN crowd is gleefully cheering out of schadenfreude whenever someone says something along the lines of "Tesla bad".
I hate to break it to you, but that's going to be a lot of people. He's made himself extremely hateable by the general public.
Hell, I'd argue they shouldn't be able to call it that legally.
Is a “garbage disposal” a device that can dispose of all garbage in your sink? No, all it can handle is light food scraps. But the inventor named it a “garbage disposal” not a “light food waste processing device.”
To me it seems stupid that so much of the debate after all these years is about the name. I think we are beyond the point where the name matters.
I wouldn’t.
A self driving system that only works during good weather is still useful and has non-zero value.
If you want to use an arbitrary name, use one, if you pick a descriptive name and don't do as described, it's just false advertising.
Again, I just think that marketing is marketing and everyone knows that. Everyone knows that a trash bag isn’t indestructible and that a paper towel can’t absorb a tsunami.
Especially for a device that has health impacts.
Typically the FTC regulates truth in advertising by (1) Is it an unequivocal statement? & (2) Does it have a health impact?
Depreciation is a normal and expected part of buying a car. If you are financing a car you should be prepared to keep it for a long period of time. While many automakers offer car loans of 5+ years, ideally you should only finance a car for 3 or 4 years and/or put more money down to ensure you are never underwater.
At no point did the author say that the car didn’t do its job as a reliable piece of transportation, but somehow it was a “financial mistake.”
The fact that cars depreciate and cost too much is a feature of all cars.
The problem is that Tesla owners are spoiled. I bought my Model Y just before the supply crunch in June of 2021, and within a year used models were selling for 20% more than I paid. They've since come down a ton as demand levelled off, but even now my car has depreciated only about 18% (including inflation!) from the cost of a brand new Y with the same configuration.
It's easy enough to imagine a young Tesla buyer over the last few years getting a very weird picture of the financials of owning a car.
What Tesla did, which historically is not a "normal and expected part of buying a car", is repeatedly slash list pricing after launch, thus meaning the used market does not trust valuing them, as so many used car dealers got very badly burned on inventory they paid too high a price for when the new car prices came down, putting them all underwater.
These aren't small or predictable market moves either; they are random swings. The Model S and Model X have been reduced new by as much as 24% at times. This Simply Does Not Happen at most car companies, most of the time, and residuals are more stable as a result.
Imagine you had just taken stock of a used model S, valuing it based on purchase price "n", then tomorrow the manufacturer slashes the price of a new one by 24%. That's very, very hard not to end up underwater.
Similarly, for existing owners - that 24% cut has just shifted the depreciation curve massively, and not in your favor. Existing owners frankly have every right to be pissed.
+ Parts availability woes
In short, buying a Tesla is a terrible decision from a financial perspective.
Maybe from a fun or cool perspective!
But if you're interested in maintaining $-value-over-time, there are much better options (Corollas, Accords, Tacomas, Subarus, etc).
lol, walk me through the process at Maserati.
If residuals plummet too much, their lease accounting can go red too. Lease rates have to account for a future predicted value of the vehicle, and manufacturers strangely like making money on leases.
Also, the huge depreciation on a Maserati isn't a surprise, most of the time. A mass produced vehicle selling in the hundreds of thousands annually seeing double digit list price shifts at random did come as a surprise, and has put car dealers underwater on Tesla inventory before.
Because those are two things that Tesla is doing that nobody else is doing at all or very well.
At daytime or ar night?
What about the weather conditions?
Considering you are putting your life, that of your family and everyone on the street around you in the hands of this software, I don't get the nonchalant "sure, it's not perfect" attitude like you were trying out the beta version of a new note taking app. It absolutely has to be 100% perfect before being allowed on the streets.
How about hardware? Is it all 100% safe? IIRC USA doesn't even require yearly warrant of fitness / MOT.
Are the roads perfectly designed and in perfect conditions? How about weather? Do you stop driving when there's blizzard outside?
If anything, FSDs are over-cautious and slow down to well below speed limits if they can't make out conditions. So, yes I'd say other drivers shouldn't complain about a system that still requires a human to constantly pay attention and keep their hands on the wheel.
It makes driving a lot more relaxing in a way that you don't appreciate until you get used to it. My feet are off the pedals and can stretch, I don't have to steer so my attention is only partly on the road ( the interior camera monitors to see if you're looking ahead and not on your phone for example ) If I have to change the music I don't have to worry about the car in front of me suddenly braking etc.
I've been using FSD for years now at day and night and I think it's been ready for Level 3 going straight on Interstates/highways for a while now. The latest update significantly improves taking turns along a route.
I commented on this thread elsewhere that I wish there was a standard suite of requirements for all automakers to certify the self-driving levels on their cars. Otherwise, I feel too much of the discussion is being polluted by people who agree or disagree with Elons politics.
I hope you don't hurt someone.
From my experience working in this space though, judging quality is extremely difficult. For one thing, different people have different criteria. I've taken rides with other people where we all had different opinions on the quality of the ride at the end. These systems (FSD especially), can also be highly sensitive to specific details of the situation. Two drives in apparently similar conditions have meaningfully different behavior because the system understands them differently. It's very difficult to compare apples to apples outside statistics and simulation.
All this is to say: don't dismiss feelings or intuition about the danger. These are good caution signs and talking about them can help companies improve the product.
I will tell you straight up that a Tesla model Y on FSD 12.3.3 is a safer vehicle operator than a human being using a phone. Period. No qualifications. Under all circumstances.
With every new release of FSD, there are people quick to say it's a breakthrough, and amazing, while others lament it doesn't nearly do what it says on the lid.
Why are we putting safety critical systems we can’t adequately measure on our streets?! There are people who confidently state that Tesla's FSD is already better than humans.
Where's the data?! Why isn't it open? Why can't we build an objective measure for effectiveness and safety of each of these releases?
I'd love to see all cars (including Waymo) compete on a standard set of tests with clear gates as to what conditions they'd be allowed to go to Levels 3, 4 or 5.
As an unrelated aside, measuring safety performance is hard. Doing it accurately without putting vehicles on the street is probably impossible. Simulation is not a fully adequate replacement for road testing and many companies are already looking at cutting the latter because of the expense. It'd be nice for regulators to outline an acceptable road test framework that better balances the goal of public safety from unsafe vehicles against ensuring safety can be demonstrated in real world environments.
https://www.dmv.ca.gov/portal/vehicle-industry-services/auto...
Of course Tesla is not participating - the real numbers would make them look disastrous.
Regarding objective measures for the quality - the famous standards for level 1-5 etc. Objectively speaking, Tesla's system is Level 2. There are other companies at level 3, and some others operating fully autonomous vehicles.
As to why Tesla chooses to base all their claims on purely subjective and highly cherry-picked numbers and videos ? We all know the answer to this question.
Which I'm happy with. I do 95% of driving on Autopilot already. Add another 4% for FSD that I'm happy to babysit and I'd be very happy.
For some reason people are so caught up on naming choice and what it implies...
I find it fascinating that a guy become the richest person in the world, by convincing 'investors' that a Level 2 system is 'almost ready' for a Robotaxi operation and will make them infinite money. To give him credit, he also did mention that Tesla is worthless if it doesn't solve FSD.
Driving is a very personal activity. There are nearly infinitely many variables driving every decision during every drive, and the default experience that people will want from something like this is "someone that mostly drives like me and never makes mistake because computers".
FSD is not that. It will be a while before it is that. Until then, I can totally understand someone "never using this again" (for a while) over a few fumbles that FSD made that they would not have (in most cases).
Tesla has to know this because they also know that if even 10k Tesla buyers decide to subscribe to FSD, that's an extra $2M/month that can be (mostly) shoveled back into training infrastructure!
If you want to get an accurate picture of FSD 12, watch random youtube videos: https://www.youtube.com/results?search_query=fsd+12.3
The majority of people are over the moon about how good the latest version is.
We have it on our Y, we're in suburban America, and its average number of attempted very-obvious moving violations per 15 minute trip seems to be about 2, and 1 attempt at curbing the wheels or something similar.
It doesn't seem to have any concept of things like conditional school zone speed limits (it consistently ignores those), school busses with their stop signs out, or any unusual scenario like that. I'm not willing to try it in anything dicier at the moment, like taking turns at an all-way stop. Even multi-lane left turns are hair raising, because it doesn't seem to consistently stay in its own turn lane, instead drifting toward the car next to us.
This experiment has ended for us for now, it's much too stressful to actually use, like being driven around by a bottom-percentile drivers' ed student. It's an incredibly impressive bit of technology, but simultaneously nowhere near good enough to not be an incredible liability.
I'm biased for them, if anything - I’ve been a shareholder since before the Model S launch - but it's not there.
There is an entire army of influencers who are both - shareholders and Tesla owners - and who promote FSD and everything Tesla pretty much as a full time job now. Whole Mars Blog is probably the most prolific one.
Self driving cars have recently been banned in British Columbia. I wonder if this applies to Tesla FSD?
Caveats: I don't let it make left turns in traffic, too stressful. It's sketchy around yellow lights (go through, please). It can aggressively brake for cars getting out of the road.
It makes me a more relaxed, less aggressively driver, er, supervisor, and it's quirks are pretty easily mitigated by taking over the accelerator or just reclaiming control.
(It was genuinely scary though!)
of course that doesn't include 1000+ serious accidents without death via FSD
only the Mercedes has real Level 3 Autonomy certification, everyone else is just using you as an alpha tester
I would say that is obviously not full self driving - but I suppose I'll never succeed in marketing with an attitude like that.
By that standard, FSD is still at Level 2 out of 5. (But probably approaching level 3 in the next few years.)
FSD as Tesla markets it (above the asterisk of course) would be level 5 autonomy where the car can complete a trip to any destination completely independent of any driver intervention. The current tech is not that.
The best autonomy I've seen is from other players like GM's Cruise were the car can complete trips without having a driver on hand to intervene... but only in certain controlled areas / conditions.
Wouldn't this be the case with Tesla's FSD? I make trips that are fully driven by FSD (I'm using it again this month, but I had a 3 month trial when I bought the car), would you consider routinely successful operation within non-prescribed boundaries a "certain area"?
I guess the disconnect between the critics and fans (I am one) is over the marketing; my understanding of "Tesla Full Self Driving" is not level 5 autonomy, it's "can drive better than my teenager most of the time".
I (think I) understand the catastrophic downside risk of computer operated machinery in an uncontrolled environment, but my experience has been very positive. I think I understand the feature for what it is, an amazing and useful if imperfect tool, regardless of what it's called.
(caveat: FSD is the only automated driving I've experienced)
Systems where the human is not directly operating the vehicle are referred to as autonomous. You may also see them referred to by the SAE level system as L3, L4, or L5, each denoting autonomy in successively larger parts of the driving space.
When the leasing contract came to an end, we got the refreshed model 3 from 2024, this time without any driving assistant, just the plain normal car. And I haven't missed any of it, just the park assistant was nice because now I can't park it on the side of the small European streets first try, but I'll learn eventually.
It was a horrible deal back then for 5000€, now it's more expensive and hasn't gotten any better.
Underappreciated observation. The "interstate-only FSD on vehicles it's uniquely suited for" seems like much better risk-reward.
A less complex and variable environment, on a vehicle with greater need for it: freight trucks, long distance buses, and RVs.
Hell, I'm pretty sure the only reason Tesla didn't just buy Greyhound/Megabus was because the stricter scrutiny from CDL regulators (FMCSA?).
Each release likely has gotten measurably better, and more usable, but at this point I seriously doubt that V12 is "the one".
Fool me once, shame on you. Fool me 11 times and I'm just going to buy a Mercedes EV.
Simple solution, get an even heavier tank. Where it starts getting tricky is when you are dealing with people who also value lives on the wrong side of the windscreen.
Lol
We know well at this point that progress, especially when it comes to AI, is anything but linear
I never owned an electric car also I refuse to pay subscription services for the car.
Aside from Musk being an online troll, if the product is great it should not impact the investment. I think lot of Tesla owners got sold on the hype.
Or if your self-driving car struggles with tight corners, or complex junctions? You won't see many of those on the freeway.
On the other hand, if your self-driving car suffers random failures to stop, unexpected swerving or a propensity to hit into stationary objects - I agree lower speeds would be safer.
I think it also would have been easier/faster to build networked vehicles that can communicate with one another over a standard protocol, but for some reason the industry didn’t opt to standardize things in that way, and now we have a bunch of rogue driverless cars that make independent decisions instead of taking into account each other’s plans.
You can call those driverless road lanes "tracks" and put some trains on it.
For what it's worth one of my clients installs charging stations and after meeting maybe hundreds of Tesla owners I assure you(anecdotally) that 19 out of 20 don't mention the CEO at all. Most Americans don't even have a twitter account(less than a third do) so this mindset of associating your car brand with all that e-drama is only for people chronically online. Most are non car people(as in, they wouldn't have bought a BMW or Mercedes, they were happy with a toyota) just excited about the tech that goes into the car. It means no more to the average driver on the road than what Henry Fords' opinions were(and they were worse)
Maybe true about Twitter or Elon specifically, but i suspect if we ignore "on Twitter" specifically, far too many people are "chronically online" or chronically drinking news cycle drama.
Ie i generally agree that people often think e-drama is more real than it is, but i also suspect we shouldn't undersell how many humans are addicted to tube fed drama. Fear sells, and we've got a lot of industries that focus on this.
Wonder what studies test this, outcomes, etc.
FSD on my Model Y is fun. There's no reason to fear because it's extremely intuitive to take over if it makes a poor decision - any resistance on the steering wheel immediately disengages Full Self Driving.
I've used FSD for thousands of both highway and city miles, and like any new experience, you shouldn't judge it based off of your first few tries.
After a short time, you learn which scenarios it handles with ease (highway driving, clearly marked roads with other traffic it can follow), and which scenarios it struggles with (dips in roads, emergency vehicles on shoulder, large open intersections without clear markings, construction zones).
FSD is basically magic. I'm a software guy, and I'm shocked how well FSD handles an insane variety of situations.
In my local town, I feel I'm quite a bit better driver than my FSD. However, when I visit an unfamiliar city, FSD often handles new situations better than I would have. We're living in the future.
Also, there's several comments here about how timid FSD can be merging or at a four-way-stop or similar; very true! Yet you just tap the gas pedal to let it know it's safe to proceed, and FSD takes it from there. Or just manually accelerate, and FSD will happily handle the steering while I handle the speed.
>And let me be clear: nope. Nope nope nope nope nope nope nope.
Saved you the click.
I absolutely love FSD. It is THE feature that keeps me locked into the Tesla ecosystem. I am absolutely willing to die on this hill.
That said, I am not surprised and can empathize with the author's experience here.
FSD is an extremely new way of driving, an activity that is integral to most people's daily lives. One's enjoyment of FSD is inversely proportional to the sum of the amount of control the driver wants while driving AND the driver's expectation of FSD being an SAE Level 5 autonomous driving solution.
Many drivers want lots of control up front, and everyone drives differently (another problem that I'll save for another post). So between FSD being short for "Full Self Driving", FSD not driving 100% like them, and it being a $12,000 add-on (despite many drivers being completely and totally fine spending way more than this in aggregate on vehicle option packages that do way less), articles and Reddit posts like these OP sprouting like wildflowers in the spring now that FSD is temporarily free isn't a surprise.
I surmise, based on this paragraph, that the OP might err on the side of "more control":
> I’ve rarely been so frightened behind the wheel of my own car. As an experience it ranks right up there with driving that one stretch of the 880 where you’re supposed to go at 45mph and everybody drives twice that, the time I was riding in a Lyft and was jackknifed by a tow truck, or that one time I was driving on the 101 North and a tech bro Lamborghini shot out of the Highway 12 offramp and came within two inches of hitting me, spun around in the road, and careened off into the distance.
All of these are driver decisions that are tied to the driver's risk profile. One person's "jack-knifed" is another person's nothing-burger. Some people feel pressured to speed due to traffic flow; other's hold firm against the posted limits. (I've been both of these people!)
---
Here is the "I'll die on this hill" portion of my response to this article.
First unpopular opinion: I paid for FSD at purchase twice; once at $7k and again (this year) at $12k. (Okay, I rolled it into my finance loans, so I'm actually paying more after interest!) I don't regret either purchase for a second and would gladly do it again. There is nothing like Tesla's FSD Beta (FSDb) program.
I am okay with FSDb being imperfect, just like me.
FSDb isn't scary for me since I know that it will do wild ass shit sometimes (it hit a cone once during construction zone). However given that FSD mostly works as expected and spares me tons of toil that comes with daily driving, I don't mind a few mess ups here and there. (I mess up all of the time while driving by hand, and I consider myself to be a good driver.)
It's insane how much toil FSDb spares me.
My wife and I road trip with our cars. 2000+ miles; all on FSD. I will absolutely not road trip without FSD; no exceptions.
---
Second unpopular opinion: Nobody else can and will touch FSD with a 100-ft pole. This one requires some explaining.
Common retorts against FSD are:
- That it's basically adaptive cruise control and lane keep assist, - Waymo, Cruise et. al are way better, and
- It's inferior to hands-free assistance systems, like SuperCruise, BlueCruise and Comma.ai.
The first retort is completely false. Anyone saying this either hasn't used autopilot at all, has _only_ used base autopilot and is projecting to FSD, or has not _really_ used TACC and LKAS.
I've driven many rental cars with these systems. Having them present is definitely better than not having them, but they don't come remotely close to Autopilot and aren't even in the same universe as FSDb.
Here are some limitations of these systems that AP does not have:
- They disengage when roads with mild curvature are detected. - They DO NOT disengage in problem areas, like construction zones. - They often disengage silently with no warning. - Many of these systems are unable to center the car on curves they can handle. - They do not interact with traffic controls, like stop signs and traffic lights. (Autopilot without FSD will not stop for these controls but will scream bloody murder if you're about to ignore them!)
Then there is the absolutely slipshod UX that accompanies many of these systems, such as: - Them having several duplicative configuration interfaces (I.e. you can configure them either through the menu on the dash or the menu through the infotainment) - Them having confusing ways of telling you when they are active - Them allowing you to keep auto-steering/LKAS on without cruise control enabled, causing you to fight with the system when you think it is not active but it is.
The second retort is closer but misses the forest for the trees. Indeed, FSD underperforms compared to systems supplemented with LIDAR and RADAR. However, they are (a) small-scale taxi networks and have been for a long time, and (b) are super geofenced and have a laundry list of exceptions whereas FSDb works in nearly all situations and geographies to the best of its ability.
This argument is like some posts on HN that argue that modern web stacks are way too complex and "you can run Uber/Yelp/whatever with a few VPSes, Redis and Kafka." All true until 100,000 people want to use your thing every day for all sorts of use cases you didn't plan for. Vertical or horizontal scaling alone won't work fast enough, and you'll go broke trying.
The final retort (BlueCruise, SuperCruise, etc.) is more correct, but ignores how Big Auto fundamentally works.
Those systems do enable a hands-free driving experience in the road segments that they support (most mapped highways).
However, the hands-free-ness of those systems is replaced with (IMO) extremely aggressive eye and face tracking that might or might not be update-able depending the system on board and the car's model year. They also rely on manufacturer-provided pre-mapping topologies and do not work at all when road conditions change (like interstate construction, which is a given on long road trips). Pre-mapped topologies also don't scale (imagine trying to support a rural road), so these systems working off-highway is unlikely, though this remains to be seen.
I am optimistic in Big Auto catching up and providing a FSD-like experience at some point. All signs are pointing to this happening, albeit more slowly than projected back in 2018. (To be fair, everyone was going ape shit over autonomous driving back then.)
However, Big Auto also considers autonomy to be mostly a feature to check a box next to. Tesla does not. They see autonomy as critical to their survival. This is a big reason why they charge so much for FSD.
Tesla is hugely outspending the competition on this effort. It's not even close. Dojo, their training system, has, what, HUNDREDS of H100's in their cluster? They also have handfuls of well-compensated PhD's in this domain working on nothing but autonomy.
I am almost positive that no other auto manufacturer has anything like Tesla's autopilot program. Investing in this would be a huge moonshot and financial investment for them, and they are (a) bleeding money in trying to keep up with Tesla's EV platform in general, (b) bleeding money in trying to do away with ICE by 2030, and (c) way too conservative to make these kinds of risks (see also: how they are handling the outdated dealer model).
The best alternative we'll probably get is Big Auto slapping MobileEye's new platform on their cars, but that's just another integration that is probably vehicle and model year dependent.
---
So there you have it: a semi-rabid FSD fanboy's take on this article. They're not wrong, but they're also not thinking big enough.
<s>Pssh. These people are letting the emotions of humans keep them from innovating, disrupting, and generally becoming certified Silicon Valley sharks.</s>
Someone trying an 11.x version and someone on 12.x will have significantly different experiences at the moment, with 12.x having some of the biggest advancements we've seen since Tesla began doing this. 12.x is still rolling out to fleet now, with only ~40% of cars on it if the online Tesla fleet version trackers can be believed.
I'm not suggesting it might change the author's feelings completely if he was on a 12.x branch of FSD, but the jump from 11 to 12 has been siginificant.
EDIT: not sure why the downvotes - its a legitimate question the original article doesn't answer...
Looks like the free trial cars are currently split across v12.3.2.1/12.3.3 depending on OS update.
My wife does this too, and has fishtailed on the highway when she has suddenly realised she’s drifted across two lanes while not looking at the road. She frequently decides to read and write messages while driving, with me and our infant daughter in the car.
I think there’s a tendency to vastly overestimate the competence of human drivers.
Further compounded by entitlement of freedom, imo. Yes, we all want freedom, but our ability to reduce public safety should be earned i suspect. We are very lax with driving safety and requirements (in the US at least). I mean hell, what percentage of people actually drive the speed limit, anyway?
Which is to say, the risks of driving have become so sanitized, that we think it's something that is easy to automate. Arguably, if all road lanes were made 18 inches more narrow, we'd have a lot of the same effects. You have to pay a lot more attention in a narrow lane than one that is super wide. Generally though, humans are very competent drivers when paying attention and patient.
The other part, it's hard to be patient given how much time is spent travelling via car and stuck in traffic jams. People get used to zooming around at 40 ~ 70mph, combined with the over-confidence of from the illusion of safety (all the road signs, extra wide lanes, road markers, obstructions removed from the road)...
I think it just combines to a place where humans are poor drivers because they become impatient and stop paying attention all the time. It's like that first day of work, where everything you do - you have to think about incredibly hard. Eventually it all becomes 'auto-pilot', and this is encouraged by the road infrastructure; even to the extent that we think it's an easy problem to automate.
I can't wait to remove the entire "fell asleep and hit a tree", "drunkenly drove through a stop sign and killed a kid" class of accidents, but that alone is not enough