EAP to FSD - bite the bullet now or risk it?

[P3dStealth]

It's funny I expected to get a lot of hate from what I wrote but we mostly seem to have the same experience or opinion.

I'm a Tesla fan and want it to work but it just doesn't right now. They should be lowering the price not raising it given the circumstances. And get these features out of beta before adding more.

 

[SammichLover]

Generally perceiving a fair amount of dissatisfaction with FSD, etc. My reaction is, hey guys, don't you like to drive? I think the S is one of the most fantastic cars to drive under all conditions and on any roads, and I don't want to cede my driving pleasure to a machine. Do you?

I covered this in another thread recently. I love to drive, and have from the start found that EAP, and now with the stoplight feature, enhances that experience. Most of the time I'm enjoying the road from my overwatch chair, my little minion taking care of the busywork. Then in a pleasantly refreshed state, since I didn't have to deal with the monotonous minutia that AP took care of, in a moment I flip the switch, in my head and the gear selector stalk, and get on it hard.

Sublime.

 

[Brexpat]

I am no more of a fan of my Tesla than I am of any of my other electrical appliances. My car is equipped with AP and sadly, what it gets right is overshadowed by the dangerous or careless things it gets wrong. Asides from random braking, AP turning itself off mid corner on a motorway and ill-judged corners; when in traffic on the German autobahn AP refuses to stay to the side of the lane and leave a way clear for emergency services. I rarely use it.

 

[SammichLover]

BTW that is why the ELDA fiasco hit me so hard, really pissed me off. Because when I say I'm in control I expect no interruptions or usurping. I remember my first vehicle with ABS, and this was late 90's at a point where ABS was already reasonably refined. It took me some time to get used to it being there. In the end I had to force myself to engage it some just so I mentally had it as part of the "expected result of my choices".

 

[kungfujoe]

I purchased a 2018 P3D that was equipped with EAP, so outside of HW3 and the newly released traffic signal features I have everything else FSD currently has. That said, I can upgrade to FSD for $4k, but in the past it has been $3k and once as low as $2k.

As a happy FSD owner myself, I say skip it. The recent hardware upgrade that let me see more cool pictures on my screen was nice, but it certainly doesn't justify the $3K it cost me when I bought the car in 2018. The most valuable FSD features are already in EAP, and the VERY few features that have been finally added for FSD are, so far, more annoying than useful (autosteer is now limited to the speed limit, instead of +5MPH, on many roads because of the feature which stops at green lights for you). More will be added later, but we have no idea when or how much longer they'll take to become legitimately useful (in the refinement process, some features have actually gotten worse - NOA generally works more poorly now than when it was introduced, though I'm confident it'll improve).

Moreover, while Tesla is indeed saying that the price will just keep going up, I don't believe it, just like I don't believe our cars will have the Level 5 capabilities necessary to make "robotaxis" by the end of 2018, err, sorry, 2019... no, I mean 2020! While Tesla is absolutely well ahead of everyone else with respect to most of these features, competitors are adding them, bit by bit, to their cars, and offering them at much lower prices. That market reality already prompted Tesla to make Basic Autopilot a standard feature, and there's good reason to believe that, in years to come, similar market realities will cause them to reduce the price on some or all FSD features, too.

I absolutely love my Model 3, and I can't say I regret paying more for FSD, but I also can't claim that the value of FSD comes anywhere near the cost.

 

[SammichLover]

when in traffic on the German autobahn AP refuses to stay to the side of the lane and leave a way clear for emergency services.

??? This is definitely in the set of roughly 5-10% of the time you should be immediately disengage and switch to manual, as soon as you see that developing. It took me some practice I guess, but the transition in and out is like breathing for me.

AP turning itself off mid corner on a motorway and ill-judged corners;

I saw comments like this from that UK owner further up, too. It sounds like the EU version of the NN isn't as baked yet?

 

[SammichLover]

the feature which stops at green lights for you.

I found this briefly off-putting, then realized it was training me to be more diligent in my overwatch job. Expect in time to be able to disabled that and have it only rely on the "nag" system instead, once enough confidence in the system is established. Just like Nav on AP went through that period, and now allows fully automatic lane changes as long as you're touching the steering wheel (or thumbing the spinners) often enough.

 

[Retired in AZ]

It's funny I expected to get a lot of hate from what I wrote but we mostly seem to have the same experience or opinion.

I'm a Tesla fan and want it to work but it just doesn't right now. They should be lowering the price not raising it given the circumstances. And get these features out of beta before adding more.

 

[Retired in AZ]

From your words, one gets the idea that you're not a Tesla fan, and my first thought is to tell you to sell your car because you'll never appreciate it for what it is - warts and all.

This is NOT the first, nor even the tenth thread where people are complaining about FSD. It would be nice to have a forum that could be searched so that someone could read others' comments and not repeat the same things....

On the other hand, P3dStealth, you come from New York, a state which is not as electric car friendly as others. Tesla has had mighty battles with the government over whether they can even SELL cars in the state without a 'dealer network'. And your viewpoint regarding the likelihood that a form of self-driving will be allowed makes me think that not much has changed in the six years since I was there.

As a retired IT guy, I want you to consider the astounding odds faced by Tesla programmers. It's easy for you to drive the car because you have a supercomputer between your ears, eyes that have a wide view of things in front of you, a huge cache of objects you identify instantly, reflexes that will permit you to operate controls semi-automatically, and so on. The task of the programmers is to figure out how to take Artificial Intelligence (AI) and train it to be able to do what you do - or at least close. In some regards, those AI programmers have to stop after the basics are in place and release the software into the wild, then gather more information from the real world experiences before going forward. That's 'BETA', and the information comes in in waves. Then, the AI programmers will figure out how to train the car to do the things it does more naturally. Autosteer is a good example: it was pretty blocky when released, but has smoothed out to get really close to what we do naturally. NOA is going through the same expansion. There are videos showing it working perfectly, changing freeways at the right intersection, exiting from them and driving in town, then parking. Under the right circumstances, everything needed to do this - for the route chosen - is already programmed. And every time you get an upgrade, you'll notice that there's also some bug fixes, taking care of the flaws or perceived flaws in the existing programs or adding sophistication to a long-standing feature, like Autosteer. Give the new release time to collect the information needed to improve and digest it. I think recognizing traffic controls and stopping at lights and signs is really basic, and annoying, because the software will stop at EVERY control point, whether you see it or not. It's a pain to drive around town and eel the car slowing down for a green light, then have to tap the accelerator or flip the stalk. But it will get there. Give it six months or so.

Enough rambllng from me.

 

[Tplamaia]

Can one ask for a refund on FSD ?

 

[hlin07]

It's funny I expected to get a lot of hate from what I wrote but we mostly seem to have the same experience or opinion.

I'm a Tesla fan and want it to work but it just doesn't right now. They should be lowering the price not raising it given the circumstances. And get these features out of beta before adding more.

I'm on the same boat, was one of the first to reserve model 3 during the night of unveiling. My wife was calling me crazy for putting in 1k on something NO ONE knew what the car even looks like.
But, in order for Tesla to become mainstream, not just in the high end luxury segment, but also say Camry/Accord territory (yes, they should be scared), price needs to be lowered.
Elon has repeatedly associate FSD as appreciating asset with Robo Taxi in the future. Well, say Elon is right, shouldn't Tesla drop price of FSD in order to have more people *beta testing* thus assuring the public the safety of this appreciating asset?

Going back to OP, whether to purchase the FSD is purely personal choice. If OP is budget minded, then NO WAY. But, if OP is someone that's comfortable in the upper middle class that have the money to spare, sure.

Personally, I'd rather pay 10k for a *sure thing* than 7k or whatever the price will be for the uncertainty in the future. My next car will likely be the CyberTruck, so I'm hoping FSD is fully operational then.

 

[SammichLover]

It's funny I expected to get a lot of hate from what I wrote but we mostly seem to have the same experience or opinion.

Hate? No, empathy.

I can see how you'd get to your conclusions. If your standard for there being value is "robo taxis", fully autonomous "rider-only" within this year authorized on public roads? Musk needs a good slap to the head for all contributions towards that expectation, and I'm at the front of the line urging that nobody buy on that.

I can see it being very easy not to grasp the depth of value in the features that are there with only a day of self-guided use. I'd guess that it took me a week to really start to grok it, and more to trust it much less rely on it. This is just EAP, forget "Full Self Driving" here because that's not what we're talking about. Is a paradigm shift in how you drive a car. Your job is different, you are the overwatch. A lot more captain, less helmsman. And not a micro-managing captain, either. The captain that focuses on the big picture and mostly lets things sort themselves out content with the way their underlings see fit to reach the ultimate goals.

I expect not everyone is going to want that. Certainly not realize it unaided, either. And like any management position there is a degree of letting go and trust. That isn't going to be built in 48 hours. Unguided I'd be surprised if most even grok at a vaguely conscious level within a week the kind of change that is needed, much less acclimatize. I know I didn't fully, only looking back a lot longer could I start to put it into words.

Until then you aren't going to be trying to let go, and there is no trust building. You can't be an effective manager without trusting, and you can't trust without listening and looking for what to trust. If you instead focus on "it should do this instead, I don't like how it is getting the job done", well that's not going to work out. Usually with people, and here even less chance since it isn't going to integrate directions and your thoughts about how to get the job done. ((That can only happen via the indirect path through Tesla.))

You have to set aside your inner control freak, your ego, to take advantage of this. Otherwise you'll be at war with it, until you do. Not everyone wants to do that, not everyone will. And it is no quick process.

Yet EAP that had a $5,000 price tag is widely liked thing among those that bought it, your assertions notwithstanding. I find in practice the current base AP by itself is a pale shadow of what the car can do, those features you dismiss out of hand really, really grow on you in weeks and months. The FSD moniker is brining a lot of baggage to this about some expectations built that have not yet been fulfilled. Musk as much as anyone is responsible for that, but that doesn't make it an accurate map of the reality behind the baggage.

 

[Zenzation]

Well, as an early adopter of FSD, I have enjoyed each new update they have sent me. It was well worth the money to me. The FSD is one of the main features that separates Flo (my car's nickname) from other vehicles. Unlike many of the experiences I am seeing reported here, my FSD gets better the longer I use it, as if it is learning my route. For example, there was a stop sign on a side road that Flo interpreted the first two times I went by it as a stop sign on my road. She slowed down the first two but has ignored it ever since, apparently having learned it was not on my road.

I love how Flo now stops at stop signs and red/green lights. I don't have to come out of autopilot any longer when I come up to a red light or 4-way stop, and tapping on the gas to tell her to keep going in a green light situation is no biggie. I can program how aggressively Flo changes lanes, and I have mine set just right. She used to take exits too fast but has since learned not to.

I think the problem many folks are having is that they are not patient with allowing the car to learn their routes and their driving habits. Also, you have to help the car a little bit, which I don't mind. Flo and I are good friends and we do everything together.

 

[SammichLover]

Personally, I'd rather pay 10k for a *sure thing* than 7k or whatever the price will be for the uncertainty in the future.

This is entirely out of wack with the reality of the market value.

10K is somewhere near an order of magnitude short of where a rider-only, full autonomous system on the general network of public would be priced any time within the next 5 years. Because a $100K of hardware per vehicle isn't able to let Waymo do this, and they aren't going to be there within 5 years, either.

That is something Musk in cloud-talk mode got dead right, even if his framing wasn't as helpful for how to utilize it, for what conclusion to draw from it. He made it far too easy to come away from it with "we're gonna be rich!" No, "it's gonna be hard, long slog" is a better takeaway. Otherwise why would Tesla be pricing it 14x under that, right?

 

[Mike K]

As a retired IT guy, I want you to consider the astounding odds faced by Tesla programmers. It's easy for you to drive the car because you have a supercomputer between your ears, eyes that have a wide view of things in front of you, a huge cache of objects you identify instantly, reflexes that will permit you to operate controls semi-automatically, and so on. The task of the programmers is to figure out how to take Artificial Intelligence (AI) and train it to be able to do what you do - or at least close. In some regards, those AI programmers have to stop after the basics are in place and release the software into the wild, then gather more information from the real world experiences before going forward.

With all due respect, it's not about how astounding the odds are or what a difficult problem this is to overcome. I don't think anyone doubts that it's a difficult problem to overcome and that's the point. Surely Tesla knew the same thing in 2017. So were they incompetent or were they intentionally misleading people so they could keep selling cars? Because it's hard to see any other option. The problem is managing expectations, releasing half-baked "features" that hardly work or don't work at all and basically selling a bill of goods that realistically was not ever even close to being delivered when promised.

Those of us that have been around for awhile have a pretty good understanding of what this is. Tesla had a falling out with Mobileye over a disagreement in how aggresive their software was being used with Mobileye's hardware. The relationship was severed and suddenly Tesla found itself still needing to sell cars, now without a sensor suite. What keeps the iron moving off the showroom floor when you don't have the product to back it up? Promises that the widget they're selling you now is going to do something grand in the future even though at the time of release the AP2.0 cars didn't even perform basic autosteer functions.

But hey, 8 cameras, long range sonar and the promise (as I recall; I might be wrong) from Musk himself stating that these cars were going to be driving themselves in a year. That was 2016. We're a full four years and two additional auto-pilot revisions since then and we still don't even have the cross country driver-less trip that was promised by the end of 2017. Do you have any confidence whatsoever that there was ever a hope they'd roll this out in mid-2017? Of course not. They're marginally further along now. This isn't a feature problem. Tesla is leaps and bounds in front of the competition so far as I can tell. It's a communication problem. It's a managing expectations problem. And at it's core, it's an arguably intentionally disingenuous claim that they were going to be able to provide a product in 2017 that they don't appear to be close to providing in 2020. So I think the FSD anger is well deserved and I don't begrudge anyone that feels they've been duped.

I can't imagine the level of disappointment I'd feel had I purchased a new car and had to shell out $7000 for auto-lane change and self parking. Insurmountable levels of disappointment. And since these options follow the car and not the owner as they reasonably should, you risk losing that investment from day one if your car is totaled. Not to mention all of those who turned in their leased FSD cars and saw absolutely nothing for their money. Uggh.

And to re-iterate, I am a huge Tesla fanboy almost to a fault. But let's call a spade a spade.

 

[SammichLover]

Tesla is leaps and bounds in front of the competition so far as I can tell. It's a communication problem. It's a managing expectations problem.

Absolutely.

Musk has a way with words that is highly aspirational, a way of using technical words to evoke emotion that isn't really what the words mean. If you don't ground yourself studiously, understand the technical meaning behind them, when listening/reading his stuff and editorialize heavily as appropriate it is easy to get drawn out upon waves of dreams.

I can't imagine the level of disappointment I'd feel had I purchased a new car and had to shell out $7000 for auto-lane change and self parking.

So then let's call a spade a spade. Lets add a key missing part into that sentence, stoplight/stop sign. Within a reasonable accounting of the beans, I did exactly that. And I'm not remotely disappointed, I'm very happy with my purchase. I captain** the freakin' future, today. At a bargain of bargains....well I did have to drop north of $50K on top of that, but I got a hellva piece of physical car to drive with that, too.

I expect a large part of why I don't feel disappointed because I didn't buy FSD in 2017 or earlier, or even on the purchase date of my Model 3 in 2018 (FSD set off all sorts of vapourware alarms with me at that time). Nor did I take seriously the particular dates of Musk's aspirations, takes a lot of the sting out of being deceived when you aren't actually misled. I guess I've always been good at picking out his words, sorting the BS wording he uses from the solid, honest stuff he does put out there.

That'd be a very rough thing to go through, what you talk about in your post, those 3 years before I finished laying down the last bit of cash for FSD. Good fortune in timing for me on when I got to the Tesla party, I suppose.

** On reflection I think I'm going to use that word going forward. "Captain" is a better one than "drive" to encapsulate using the software formerly knows as EAP and the current FSD.

 

[ACofNY]

The stock owner in me says buy it. The car owner in me says, save your money.

If you are like most Tesla stock owners, why not pay back with some of your gains and get the FSD? You probably had more than $7k gains.

 

[Bob Denny]

I’m right there with you. My 3 month old MX, with the beta 2020.16.2.1 software is OK, but nothing resembling autonomous. I’ve resigned myself to expecting it to do OK on the freeway, but it has problems there too. If it needs to get over to exit or change highways, and there is a vehicle in the lane to the right going the same speed as you within your car-length spacing, it just sits there until you flip out and take control to slow down and pull over behind thatvehicle. It doesn’t know how to slow down to get into the lane to the right? Also I hate driving in people's’ blind spots, especially on the right side of a big semi! It will happily drive in the very most dangerous places like that. My friend in the auto body collision shop says that the #1 cause of crashes is being in someone’s blind spot. FSD doesn’t care at all. I’ve seen all the goofy and dangerous behavior on surface streets. Slowing down suddenly approaching a green light is prescription for getting rear-ended! Plus you can’t twig the stalk too early or a yellow light will get you!! So it’s a balancing act. Navigating on AP, exiting at a ramp, approaching a green light where you need to make a turn, it freaks out. It should be telling you that it needs you to take over MUCH sooner. It kills me that I spent all that $$$ on a promise.

And I don’t care how hard it is for the programmers. I’m a lifetime developer and I learned early on that promises are cheap.

 

[Knightshade]

Maybe in the next year FSD can stop at traffic lights.

It already does in the US, with other countries coming later this year.

Can one ask for a refund on FSD ?

Sure. Though if it's been more than 48 hours since you bought it the answer will be no.

Unlike many of the experiences I am seeing reported here, my FSD gets better the longer I use it, as if it is learning my route. For example, there was a stop sign on a side road that Flo interpreted the first two times I went by it as a stop sign on my road. She slowed down the first two but has ignored it ever since, apparently having learned it was not on my road.

The car absolutely does not learn individual routes like that.

The only reasons behavior would change in an identical situation is either:

A firmware update to improve all cars behavior on correctly recognizing what road a sign it sees applies to.

or

A map update clarifying which road goes with which sign (since the system is using maps as a primary source for traffic control device locations- which is how it can alert you about an upcoming one from 600 feet away even if it's not yet visible)


if each car actually learned and changed behavior individually that'd be an absolute nightmare for Tesla to ever troubleshoot problems or improve the fleet-wide code.

I think the problem many folks are having is that they are not patient with allowing the car to learn their routes and their driving habits.

Again- not something the car actually does.

Apart from things the driver can manually set (like follow distance or aggressiveness of passing while using NoA) behavior only changes with firmware updates (same changes for all comparable HW cars)- or map updates.

 

[drj3]

I also have EAP and love the NOA and auto lane changes. To me the additional features of the FSD for those relatively few of us with the long discontinued Enhanced Autopilot still are not worth it. My wife talked me out of the FSD upgrade when it was still $2000. I spent the money on Tesla stock instead and am much happier. I could sell and use the profits to pay for the FSD upgrade, but to me it still is not worth it. My advice - spend the $4000 on Tesla stock. If FSD becomes worth it to upgrade, the increase in stock price will keep up with or surpass the increase in FSD price - and you are still supporting Tesla.