Poll: Now that they are increasing FSD from $3k to $5k, will you be purchasing it?

[Electroman]

On highways with decent lanes, the current AP does an excellent job. Just reduce the nag and voila we are at L3 on highways.

 

[JeffK]

@JeffK that is not evidence of anything being used.

It actually is evidence the hardware is being used because that data is from the car... What it displays to the user on the in-car screen(s) is not indicative of what it's actually doing behind the scenes or what it's capable of.

I would say the latter.
Nothing I have seen up to now - from vehicles that are road-legal that is - makes me think that FSD is anywhere near ready for market

Have you seen the Waymo cars or the MobilEye demos? Both are in street legal vehicles. Using FSD everywhere is not yet legal, but that doesn't necessarily mean it's far from market. The data collection alone to prove it's safe could take years depending on how many cars are on the road. Tesla has a much better shot than everyone else on this front.

There are Lvl 4 vehicles right now, that's not debatable. They exist, but the barrier right now is proving safety and handling edge cases.

 

[AustinPowers]

True, but I am not interested in paying 3K (or 5K) in order to provide data to Tesla to develop a certain feature that might one day become of use to drivers. I see no added value whatsoever in doing so.

 

[ℬête Noire]

There are Lvl 4 vehicles right now, that's not debatable. They exist, but the barrier right now is proving safety and handling edge cases.

Then it's not actually successfully Lv 4 yet.

After that there is a vast sea between Lv 4 and Lv 5 when you implement Lv 4 as Waymo has, heavily reliant on their pre-mapped data within the system's geographic limits. The same reason that SuperCruise is a dead-end.

 

[TheLocNar]

Tesla suckered me in. What’s another $3k after spending $75k?

 

[d21mike]

I don’t think EAP will ever allow you to go hands free, with the exception of limited Summon which is grandfathered in. It’s even possible they’re getting more aggressive on the nags to make the FSD option (like the example above in traffic) more enticing.

I agree with @Daniel in SD. I believe what you are describing is EAP. I do not believe you can have ON-Ramp to OFF-Ramp and also have a lot of nags. To me FSD is more about surface streets from when I leave my house to when I get to my destination. But since I got FSD I am not worried covered if I am wrong.

 

[insaneoctane]

While FSD has the more complex job of navigation on complicated surface streets, if EAP is expected to offer level 3+ on the freeway, I would argue that it has more liability due to the significantly higher road speeds found on divided highways. Crashes at 75 MPH are far more dangerous than crashes at 35 MPH

 

[JeffK]

Then it's not actually successfully Lv 4 yet.

After that there is a vast sea between Lv 4 and Lv 5 when you implement Lv 4 as Waymo has, heavily reliant on their pre-mapped data within the system's geographic limits. The same reason that SuperCruise is a dead-end.

Unfortunately you've just contradicted yourself. I cited that there exist level 4 vehicles, you disagree and say there aren't and then say "when you implement Lv 4 as Waymo has"

This is by definition level 4 (in a defined use case):

Just so we're all clear, Tesla is shooting for level 4 first. Level 5 is still a ways off.

 

[ℬête Noire]

Unfortunately you've just contradicted yourself. I cited that there exist level 4 vehicles, you disagree and say there aren't and then say "when you implement Lv 4 as Waymo has"

As Waymo is implementing but hasn't finished yet. They're getting closish (they've put huge pressure on themselves to try push out a commercially operating test fleet soon). Saying there's Lv 4 vehicles is a bit like saying there's fusion power plants (as a more extreme example). We are able to create mostly contained fusion reactions and collect power from that but it's not like they work in a functioning commercial sense.

The difference between LV 4 and LV 5 is really just a matter of scope definition, what you decide to define as "all use cases" can place you in one or the other. And Waymo has a MUCH narrower scope than what Tesla is aiming for, geographically. In some senses going from LV 4 to LV 5 is super easy if you want to play philosophy semantic games, but I'm talking in a much more real sense.

What Tesla is doing is attacking the problem in a direction that's a huge reach to get to 4-5 but when they get there their coverage is very broad and flexible and inherently putting you close to "real" Lv 5. Waymo (and pretty much everyone else, but Waymo's clearly ahead of that pack) is coming at it from the direction to shortcut rush to 4, with a much larger per unit budget so aiming at an entirely different market at this point, with the consequence that the last leg to "real" 5 is a really long way. That's at the root of what Musk's comments about LIDAR leading towards a "local minimum" was getting at.

 

[barjohn]

@JeffK, when I say being used, I mean applied to EAP, not merely capturing data. As I recall, the poster of that video superimposed CAN bus data on the video information. He created those images, not the system. (I'm going from memory so I could be wrong). If the system were capable of producing and using data as shown, how hard would it be to create appropriate bit map images to display on the screen? As a programmer, I can tell you that would be simple. Further, I would correlate the radar targets to my image targets and lock the two together. I understand how difficult the real-time image processing and identification is to identify and track multiple targets simultaneously in real-time. In the military application we do it with radar targets all the time. I think visual is harder. To reiterate, I see no evidence that EAP as it currently exists is using visual information other than for lane identification and even there, it is very short range. Have you noticed that curves are displayed as straight lines? Compare to Comma.ai display that actually shows the road curvature. If a handful of programmers can do that, why can't Tesla's large team do the same?

I would love to see EAP capable of delivering level 3 or 3+ performance; however, it has a long way to go given the current rate of change/improvement and the fact that each new release appears to have areas in which it regresses. Reading the comments from the Tesla software developer on reddit (assuming he is real), it appears to me that they lack some of the disciplinary processes needed to adequately test new releases and minimize bugs. At least, that is my impression.

I'm a big believer in the small team approach where each member is hand picked for their skill at a particular function. Small teams produce far better code and produce it faster with fewer bugs and complexities. Large teams are like trying to give birth in one month with nine people. It doesn't work out too well. More time goes into management than production and testing is made much more complicated.

 

[JeffK]

@JeffK, when I say being used, I mean applied to EAP, not merely capturing data. As I recall, the poster of that video superimposed CAN bus data on the video information. He created those images, not the system. (I'm going from memory so I could be wrong). If the system were capable of producing and using data as shown, how hard would it be to create appropriate bit map images to display on the screen? As a programmer, I can tell you that would be simple.

Two things, as a programmer you will completely understand that a backend system doesn't have to render anything graphically in order for it to be used. If you are saying that although the data is clearly generated by the hardware/software combo that it's not used in the front end display then sure, totally agree. Average drivers don't need that information. Even just to render it, you'd need to transform the data from a computer usable form to a form a human can see. What's the point? It's good for debugging, but your role in the Tesla is as a driver.

Would you rather them improve and spend more time on the backend or uglify the front end and overload the driver with information gathered by the system?

To reiterate, I see no evidence that EAP as it currently exists is using visual information other than for lane identification and even there, it is very short range.

AEB uses cameras and radar. So to sum it up EAP is using vision for EAP tasks which require vision such as lane keeping and AEB.
What else did you need EAP to do? It's not FSD.

 

[ℬête Noire]

AEB uses cameras and radar. So to sum it up EAP is using vision for EAP tasks which require vision such as lane keeping and AEB.
What else did you need EAP to do? It's not FSD.

His point is there is scant, if any, evidence that the cameras data is used to detect anything beyond the lane lines.

They're going to need be processing that camera data feed for object recognition to get much further, because of the limitations of the radar data stream.

 

[JeffK]

His point is there is scant, if any, evidence that the cameras data is used to detect anything beyond the lane lines.

They're going to need be processing that camera data feed for object recognition to get much further, because of the limitations of the radar data stream.

We just demonstrated that as a matter of fact there is data showing that it detects vehicles using vision. Even posted a picture.

 

[ℬête Noire]

We just demonstrated that as a matter of fact there is data showing that it detects vehicles using vision. Even posted a picture.

ROFL

 

[tinm]

Zero interest in FSD or EAP. Just give me tons of sensors and intelligently alert me now and then to potential dangers but I prefer to keep on doing the driving. While everyone else seems gladly willing to surrender their personal agency to Tesla, I think I'll hold on to mine.

And look: I know 99% of you disagree. I respect that and stipulate that, no disputes. Enjoy your FSD and EAP. Knock yourselves out. But please don't slam me with lots of "Disagree" votes to this post just because I have a different viewpoint. Thank you.

 

[JeffK]

Zero interest in FSD or EAP. Just give me tons of sensors and intelligently alert me now and then to potential dangers but I prefer to keep on doing the driving. While everyone else seems gladly willing to surrender their personal agency to Tesla, I think I'll hold on to mine.

And look: I know 99% of you disagree. I respect that and stipulate that, no disputes. Enjoy your FSD and EAP. Knock yourselves out. But please don't slam me with lots of "Disagree" votes to this post just because I have a different viewpoint. Thank you.

Many people feel the same way. One thing Tesla could do better is making blindspot detection more obvious for the driver.

The only problem with people feeling this way is that over 94% of all fatal accidents are due to human error. The goal would be to reduce that. Even without EAP and FSD you'll still get AEB which will try to compensate for the slow reaction time of a human. Not all human drivers are created equal and also sometimes "stuff" happens to the best of us.

If you're bored, this is hopefully where the industry will go in terms of safety:

 

[pilotSteve]

Many people feel the same way. One thing Tesla could do better is making blindspot detection more obvious for the driver.

Yes indeed! I gave Tesla feedback when adjacent lane traffic "icons" were finally displayed that this was NOT a safety improvement until vehicles in the blind spots (requires the side cameras be active and contributing to vehicle identification) are displayed as well.

 

[kbM3]

Question is worded weirdly.

You understand that the price is still $3,000 during vehicle purchase?

If the question is "Now that Tesla is increasing the post-purchase price difference for FSB from $1000 to $2000, will you be including FSD in your initial purchase?"

Very unlikely. The feature they roll out in August would have to be blow-away amazing by itself, because I still don't have confidence Tesla (or anyone on the planet) is anywhere near [Level 5] FSD.

It doesn't have to be level 5. Level 4 is plenty.

I don't think the current rendition of Tesla vehicles have sufficient hardware and compute capacity to go beyond L3.

I would be surprised even if they barely touched L3.

I am Tesla fanboi allright, but I am also educated in STEM with a computer science major and by profession. So I am not convinced.

I'm a software Dev and my instinct is that you're right. BUT, I make it a general policy to never bet against Elon. That is a quick way to join the legion of people that have done so and been proven completely wrong.

 

[ℬête Noire]

It doesn't have to be level 5. Level 4 is plenty.

Tesla ain't anywhere near that, either. In part because they are effectively targeting Lv 5 (see Musk's cloud talk comments about long distance remote summon across public roads)

 

[tinm]

Imagine: it's 2028. You're driving in your Tesla down a desolate, straight rural road at 105 mph, using full EAP and FSD. Cop, hiding behind a hedge, sees you, clocks you at 105, starts following, turns on siren and flashing lights. A camera on the front of his cop car's grille zooms in on your license plate, looks it up in the cloud, sends a command to a state-law-approved system that all car manufacturers must abide by that in turns sends a signal to your car. Suddenly a loud alarm sounds in your car, and a warning dialog appears on your screen: 'Law enforcement orders you to immediately safely pull this car over. If you do not comply the car will automatically do so on its own in 10 . . . 9 . . . 8 . . . 7 . . " etc.

Or imagine it's 2028 you're driving your Tesla in a busy city. Maybe you are a minority, just driving along perfectly legally. Suddenly a cop is behind you, sirens and flashing lights, and the same warning appears on your screen.

This is the price of EAP, FSD, and the cloud-based, always-connected world that Tesla is also leading us into. All Tesla's reasons may right now be perfectly sensible (OTA software updates, detecting problems in the car in real time, notifying law enforcement of your car's whereabouts if it's stolen, etc) but at some point when the whole world catches up, so will the laws on the books, to take advantage of cars that are controlled by computer and the internet.

I'm not sure I like that world.