-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
4 months left in 2020 to make a Tesla FSD on city streets
- Thread starter CertLive
- Start date
These are articles from NY times in 1903 hinting that manned flight would take at least 1 to 10 million years. Nine weeks after these articles were published the Wright brother took their first flight
Given how poor some of Tesla’s current driver assistance features are, it may seem like they’re far off. But the fact they’re still using antiquated code for things like park assist is in some ways encouraging. For whatever reason, they haven’t even tried to give quality updates to sub-features, so I’m optimistic a big leap could occur in the near future.
A new software rewrite that isn’t piecemeal code representing a patchworked evolution of features should help a lot. One of Tesla’s biggest advantages is the data it can feed its AI. With an increasing amount of miles driven, hopefully the rewrite will undergo growth and evolution at a rate faster than anything we’ve seen.
But in regards to value of FSD on a current purchase? That would still seem to be getting little in return for the prime life of the car. I’d rather spend 8k to get the driver assistance features currently found in other cars.
A new software rewrite that isn’t piecemeal code representing a patchworked evolution of features should help a lot. One of Tesla’s biggest advantages is the data it can feed its AI. With an increasing amount of miles driven, hopefully the rewrite will undergo growth and evolution at a rate faster than anything we’ve seen.
But in regards to value of FSD on a current purchase? That would still seem to be getting little in return for the prime life of the car. I’d rather spend 8k to get the driver assistance features currently found in other cars.
Atari2600
Active Member
Well, holding the lane when going into a turn too fast. Sorry for not being more clear. Even then it's probably too late to do anything at that point. This is for when you use AP on roads Tesla tells you to not use it on.
Ok thats it I have decided! I cant take this *sugar* no more! It is time to Send it back!
I shouldn't be spending my life in service centres fixing problems with a car I saved up for years! I should be getting paid to bug test AP not the other way around..... Its like some strange Tesla world where they think it is funny in a car travelling at 70mph it can just decide to turn off or phantom break. All because some HAL 9000 beta is overlaid onto my car whether I like it or not. What if I was to break a finger catching that steering wheel... my streaming career could be over in a second! Full refund time.
I shouldn't be spending my life in service centres fixing problems with a car I saved up for years! I should be getting paid to bug test AP not the other way around..... Its like some strange Tesla world where they think it is funny in a car travelling at 70mph it can just decide to turn off or phantom break. All because some HAL 9000 beta is overlaid onto my car whether I like it or not. What if I was to break a finger catching that steering wheel... my streaming career could be over in a second! Full refund time.
I agree no holding back here. The majority seem to feel this will not happen which is the sensible side to be on given the current software examples. Autopilot is not as good as other car offerings for doing the sole task of keeping the car in the lanes selected and at the speed required by reading actual road signs. Until of course this apparent software rewrite. Its still the electric car to beat in other area's but the car struggles on the one big thing its meant to be good at and sold at great cost to access autonomy. This post is also an eye opener for the much longer invested users in the technology. So far things have not moved hardly at all. I find it hard to wrap my head around a vision based system just like a human. If your eye is blocked by the sun how do you see the danger to take action? Its going to be an interesting few years to see who wins the race here. I would like to see a cheaper Tesla for all, I do not think the model 3 is cheap enough to really save this planet and make everyone turn. Need a model E to start S3XY"e"r. Small runabout that does 250miles+ for say 20K.
The whole point is that the car will do it better than a human to become safer. As of today that is 2030. Being optimistic 2025. Being on Elon time 2020 Christmas coming early.
Which EVs have better autostreer/lane-keep? I know it’s not the Audi or Porsche.
Imo you’re right about cost. Paying such a premium for acceleration and range makes more sense for personal car use than FSD.
Granted, it is/will be new tech, which is in the premium car space, but as FSD becomes more prolific, it’ll be tough to justify spending more on a car with a spartan interior.
A few points here. First, no other car on the market is able to automatically stop and go at stoplights, only Tesla. That shows real progress against the competition. Second, no other carmaker is using Tesla's approach of visual identification and neural networks for FSD and if they are, they don't have nearly as much real world training. Musk has also said that they are rewriting this code to switch from 2D to 3D identification. This feature is also clearly labeled as Beta software. YMMV, but I bought this car and this feature knowing that this stuff isn't a finished product and part of the fun of owning a Tesla is being in the edge and getting to test things out. Yeah it brakes when it shouldn't and some things don't work well. It's a brand new technology and it takes a while to get it right. If it was easy, it would have been done already. Timelines don't mean anything because until you hit the point where improvements happen exponentially, the progress is going to seem glacial for any technology. If you don't have the patience to be an early adopter, then don't buy the car or those features. It's really that simple.
Good thread.
I upgraded to extended range and AWD. IMO easily justifiable.
Since the day I bought this car I've said that FSD is a complete waste of money but that I'll buy it when I get my next Tesla. Nothing has changed that opinion. It seems like it has been getting everyone's hopes up to be better, so the frustration with it is increasing.
I upgraded to extended range and AWD. IMO easily justifiable.
Since the day I bought this car I've said that FSD is a complete waste of money but that I'll buy it when I get my next Tesla. Nothing has changed that opinion. It seems like it has been getting everyone's hopes up to be better, so the frustration with it is increasing.
The official blog post didn’t mention anything about needing new hardware, so I think it is valid that AP1 users thought this announcement applied to their current cars. I mean, this was the official communication and all. Probably went out via email to owners even. So, maybe Elon didn’t promise that feature, but Tesla did.
Summon Your Tesla from Your Phone
Last Fall, Tesla Version 7.0 software introduced a range of new Autopilot active safety and convenience features to give you more confidence behind the wheel, increase your safety on the road, and make highway driving more enjoyable. The release of Tesla Version 7.1 software continues our improvements to self-driving technology. This release expands Autopilot functionality and introduces the first iteration of Summon.
Using Summon, once you arrive home and exit Model S or Model X, you can prompt it to do the rest: open your garage door, enter your garage, park itself, and shut down. In the morning, you wake up, walk out the front door, and summon your car. It will open the garage door and come to greet you. More broadly, Summon also eliminates the burden of having to squeeze in and out of tight parking spots. During this Beta stage of Summon, we would like customers to become familiar with it on private property. Eventually, your Tesla will be able to drive anywhere across the country to meet you, charging itself along the way. It will sync with your calendar to know exactly when to arrive.
The release of Tesla Version 7.1 software is the next step toward developing fully autonomous driving capabilities and delivering them through over-the-air software updates, keeping our customers at the forefront of driving technology in the years ahead.
I had a Kia Eniro before the Model 3 and it was able to lane keep and move with traffic without any issues at all on the same trips and roads. So a very up to date comparison on capability compared to AP. I am not going near what it can do with updates but I am comparing my ownership from 2019 to now and as far as autonomy in this regard I believe KIA's LKA & Lane following assist more stable, does not suddenly break or turn off with different road markings. I even wonder if its FCA forward collision avoidance assist would have worked in that white truck across the motorway scenario that the model 3 did not see because of how it does it. Not sure about others yet to drive different systems. I felt so far my model 3 has not quite made it there yet.
Or is it because the other car manufacturers will not put people behind the wheel of beta autonomy? Maybe they do not feel its ready or safe enough yet to have their fleet doing FSD? I do not actually think if you drive other cars with road autonomy Tesla are in front with anything unless you like the potential hazardous bugs, like posted in my video earlier in this post. Traffic lights still require human say so?
Knightshade
Well-Known Member
FWIW I recently (couple months back anyway) drove a 2020 Kia with LKA.
Maybe it's different in the UK, but it was complete garbage compared to AP on a Tesla.
It couldn't be turned on at all under I think 40 mph, it ping-ponged noticeably in comparison, and it'd randomly turn off with no notice of the fact it did besides the color on dash indicator changing (nothing audible) and for no apparently reason- and did so numerous times in only a few hundred miles of driving.
Here's Bjorn BTW testing the Hyundai/KIA LKA system and finding it... not great...compared to Teslas system
His experience was even worse than mine honestly
Maybe it's different in the UK, but it was complete garbage compared to AP on a Tesla.
It couldn't be turned on at all under I think 40 mph, it ping-ponged noticeably in comparison, and it'd randomly turn off with no notice of the fact it did besides the color on dash indicator changing (nothing audible) and for no apparently reason- and did so numerous times in only a few hundred miles of driving.
Here's Bjorn BTW testing the Hyundai/KIA LKA system and finding it... not great...compared to Teslas system
His experience was even worse than mine honestly
Last edited:
There’s undoubtedly a different manufacturing approach.
In the AI/software world, I don’t think investing billions and years to bring technology to maturity in a controlled environment will work as well as it did last century.
That said, Tesla’s approach gives competitors a chance when it comes to other aspects of the car, particularly in combination with their poor service.
Tesla’s current offerings are lacking in things nearly every other car manufacturers do well, like parking, blind spot, HUD etc. But I have yet to find one that does the more “advanced” driving assistance better (I have not tired SuperCruise(
LKA should not be compared to Tesla's EAD. The lane keep assist systems just assist you, they don't drive the highway for you. They all bounce around because that's how they work. They sense a line on the right, they steer left and then the same thing on the other side.
Supercruise would be a better comparison.
So Musk has said that if self-driving is safer than regular driving it would be unethical for them to withhold that from people until it is 100% perfect. Anyone can disagree with that as it's totally a judgement call. Certainly there have been autopilot fatalities already and phantom braking is a serious, dangerous thing to be happening for sure.
The defense of that position though, is that long-term, Tesla's approach fundamentally requires collecting real world driving data. It's necessary to make the product better. Other manufacturers are using maps and lidar which may be better or worse at handling new situations they don't already have prior knowledge of. Tesla's argument has always been that its system will outperform these in the long run because it is capable of learning and making decisions on the fly. Again, it's completely valid to disagree with this approach until it is proven better but it's a completely different approach from what others are doing. This thread is hitting the intersection of these two ideas that Tesla both 1) needs people to use it's systems to make them better and 2) isn't very good right now but at some point could surpass everything else by a wide margin given enough time.
Knightshade
Well-Known Member
Teslas system does not learn on the fly though.
The only time behavior changes is with a firmware update from the mothership.
Most Tesla data collection is done passively via campaigns- regardless of what the human behind the wheel is doing, or what version of driver assists he paid for.... the bulk of it is just to collect pictures with the cameras to use to train the master NNs back at HQ.
I'm not the one who originally brought up LKA
That said- they advertise an "active" LKA on newer cars that is absolutely supposed to do more than just ping pong you around
It is hopeful and worrying at the same time. Like you, I'm hopeful their apparent neglect of the sub-features means something worthwhile is just over the horizon. However, it could also mean we have to live with the issues which would be a disappointment to many.
Good point and I agree. We'll have to wait and see what's in store.
Stating Tesla’s FSD system is “capable of learning and making decisions on the fly” isn’t the same as saying one’s car is “learning on the fly”.
Behavioral (thinking) changes from the learning of the FSD are limited to discrete updates, currently. Applications of 5G should change this.
Right, which needs people to use its system Ike the poster said.
Similar threads
