Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Full Self-Driving - feels like a long way off to me...

This site may earn commission on affiliate links.
It's situations like this that reminded me how far we are away from having full self driving cars - able to operate without any humans behind the steering wheel, monitoring operation. Safe driving takes much more than doing lane keeping and looking for turns/entry/exit ramps.

That said, I do expect Tesla to continue improving the AP software, operating safer (under most conditions) in more situations, and requiring fewer driver interventions, especially on limited access, multilane, highways which are relatively simple, compared to side/urban streets.

Totally agree with you bob_p... I don’t think I can make many drives that don’t have any edge conditions and I forsee programming a system to reliably deal with them very, very difficult.

I too expect things to get better as time goes by, but thinking fully autonomous driving is just around the corner as some do (just look at some of the quotes I posted above) is ridiculous to me. Look where we are now... FSD is not happening by the time Tesla’s next full redesign (if it is 2021) rolls out. Sorry to say to some of these folks, but we’re going to have human controls for a long, long time.
 
  • Like
Reactions: Alex362
Honestly, I think some people underestimate AI in self-driving cars. Everything you mention are all things that a self-driving car will be able to handle with the proper code. They are doable. I am not saying it will be easy to do but they are definitely doable. After all, you learned how to do these things correctly, so why can't a computer also learn them? An AI can learn them too, it's just a matter of teaching it, just like you were taught.

I have to disagree with you a little bit here... human learning and machine learning are not apples to apples. It’s more like apples to oranges.
 
FSD is a long way off and when it comes, why would I need a performance car like a Tesla? Autonomous cars will be speed regulated by legislation so why not just get a large battery car with a weak motor and ultra luxurious interior? That doesn't fit the Tesla MO. By the time FSD gets here, charging networks will be everywhere so their long distance competitive advantage will be gone. I don't understand why Tesla keeps pushing this feature when it doesn't fit their "we don't make slow cars" mantra. I wish they would focus on the driver while who will be doing the operating for the foreseeable future.
 
I have to disagree with you a little bit here... human learning and machine learning are not apples to apples. It’s more like apples to oranges.

Obviously, they are not exactly the same. However, machine learning can be quite powerful nonetheless. Machines actually are better than humans in terms of the amount of stored data and the speed of processing data that they are capable of.

"It is proven that machine based knowledge far exceeds the capacity of the human brain as far as memorizing knowledge, understanding and comprehending are concerned" What is the Difference Between Machine Learning and Human Learning?

Those are two things that should make computers good at self-driving once they get enough data. I mean, if a computer is able to store vast amounts of data and is able to process that data incredibly fast, even better than a human, I don't see why it wouldn't be able to self-drive, if given the right algorithms.
 
I mean, if a computer is able to store vast amounts of data and is able to process that data incredibly fast, even better than a human, I don't see why it wouldn't be able to self-drive, if given the right algorithms.

Hey, I hope you’re right but I just don’t see it. We have made leaps and bounds with AI, but having a vast amount of knowledge doesn’t mean you can apply it properly. I can memorize a bunch of facts, but if I don’t understand the meaning behind them it doesn’t do me much good other than being good at trivia contests. It’s like my mom trying to use her universal remote... she writes down every button press to switch between inputs and functions. If say, the TV doesn’t power on right away she’s stuck. I keep trying to teach her not to memorize the steps but to understand what each button does. Another example was DVD’s... all the menus are different... she would literally write down press left arrow, press okay, etc. but when the next DVD’s menu layout was different she was stuck.

Knowing a bunch of info is not power, knowing how to use, understand and apply it is... and that is hard to teach to a machine... and my mom. :)
 
Last edited:
  • Funny
Reactions: Inside
FSD is a long way off and when it comes, why would I need a performance car like a Tesla? Autonomous cars will be speed regulated by legislation so why not just get a large battery car with a weak motor and ultra luxurious interior? That doesn't fit the Tesla MO. By the time FSD gets here, charging networks will be everywhere so their long distance competitive advantage will be gone. I don't understand why Tesla keeps pushing this feature when it doesn't fit their "we don't make slow cars" mantra. I wish they would focus on the driver while who will be doing the operating for the foreseeable future.

Tesla is pushing FSD because a) it will happen sooner than people think and b) there is a lot of money to be gained from FSD.
 
  • Like
Reactions: Cowbell
Tesla is pushing FSD because a) it will happen sooner than people think and b) there is a lot of money to be gained from FSD.

A) I hope you’re right, but I’m skeptical. Look where the current autopilot is now... do you really think within 5, 10, 15 years we’re going to have cars with no human controls? Look at the progress so far... sure we’ve come a long way and technology growth is usually exponential, but we’ve got a long way to go.

B) 100% agree with you there.
 
I just want to let everyone know I’m not trying to be antagonistic or excessively argumentative... I want Tesla to succeed and I think it would be amazing if we had a FSD breakthrough.

I seem to remember we had somewhat of a breakthrough with machines understanding voice. I feel like we went from Dragon Naturally speaking back in the day where you had to train the software to understand only your voice to where we now have devices that can understand almost anyone right out of the box (unless you have a strong Irish accent, LOL)... and they can also adjust for sentence context. Pretty awesome. Hopefully we see some leaps like that with regards to FSD in the near future... only time will tell.
 
  • Like
Reactions: diplomat33
A) I hope you’re right, but I’m skeptical. Look where the current autopilot is now... do you really think within 5, 10, 15 years we’re going to have cars with no human controls? Look at the progress so far? Sure we’ve come a long way and technology growth is usually exponential, but we’ve got a long way to go.

Your skepticism is totally understandable. I think it is hard for people, myself included, to understand how AI works because we don't really have direct experience of it. I mean, we have direct experience with how our brains work. We know what they can do. And we know our brains can drive a car because we do it everyday. But we don't have any real experience with machine AI. They are an unknown to us. So can this "unknown" really do what I know my brain can do? It makes sense to be skeptical especially since we know how difficult driving can be but also how easy it is for humans. We don't need to think about driving. For example, we instantly see a parked car and know how to avoid it. It is almost a reflex. But teaching a machine to do the same is difficult. The machine has to identify the parked car correctly, know the path of the car, calculate if a collision will occur, identify other potential collisions, figure out a correct path to avoid all collisions, etc, which we humans do all in an instant without thinking about it. People also get stuck on the "edge cases" and think "this one situation is super hard for a computer and until it can handle all these difficult situations, it can never be truly L5 autonomous." True, but we need to take things one step at a time. Self-driving cars will improve over time. They don't need to jump straight to L5 right away. Just look at Waymo. No, their cars are not L5 autonomous yet but they have achieved pretty good L4 autonomous cars that can operate in certain geofenced areas. One step at a time!

And I tend to think that AP is pretty good for what it does now. It will get better over time. Add the ability to see traffic lights and traffic signs, add the ability to handle merging traffic, etc... and AP will get even better. No, it won't be able to handle every edge case right away, but it will improve over time and become more and more capable of self-driving as time goes on.

But in 15 years, we will definitely see self-driving cars with no controls at all. As you even say, the tech growth is exponential. It is not just moving faster but the rate of growth is also increasing faster and faster. Just look at a classic log growth:

1
10
100
1,000
10,000
100,000
1,000,000

The first generation was only an increase of 9 but in just 6 generations, it went from 1 to 1 million!!

FSD may take 10 years for the first step, 5 years for the next step, 1 year for the next step and 1 month for the last step. it looks slow at first but in the end, it comes very quickly.
 
Last edited:
  • Like
Reactions: Theflash95
FSD may take 10 years for the first step, 5 years for the next step, 1 year for the next step and 1 month for the last step. it looks slow at first but in the end, it comes very quickly.

Personally I’m not in a hurry, so whatever the timeline ends up being I’m okay with it. I just hate to see all these folks thinking it’s right around the corner and getting impatient that things aren’t improving fast enough. It’s a complicated task to program/engineer/what have you... people need to be patient. I’ll keep my fingers crossed that things move forward faster than I expect.
 
  • Like
Reactions: diplomat33
Unfortunately it's a fallacy to think that all technology will have an exponential improvement curve. Take a look at the OS on your laptop. It's probably running Windows 10. Sure, it has thousands of times more RAM, CPU cycles, and storage than say, 20 years ago, but ultimately Windows as an experience is still pretty horrible. The OS hasn't improved all that much. The analogy here is saying that the sensors and CPU/GPU may improve exponentially, but the core functional may only improve at a small incremental and linear rate.

Also I think it's fair to say that FSD is going to have a strong dependence on the network effort. Meaning its effectiveness is going to be a strong function of how many other FSD vehicles are on the road. The more FSD vehicles, the better the system becomes. The break through will start to happen when FSD become the majority, slowly forcing non-autonomous out. That is realistically around 20+ years away IMO.
 
Hey, I hope you’re right but I just don’t see it. We have made leaps and bounds with AI, but having a vast amount of knowledge doesn’t mean you can apply it properly. I can memorize a bunch of facts, but if I don’t understand the meaning behind them it doesn’t do me much good other than being good at trivia contests. It’s like my mom trying to use her universal remote... she writes down every button press to switch between inputs and functions. If say, the TV doesn’t power on right away she’s stuck. I keep trying to teach her not to memorize the steps but to understand what each button does. Another example was DVD’s... all the menus are different... she would literally write down press left arrow, press okay, etc. but when the next DVD’s menu layout was different she was stuck.

Knowing a bunch of info is not power, knowing how to use, understand and apply it is... and that is hard to teach to a machine... and my mom. :)
And yet your mom can drive a car? :)
 
  • Like
Reactions: tranzndance
Unfortunately it's a fallacy to think that all technology will have an exponential improvement curve. Take a look at the OS on your laptop. It's probably running Windows 10. Sure, it has thousands of times more RAM, CPU cycles, and storage than say, 20 years ago, but ultimately Windows as an experience is still pretty horrible. The OS hasn't improved all that much. The analogy here is saying that the sensors and CPU/GPU may improve exponentially, but the core functional may only improve at a small incremental and linear rate.

Mac/Apple fan? Just kidding. I agree with you... I hope there is some breakthrough that happens, but I feel like it’s going to be a long road (no pun intended) to get to FSD. Again... I would love to be proven wrong and see it happen sooner than I expect. I’m skeptically optimistic... is that a thing?
 
Welcome to the club. On FSD timelines, I always thought Musk was too optimistic, or as the shorts would say lying.

I believe FSD is possible only if:

- you make some key changes to our road infrastructure - from small roads to highways - that aids FSD. Like embedded sensors on the road that aids the car to make key decisions at certain junctures.

- have a common protocol that cars can share and send some basic information to each other.

We are far far off from that. I think we will have the first probe landing in Mars from SpaceX before we have any semblance of FSD for Joe average.
Don't have to be short to think that he was lying. He knew that the car wasn't going to drive from NY to LA without the driver lifting a finger. Anyone who believed that is a fool. So is he a fool or a liar? I'd buy buckets of food from Jim Bakker before I'd buy FSD from Musk. At least you are receiving a product as-advertised on-time from the ex-con.

Is it so hard to set a reasonable timeline and goals? You have a fantastic product, just sell it, you don't need the extra stuff. Take FSD completely off the table and you will still have 400,000 Model 3 reservations. So why do this? It is only going to lead to lawsuits and bad press. How about focusing on things that really matter to every owner... Quality Control and service.
///rant
 
And yet your mom can drive a car? :)

Yup... technology just kicks her butt. Trying to explain Bluetooth and what she needs to do (change the radio input to Bluetooth) if she wants to listen to her phone’s driving directions is exhausting. For years she drove a manual transmission car and was reluctant to go to an automatic. She gets a pass though, she raised me to be a decent human and she’s pretty awesome despite her Luddite tendencies.
 
  • Love
  • Like
Reactions: d21mike and Matias
I'm extremely doubtful that the current AI methodology is ever going to get to FSD because it's really aimed at improved pattern recognition rather than what's been dubbed "Artificial General Intelligence", that is, what most people think of as "intelligence". FSD seems to be to be in the realm of AGI, not pattern recognition.

Progress in AGI has been very slow to say the least.
 
Part of the issue too is that people have different expectations for what constitutes true FSD. For many, they will only accept FSD when it is perfect L5 autonomy that handles every single road in the entire world and every single driving situation always perfectly. If that is your standard, then yeah, you are going to find a lot of reasons to be skeptical. Personally, I define FSD as L4 autonomy.
 
Part of the issue too is that people have different expectations for what constitutes true FSD. For many, they will only accept FSD when it is perfect L5 autonomy that handles every single road in the entire world and every single driving situation always perfectly. If that is your standard, then yeah, you are going to find a lot of reasons to be skeptical. Personally, I define FSD as L4 autonomy.
In this case it is defined:
Press a button and your car will drive across the country to come to you.
Your car will go out and rideshare when it is called upon. (tesla network)
Your car will drop you off, then go and find a parking spot.

What level do you consider a car going from NY to LA without a driver? Because this is what is promised.
 
  • Love
Reactions: Matias
In this case it is defined:
Press a button and your car will drive across the country to come to you.
Your car will go out and rideshare when it is called upon. (tesla network)
Your car will drop you off, then go and find a parking spot.

What level do you consider a car going from NY to LA without a driver? Because this is what is promised.

That could all be L4 autonomy.
 
I'm extremely doubtful that the current AI methodology is ever going to get to FSD because it's really aimed at improved pattern recognition rather than what's been dubbed "Artificial General Intelligence", that is, what most people think of as "intelligence". FSD seems to be to be in the realm of AGI, not pattern recognition.

Progress in AGI has been very slow to say the least.
I'm a full self driving skeptic but totally disagree with you. AGI is a quantum leap past full self driving. Don't confuse the two. FSD requires ANI, not AGI.