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

2 years: Musk's new timeline for door-to-door full autonomy

This site may earn commission on affiliate links.
Software done meaning it meets the technical requirements of being Level 5 (can drive without any involvement of the driver or even a driver at all). Technically a level 5 car should never require disengagement. Level 4 however, may still require disengagement, although it should fail safely even if driver does not respond. The initial aim was level 4 back then, but I believe the industry has moved on to target level 5.

That's not the definition of Level 5. As Level 4 can drive without any involvement of the driver or even a driver at all. The difference between Level 4 and Level 5 is operational design domain (ODD) like geofencing.

Level 5 on the other hand can drive anywhere a human can from bright and sunny roads, to 10 inches of snow blizzard.


to dirt road and un-marked desert and off roads, to muddy and upstreams, to driving ultra dense traffic like india.
 
  • Disagree
Reactions: shonline
@croman - my personal belief is that bladerskb has psychological issues. I'm not being snarky or snippy. He isn't a typical troll - but he does fit the profile of a certain kind of maladjusted forumite. There's one on the biggest home theater site - avsforums - who has has made it his personal mission to "educate" or "correct" owners and potential buyers of Sony's most expensive home theater projectors ($15-$50K). He can't afford one himself but he is a geek and he does have some technical knowledge. He has been hanging out there for several years and will spend no end of time writing detailed critiques of Sony's products.

Instead this guy prefers any other brand, especially JVC. Just like @Bladerskb he gives off that whiff of a guy who shows up to a Ferrari club meet without a Ferrari and wants to argue with the Ferrari owners in great detail why their cars aren't "all that."

Just like @Bladerskb this anti-sony guy can't seem to talk about anything else and has a really odd "fixation" about trying to prove to Sony owners that their products are flawed, that Sony is flawed as a company etc.

Bladerskb clearly gets emotional about this topic - witness the way he keeps saying "you people" and going ad hominem in his thread replies. Something is wrong with this guy.
 
Pretty stupid and extremely non exciting.


I will forever drive my car. I want my Tesla with no useless safety features and no autopilot. CPO 13-14 it is....

Oh and if you want your car to drive itself then you might as well have your arms and legs chopped off since you wont be using them to drive. And might as well just sit in a wheelchair for the rest of your life since humans nowadays are too lazy and want technology to do everything for them. ROFL.

Our brains have gotten larger but our intelligence continues to shrink.... so sad.

With all do respect, I think you are missing the purpose that many of us want a viable autonomous vehicle because you don't yet own one.

It is not so I can put my kids in the car and have them dropped off at school, or I want to do anything but drive on my way to work, or I want my Model X to be a taxi when I am not using it, etc

Us current owners know about the safety factor added when AP is activated, and theoretically in time it will be more safe, possibly the safest when the human factor is eliminated.

You say you don't want any of that, and the certainly is your perspective, but I am pretty sure you will never use it, especially once it is fully developed and approved.

Show me your confirmed order with EAP not selected, and maybe I will believe you are true to your words, and you will be one of the very few to purchase a current Tesla that way.
 
@croman - my personal belief is that bladerskb has psychological issues. I'm not being snarky or snippy. He isn't a typical troll - but he does fit the profile of a certain kind of maladjusted forumite.
Bladerskb clearly gets emotional about this topic - witness the way he keeps saying "you people" and going ad hominem in his thread replies. Something is wrong with this guy.

Pretty bold psychological assessment, but I don't understand his last post, so either I am ignorant, or you may be correct.

Its not driving in India. Its praying that Lord Ganesha keeps you alive and making it to your destination.

I really hope you are being funny and not actually praying to Lord Ganesha, but I support your right to do both.

And you know this - how? The issue will come down to statistics - whether or not camera-based self driving will be demonstrated to be statistically safer than human driving. AP1 got a 40% crash reduction. So far the stats are in Tesla's favor.

If the issue came down to sensors the problem would be solved already. Anyone can stack up sensors - the hard part is the software. What Tesla is doing is arguably a more challenging and ambitious approach than those using lidar to map out urban centers. Tesla is attempting to have neural networks learn to interpret visual scenes accurately enough to drive as safe/safer than a human being without precision cm maps.

Not sure I fully agree, as I will agree to AP1 adding safety, but there is a huge difference between a driver assist process and autonomous driving.

My perspective is that most other manufacturers are not just relying upon image interpretation, which relies upon accurate visualization of the environment which can be impaired by many factors, including weather, Having alternative sensory input no doubt will demand more advanced processing, but utilizing multiple input types may prove to be more successful.

We can predict all day long, but in the end what matters is what gets approved for public roads, and how much confidence do you have in a government entity to make that assessment, free of subjective bias.

Even more stringent may be the insurance companies, since they now have a surcharge just be being a ride-share driver, imaging the premiums for those with autonomous vehicles. Now they are the experts in stats.


Shattered dreams? Come on dude - this is so much hyperbole and drama. I have an AP1 car - I have no "dreams" to be "shattered." Nobody expected it to do FSD. Use logic here - AP2 has 360 visual coverage and a replaceable board.

There is no particular reason - other than invented worries and hand wringing of the hyperventilating nerds around here - to think that regulators won't give approval under some conditions if it is demonstrated to be safe. It may take a while but the sensors are there. There is no data on this thread - just a bunch of geeks wringing their hands.

If AP2 ain't tha ticket I'll sell mine and buy AP3. Who frickin' cares? This is the price you pay to be on the bleeding edge. HOWEVER - while there are/were obvious reasons why AP1 would never go very far there are no reasons for AP2 except, once again, data-free rants and speculation of forum nerds.

Tesla owns the software now, they built in a replaceable brain, and they have 360 degree visual coverage. At this point it is a question whether software and deep learning can get to safer-than-human in-at-least-some-conditions driving.

Agree to disagree, but my dreams were shattered when I realized my AP1 resale value just tanked by over a third of the purchase cost, once no hardware upgrading was offered, just 2 months after delivery of my X, plus the reality that no further enhancements to AP1 were forthcoming, never expecting FSD.

And you are way off base about the approval process, for as stated above, it's not just the govt and municipality's approval, but that of the insurance carriers.

"Who frickin' cares?"... those of us who invested high dollars for an expensive high-tech automobile do, clearly not you. Have you so much money that upgrading vehicles is meaningless to you?

Your patronizing and aggressive characterization of those of us thinkers who are debating these details indicate you truly are a hater, as does the number of your down ratings.

Why are you even on the forum other than to chastize and bully others??? Been of your meds lately?

I can wait to see your expected aggressive response.
 
  • Like
Reactions: mmd
Timeline: The future of driverless cars, from Audi to Volvo

“We’re going to end up with complete autonomy, and I think we will have complete autonomy in approximately two years,"
""When I say level 4, I mean level 4 autonomy with the probability of an accident is less than that of person," he says."
"Its a much easier problem than people think it is"

Elon Musk: Tesla is 2 years away from a fully self-driving car

That is the rational goal. The rest is fluffy marketing.
 
@croman - my personal belief is that bladerskb has psychological issues. I'm not being snarky or snippy. He isn't a typical troll - but he does fit the profile of a certain kind of maladjusted forumite. There's one on the biggest home theater site - avsforums - who has has made it his personal mission to "educate" or "correct" owners and potential buyers of Sony's most expensive home theater projectors ($15-$50K). He can't afford one himself but he is a geek and he does have some technical knowledge. He has been hanging out there for several years and will spend no end of time writing detailed critiques of Sony's products.

Instead this guy prefers any other brand, especially JVC. Just like @Bladerskb he gives off that whiff of a guy who shows up to a Ferrari club meet without a Ferrari and wants to argue with the Ferrari owners in great detail why their cars aren't "all that."

Just like @Bladerskb this anti-sony guy can't seem to talk about anything else and has a really odd "fixation" about trying to prove to Sony owners that their products are flawed, that Sony is flawed as a company etc.

Bladerskb clearly gets emotional about this topic - witness the way he keeps saying "you people" and going ad hominem in his thread replies. Something is wrong with this guy.

While I will not jump into this directly, I will say that this forum has many folks who have an agenda. Stuff from stock manipulation to Lord knows what. Many are non owners, not even want-to-be owners who tend to be negative and dismissive for whatever reason draws them here.

I tend to look for a signature showing car ownership (even though that can be faked) and then content of posts when I am deciding how much weight I'll give someone's thoughts or meandering...
 
  • Like
Reactions: spentan and essmd
While I will not jump into this directly, I will say that this forum has many folks who have an agenda. Stuff from stock manipulation to Lord knows what. Many are non owners, not even want-to-be owners who tend to be negative and dismissive for whatever reason draws them here.

I tend to look for a signature showing car ownership (even though that can be faked) and then content of posts when I am deciding how much weight I'll give someone's thoughts or meandering...

Sure. And to be fair, equally a lot - probably even more than on the negative side - TSLA investors who have a vested interest in Tesla the company on the positive end of the spectrum.

All biases can twist conversations and objectivity should be the goal, of course.
 
  • Like
Reactions: shonline
That's not the definition of Level 5. As Level 4 can drive without any involvement of the driver or even a driver at all. The difference between Level 4 and Level 5 is operational design domain (ODD) like geofencing.

Level 5 on the other hand can drive anywhere a human can from bright and sunny roads, to 10 inches of snow blizzard.


to dirt road and un-marked desert and off roads, to muddy and upstreams, to driving ultra dense traffic like india.
I'm just going with the SAE definition:
Level 4: the driving mode-specific performance by an automated driving system of all aspects of the dynamic driving task, even if a human driver does not respond appropriately to a request to intervene.
Level 5: the full-time performance by an automated driving system of all aspects of the dynamic driving task under all roadway and environmental conditions that can be managed by a human driver
https://www.sae.org/misc/pdfs/automated_driving.pdf

As you can see, Level 4 still has "request to intervene" (AKA disengagement), Level 5 never has "request to intervene". Yes, the level 4 car can still fail safely in environments it can't handle, but it requests the driver to take over first (if driver ignores, it stops safely). Level 5 will never have that need to ask, since if it fails, that means it's a situation a human can't handle anyways.
 
Last edited:
"Who frickin' cares?"... those of us who invested high dollars for an expensive high-tech automobile do, clearly not you. Have you so much money that upgrading vehicles is meaningless to you?

Concede your point. As for why I'm on the forums - I enjoy the discussion, I love tech, AI and robotics and I think if you look up my history you'll see far more likes than disagrees. But like everyone I have things that set me off - mine personally are: 1) people wringing their hands about theoretical obstacles in the future which loom large in their worries. 2) Forumites who lecture others about safety - I do get a guilty pleasure from making fun of these self-appointed nannies. 3) Pseudoscientific discussions involving people arguing over arbitrary autonomy-level distinctions, hypothetical future government regulations, and future-predicting-based-off-youtube-videos. This kind of discussion is a far cry from people who understand the details of machine learning having high level technical discussion - THAT would be discussion worth reading instead of poking fun at. I wish there were more experts here (I'm not one of them), I truly do.

I'll also concede I can be aggressive. Take it how you will.

Pretty bold psychological assessment, but I don't understand his last post, so either I am ignorant, or you may be correct.

Or - you are ignorant and I am also wrong. Or you're not ignorant and I'm wrong. Or, or, or...

Not sure I fully agree, as I will agree to AP1 adding safety, but there is a huge difference between a driver assist process and autonomous driving.

Agreed - and there is a huge difference in both the sensor suite AND the processing power of the computer between AP1 and AP2. 1 camera vs 8. 40X more computing power. A swappable board (I think this is key) that can instantly multiply the processing power should it turn out this board can't run enough teraflops to run a neural network deep enough to safely handle the most complex driving situations. And Tesla now owns its software - no chance it will lose the license and have to start over. Finally - Tesla has gone from zero to a very usable autopilot in 6 months. It is still using a fraction of the computing power and only 2 out of 8 cameras. All signs point to lift off. If you see evidence the other way - what is it?

My perspective is that most other manufacturers are not just relying upon image interpretation, which relies upon accurate visualization of the environment which can be impaired by many factors, including weather., Having alternative sensory input no doubt will demand more advanced processing, but utilizing multiple input types may prove to be more successful.

Having more sensory input will DOUBTLESS prove more successful eventually. The more you know the more you can predict. This is not the same as making the claim that more than vision is needed to drive as well or better than a human.

In this case however I think it's reasonable to look at the history of Musk's operations - which have always been cheap hardware + brains = world beating performance. He did it with SpaceX (read the history of that cowboy operation - kind of astounding - it's in his biography) and then with AP1 he produced the best performing autopilot available with far fewer sensors than others (see MBZ's car bristling with radar and cameras that couldn't perform nearly as well). So - in the hands of any other company I'd agree with you. But Musk has shown us more than once that software is what is key. And he has the track record so far to back up his philosophical position. If he fails then he fails - but I think the safer bet right now based on history is that he will succeed.

And YES - of course you are right that eventually Teslas will have more sensors. It isn't an either/or proposition. You could have acceptable safety Level 5 with "suite a" and increased safety with "suite b." Like in planes - pay more you get more redundancy, more engines, etc.

We can predict all day long, but in the end what matters is what gets approved for public roads, and how much confidence do you have in a government entity to make that assessment, free of subjective bias. Even more stringent may be the insurance companies, since they now have a surcharge just be being a ride-share driver, imaging the premiums for those with autonomous vehicles. Now they are the experts in stats.

Agree that carriers are experts in stats - therefore what, exactly? These carriers are not planning for the inevitable? AP1 reduces crashes significantly. My AP2 car makes fewer "potential crash" errors than my AP1 car does despite being 6 months old on brand new software using a fraction of its capabilities. You don't think the insurance companies understand this? You don't think they see where this is going?

Agreed - I have no confidence in government to make bias-free assessments. I do have confidence the government will approve Musk's camera-only system for at least some conditions.

But as someone pointed out in an article recently - this "government must approve" worry is a bit of a false worry. Lack of regulation didn't stop Musk from deploying Autopilot. Do you remember the moaning/groaning/crying/hand wringing here and in the press at the time? He did it anyway. Why do we think the government is an obstacle now more than then - considering stats are racking up showing semi-autonomous systems are reducing crashes?

Agree to disagree, but my dreams were shattered when I realized my AP1 resale value just tanked by over a third of the purchase cost

Okay, agree to disagree.
 
Concede your point. As for why I'm on the forums - I enjoy the discussion, I love tech, AI and robotics and I think if you look up my history you'll see far more likes than disagrees. But like everyone I have things that set me off - mine personally are: 1) people wringing their hands about theoretical obstacles in the future which loom large in their worries. 2) Forumites who lecture others about safety - I do get a guilty pleasure from making fun of these self-appointed nannies. 3) Pseudoscientific discussions involving people arguing over arbitrary autonomy-level distinctions, hypothetical future government regulations, and future-predicting-based-off-youtube-videos. This kind of discussion is a far cry from people who understand the details of machine learning having high level technical discussion - THAT would be discussion worth reading instead of poking fun at. I wish there were more experts here (I'm not one of them), I truly do.

I'll also concede I can be aggressive. Take it how you will.



Or - you are ignorant and I am also wrong. Or you're not ignorant and I'm wrong. Or, or, or...



Agreed - and there is a huge difference in both the sensor suite AND the processing power of the computer between AP1 and AP2. 1 camera vs 8. 40X more computing power. A swappable board (I think this is key) that can instantly multiply the processing power should it turn out this board can't run enough teraflops to run a neural network deep enough to safely handle the most complex driving situations. And Tesla now owns its software - no chance it will lose the license and have to start over. Finally - Tesla has gone from zero to a very usable autopilot in 6 months. It is still using a fraction of the computing power and only 2 out of 8 cameras. All signs point to lift off. If you see evidence the other way - what is it?



Having more sensory input will DOUBTLESS prove more successful eventually. The more you know the more you can predict. This is not the same as making the claim that more than vision is needed to drive as well or better than a human.

In this case however I think it's reasonable to look at the history of Musk's operations - which have always been cheap hardware + brains = world beating performance. He did it with SpaceX (read the history of that cowboy operation - kind of astounding - it's in his biography) and then with AP1 he produced the best performing autopilot available with far fewer sensors than others (see MBZ's car bristling with radar and cameras that couldn't perform nearly as well). So - in the hands of any other company I'd agree with you. But Musk has shown us more than once that software is what is key. And he has the track record so far to back up his philosophical position. If he fails then he fails - but I think the safer bet right now based on history is that he will succeed.

And YES - of course you are right that eventually Teslas will have more sensors. It isn't an either/or proposition. You could have acceptable safety Level 5 with "suite a" and increased safety with "suite b." Like in planes - pay more you get more redundancy, more engines, etc.



Agree that carriers are experts in stats - therefore what, exactly? These carriers are not planning for the inevitable? AP1 reduces crashes significantly. My AP2 car makes fewer "potential crash" errors than my AP1 car does despite being 6 months old on brand new software using a fraction of its capabilities. You don't think the insurance companies understand this? You don't think they see where this is going?

Agreed - I have no confidence in government to make bias-free assessments. I do have confidence the government will approve Musk's lcamera-only system for at least some conditions.

But as someone pointed out in an article recently - this "government must approve" worry is a bit of a false worry. Lack of regulation didn't stop Musk from deploying Autopilot. Do you remember the moaning/groaning/crying/hand wringing here and in the press at the time? He did it anyway. Why do we think the government is an obstacle now more than then - considering stats are racking up showing semi-autonomous systems are reducing crashes?



Okay, agree to disagree.


Ok, so your are more reasonable and thoughtful than you came in that earlier post.

Regarding the insurance carriers, their role will be adjusting the premiums for cars used autonomously, I can'y even imagine how they will react once the feature is available. Perhaps in time the stats will show an increase in safety, but there is now way they will extrapolate the current AP data and the presumed increase of safety to determine the risk... it's truly apples and oranges due to presence of a focused driver.
There absolutely be approval required prior to public use of autonomously-driven vehicle, by the NTSB. They also know that there is a huge difference between a driver-assist product and a driverless vehicle. Whatever that article said, it was just one persons perspective. Think of the NTSB like a FDA for a new drug, it will not just be released for purchase until vetted and approved.

In fact, here is their perspective:
https://one.nhtsa.gov/nhtsa/av/pdf/Federal_Automated_Vehicles_Policy.pdf

I also share most of your forum intolerances, but it is a community, and it takes all kinds to make a village, including a few idiots, trolls, etc....

I was under the impression that this forum was moderated to avoid bad behavior, but with what I have read, I doubt it.