Full Self Driving ability by end of the YEAR ?

[Starno]

So we have to make the distinction between ability and availability. We all know that for questions of safety, and regulation, Tesla won't release something if it isn't completely reliable.

From the video of Tesla we see in Palo Alto, the ability seems to be here. But I suspect it isn't yet reliable enough. It's possible they shot the video multiple times before they get a shot where the driver doesn't need to take control of the car.
---------------

 

[sandpiper]

There was a bit info floating around the internet, that talked about the making of that video. And reality did not match what the video implied.

I think that there is no chance that we see any legal self driving on the road within the next decade.
The problem is that the current systems lack contextual awareness and I can't see how any amount of neural network training will get around that to the point where the vehicle is safe to run free. I think, before it can be legal, you will need some sort of redundant system that ensure that the vehicle won't contact obstacles and other cars, follows traffic signaling, and is tolerant to adverse weather conditions.

At a minimum I expect that you will need both cameras an independent LIDAR system. And I expect that you will need some combination of:

- high precision GPS and high resolution mapping
- simple on-ground infrastructure (like short range radio that could signal intersections, traffic signals, etc..)
- continuous car to car communications.

I also expect that there will be a slow roll-out development of regions / roads that have the installed infrastructure where auto-drive is permitted on properly equipped vehicles. That would initially be mostly controlled access highways and increase to include the feeder roads as the systems prove themselves out and the infrastructure is installed.

I do expect that Tesla will get to the point where self driving is mostly possible with the current cars. But I don't see it ever being reliable enough that it will ever be turned on beyond the factory demos.

 

[Burnt Toast]

There was a bit info floating around the internet, that talked about the making of that video. And reality did not match what the video implied.

I think that there is no chance that we see any legal self driving on the road within the next decade.
The problem is that the current systems lack contextual awareness and I can't see how any amount of neural network training will get around that to the point where the vehicle is safe to run free. I think, before it can be legal, you will need some sort of redundant system that ensure that the vehicle won't contact obstacles and other cars, follows traffic signaling, and is tolerant to adverse weather conditions.

At a minimum I expect that you will need both cameras an independent LIDAR system. And I expect that you will need some combination of:

- high precision GPS and high resolution mapping
- simple on-ground infrastructure (like short range radio that could signal intersections, traffic signals, etc..)
- continuous car to car communications.

I also expect that there will be a slow roll-out development of regions / roads that have the installed infrastructure where auto-drive is permitted on properly equipped vehicles. That would initially be mostly controlled access highways and increase to include the feeder roads as the systems prove themselves out and the infrastructure is installed.

I do expect that Tesla will get to the point where self driving is mostly possible with the current cars. But I don't see it ever being reliable enough that it will ever be turned on beyond the factory demos.

No, it will work, but it wont be level 5. It will require a human to take over the wheel at a moments notice. The thing I'm most concerned about oddly enough are traffic lights. You need 100% reliability or FSD is a dream.

 

[BestRadar]

FSD is a fantasy that I am many others bought into seeing their FSD video that they still tout today. I love the description on how it says that our Self Driving Tesla can be used for car sharing and ride hailing for friends and family. There is zero chance from what I have seen from Tesla lately that this is a reality. The car can un-park itself and pick me up at the door. Cant wait to see this happen. They gave us summon and all we can do is make the car go back and forward remotely.

 

[MSullivan]

No way, I don't even think we will have the new browser kernel by the end of the year.
Elon is horrible at managing expectations.
Example, announced in April the semi will be unveiled in September. Why?
Apple sends out invites to journalist a week out. Why not do the same?
Elon should stop pre-announcing anything.

 

[McRat]

Any particular year you have in mind?

 

[sandpiper]

No, it will work, but it wont be level 5. It will require a human to take over the wheel at a moments notice. The thing I'm most concerned about oddly enough are traffic lights. You need 100% reliability or FSD is a dream.

So I don't think we disagree. FSD, to me, means L5. And the video and Tesla rhetoric imply that as well. I do believe that it will generally work, but it won't be reliable enough that you can turn it on and expect it to run around the city while you take a snooze. You'll have to be sitting there, paying attention, maybe hands-off, maybe not.

To have a truly usable and safe system requires redundancy - hence my list of requirements above. You should be able to have a big splat of seagull turd completely obscure your front camera, and the system will continue on safely, perhaps pulling over to ask somebody to clear the camera. The current hardware will simply not do that.

 

[Starno]

So I don't think we disagree. FSD, to me, means L5. And the video and Tesla rhetoric imply that as well. I do believe that it will generally work, but it won't be reliable enough that you can turn it on and expect it to run around the city while you take a snooze. You'll have to be sitting there, paying attention, maybe hands-off, maybe not.

To have a truly usable and safe system requires redundancy - hence my list of requirements above. You should be able to have a big splat of seagull turd completely obscure your front camera, and the system will continue on safely, perhaps pulling over to ask somebody to clear the camera. The current hardware will simply not do that.

I don't think it will be more than 2 years after what Elon predicted, meaning maximum 2020 I think.
Elon might over promise often but not at the point of missing a target from 10 years...



Elon Musk: We Will Have "Full Self-Driving Capabilities" in 3 to 6 Months

So yes, it's really probable that Elon misses the target of having it in 3-6 months. But do you really think he would dare making such assumption if he would think it is impossible for the next 10 years ?

 

[NerdUno]

Tesla may have outsmarted themselves on the FSD option. If it turns out that the cars need more hardware, Tesla will have no choice but to retrofit the existing vehicles before releasing any sort of enhanced hardware vehicle. Remember, it's a contract.

 

[ancreserve]

Previous Episodes. Link to self driving(almost Leaf

 

[stopcrazypp]

I don't think it will be more than 2 years after what Elon predicted, meaning maximum 2020 I think.
Elon might over promise often but not at the point of missing a target from 10 years...



Elon Musk: We Will Have "Full Self-Driving Capabilities" in 3 to 6 Months

So yes, it's really probable that Elon misses the target of having it in 3-6 months. But do you really think he would dare making such assumption if he would think it is impossible for the next 10 years ?

Bad quoting in that article that removed the context.

This is the actual chain:
@tsrandall
At what point will "Full Self-Driving Capability" features noticeably depart from "Enhanced Autopilot" features?

@elonmusk
3 months maybe, 6 months definitely
Elon Musk on Twitter

So what Elon means is that in 3-6 months there will be a noticeably different feature for those that opted for the FSD vs those that just opted for the EAP. It does not mean that they will be releasing FSD (level 5) driving at this point. EAP uses 4 cameras, FSD uses 8 cameras, so there are plenty of other features they can release for FSD (not related to level 5 driving) that won't be available in EAP.

The only time-line related to level 5 is that there will be a cross country demo by the end of the year. But there is no promised timeline for when level 5 driving will be released to the public.

 

[davidc18]

I do not see how Tesla can avoid insuring the vehicles if they are ever going to get to L5 capability.

Is any insurance actually going to accept liability for a Tesla driving itself?

I just do not see that happening until/unless Tesla offers to accept full responsibility for what ever happens while the vehicle is driving itself.

 

[Tam]

...no promised timeline...

In 2014, Elon guessed that it would take 10 years (2024).

In 2015, he guessed that it would roughly be in 5 years (2020).

In 2016, he guessed that it would be in 24-36 months (2018-2019).

In July, 2016, Business Insider said:

"Another bold promise Musk has made for Tesla is that its cars will be autonomous by 2018. And it seems to be on track to make that happen."

 

[stopcrazypp]

In 2014, Elon guessed that it would take 10 years (2024).

In 2015, he guessed that it would roughly be in 5 years (2020).

In 2016, he guessed that it would be in 24-36 months (2018-2019).

In July, 2016, Business Insider said:

"Another bold promise Musk has made for Tesla is that its cars will be autonomous by 2018. And it seems to be on track to make that happen."

That is just general guesses Elon made about when they will have it, not a concrete promise for FSD level 5 release (in fact FSD suite didn't even exist when Elon gave those timelines).

Linking to the actual quote for the supposed 2018 promise. This is what Elon said in September 2015:
"The Tesla that is currently in production has the ability to do automatic steering autopilot on highway. That is currently being beta tested and will go into wide release early next month. So, we are probably only a month away from having autonomous driving at least for highways and for relatively simple roads," Musk said. "My guess for when we will have full autonomy is approximately three years."
Elon Musk says Tesla’s fully autonomous cars will hit the road in 3 years

As you can see, it is just Elon's vague guess of when it is possible, not anywhere close to a concrete promise. From your own paraphrasing, apparently in the same year he said 5 years also, which shows how vague this is.

 

[Tam]

...vague...

I agree that the timeline is not official and not concrete, just vague guesses but I think it does seem to say that with exponential progress, Tesla thinks it can achieve autonomous goal much quicker than other companies' guideline such as for Ford's goal for 2021...

 

[stopcrazypp]

I agree that the timeline is not official and not concrete, just vague guesses but I think it does seem to say that with exponential progress, Tesla thinks it can achieve autonomous goal much quicker than other companies' guideline such as for Ford's goal for 2021...

But this thread is about a concrete release of FSD/level 5, not some vague general talk about when autonomous cars might be possible. There are plenty of industry predictions for that (Elon's guesses is just part of it) and they are largely on theory.

But I believe this thread is more concerned about the practical aspect.

 

[Tam]

...practical aspect...

It's hard to get any reliable estimated timeline until I can see a success on current AP2 transition which it is trying to catch up with AP1.

A cross country driverless demo in December will be a good indication of how well the progress is.

But of course it could be a well planned rehearsed setup just to pass the predicted routes.

Even if that's the case, not all demos can pass the grade such as the last 2017 CES self-parking demo on stage from Faraday Future (OK in parking lot though).

I think an objective way to measure the progress is California DMV Disengaement Report (how much human intervention required on the roads).

Without objective data (how many times human have to intervene, how many collisions...) the timeline is just speculation.

 

[googlepeakoil]

No way, I don't even think we will have the new browser kernel by the end of the year.
Elon is horrible at managing expectations.
Example, announced in April the semi will be unveiled in September. Why?
Apple sends out invites to journalist a week out. Why not do the same?
Elon should stop pre-announcing anything.

I've heard it said before - making 1-5 cars is really easy compared to having an assembly line making them. He's keeping his design and early-prod team busy doing something. So I can believe they'll have a first truck built - but that's still many years away.
Why he does it? Hype and SP influence. Oh, and he can't help himself

 

[SlyWombat]

L5 Driving does not have to be foolproof, you getting in the back seat of an Uber has a chance of running a red light and crashing (with 10 million car accidents per year). You don't need LIDAR, your head does not have this and you can drive a car with just vision, so the hardware is fine. Now the advances with deep learning have accelerated capabilities of the software, it was a breakthrough. That said, my vote is no way by end of year due to regulatory reasons, but an indepth demo of capabilities will be shown by end of year.

 

[SlyWombat]

BTW, 10 million car accidents was not Uber, but a 2011 statistic in general How Many Times Will You Crash Your Car?

Tesla just has to be "safer" by an order of magnitude and they win.