Predict When will Tesla release Fully Self Driving to cars? 2018/2019/2020...

[Bladerskb]

I ran into this today and i thought about how many times bird poop has fallen on the side pillars of my BMW 3 series constantly and how a tesla at a stop light would handle it. Turns out to be a big problem on the google cars.

So my question to you is this, with meeting AP1 parity taking up to 9 months and regardless of regulation, when do you think Tesla will have/release Fully Self Driving software update ready (L4/L5) that is at or better than humans and why?

Like..

A) First Half of 2018
B) Second Half of 2018
C) First Half of 2019
D) Second Half of 2019
E) First Half of 2020
F) Second Half of 2020


E) 2021 and Beyond...

 

[Tam]

...regulation..,

Some states are not ready yet but at least Michigan and Florida have laws for driverless cars already.

So at least, "regulation" will no longer be an excuse for the delay.

 

[Bladerskb]

I should also mention California is gonna vote on one (which includes actually selling FSD like MI & FL law allows) at the end of the year and Georgia just got their law aswell, also German.
so when do you think it will happen?

 

[xyeahtony]

Elon thinks by end of 2017, i'll go with that. AP1/AP2 parity is happening this summer and if the LA-NYC self-driving test is successful this year, i dont see why FSD can follow shortly thereafter.

(a lot of people are going to laugh at this, but whatever lol).

 

[Sir Guacamolaf]

Last night I was driving on AP1 on the freeway. A plastic bag flew in front of the car, and the car panicked, showed a red car ahead and started slamming the brakes.

FSD, where you can sit back and sleep is safely a few years away.

 

[Skotty]

I'm going to go with 2036.

However, highway only maybe as soon as 2020 / 2021.

 

[Mo City]

I think L4 commences late 2020 or early 2021 .

AP2 will improve in S-curve fashion mirroring the Model 3 ramp up. The explosion of AP2 vehicles on the road will allow Tesla to build a neural network that puts them substantially ahead of everyone else. Simultaneously, AI will devour the explosion of data and allow Tesla to completely bridge the gap with Waymo. By the end of 2019, Tesla will be at the point of ironing out the last, difficult 5%.

Incremental gains over the next 12-18 months will be followed by rapid advances in 2019/2020.

To start with, L4/L5 will apply only on limited routes and grow over time to apply nearly everywhere. Some roads won't be L5 for a very long time. For this reason, in 2022, the Tesla Network may not take you from Point A to Point B via same route you would drive.

 

[oktane]

5 years

 

[NikeWings]

So my question to you is this, with meeting AP1 parity taking up to 9 months and regardless of regulation, when do you think Tesla will have/release Fully Self Driving software update ready (L4/L5) that is at or better than humans and why.

If your FSD bench is "at or better than humans", then it needs to solve for the lowest common driver-denominator which is pretty low and requires extensive oddball, bizarre, unimaginable, and what-the-hell-are-you-doing scenario design and test. Bird poop, flying debris, and other brainless threats might be less difficult than simulating threats with a brain. I have no idea what degree of challenge this presents to development but it's got to be more complicated than consistently and reliably reading a painted line (not there yet), so I'm not expecting human grade FSD sooner than 2020 (although sold earlier in true Tesla style).

 

[AMPUP]

I'm still not convinced they won't need to make another hardware change, but there are plenty of other threads discussing that. I'll say 2019 with a refreshed model S coming with new hardware in late 2018.

 

[Bladerskb]

I'm still not convinced they won't need to make another hardware change, but there are plenty of other threads discussing that. I'll say 2019 with a refreshed model S coming with new hardware in late 2018.

refresh late 2018 with lidar perhaps?
isn't that abit too close together...less than 1 year of development with the new hardware?

 

[malcolm]

Well, having just read the pattern of bird poop on my car, I'm going to forecast December 17 to May 18.

The birds have spoken. Or possibly dysentery.

Hard to tell.

 

[Bladerskb]

I'm going to go with 2036.

However, highway only maybe as soon as 2020 / 2021.

I agree with highway in 2020/2021. But any particular reason why so far for urban driving?

If your FSD bench is "at or better than humans", then it needs to solve for the lowest common driver-denominator which is pretty low and requires extensive oddball, bizarre, unimaginable, and what-the-hell-are-you-doing scenario design and test. Bird poop, flying debris, and other brainless threats might be less difficult than simulating threats with a brain. I have no idea what degree of challenge this presents to development but it's got to be more complicated than consistently and reliably reading a painted line (not there yet), so I'm not expecting human grade FSD sooner than 2020 (although sold earlier in true Tesla style).

i agree and think thats what alot of people are not taken into account. Dealing the the unpredictability on the road. You don't realize it till you actually have a mature FSD software running unto these unpredictable situations and you are left scratching yourself on how to solve it.

 

[Tam]

...German...

For clarification, unlike Michigan and Florida laws that allow cars without a licensed driver, Germany passed a law which is similar to Telsa Autopilot principle:

A human driver in the seat at all times,

there must be controls for driver to take over (Michigan allows cars with no steering wheel and no braking pedal),

However, with few improvements:

Human is only responsible during manual operations

Manufacturers are responsible if there's an accident during an autonomous session.

Drivers don't get ticketed for texting during an autonomous session

-------

It is hard to predict autonomous timeline until I can see something more concrete.

Autonomous dream has been around in the airline industry for a long time but it is still uncertain that the industry will be able to get rid of human pilots. It's understandable because flying is hard.

On the other hand, autonomous trains should be more simple to do due to its very strict geofencing routes but there is no prediction of when train industry will be able to replace train conductors.

There are already autonomous trains in private tracks such as between airport terminals and theme parks but they don't even have frontal obstacle sensors because the tracks are assumed free and clear at all times.

It is good to be enthusiastic but third parties still need to test them out to make sure autonomous cars can reliably avoid accidents.

 

[whitex]

The standard Tesla time answer was not one of the options: "soon", which of course means " shortly before eventually"

 

[Tangible]

We tolerate 30,000 deaths a year in the US in human-driven cars. If FSD is 100 times safer but not perfect, there will be a media outcry about the "killer robots" prowling the streets. Manufacturers will drown in civil suits and criminal complaints.

I don't think things will take off until we have cars that actively intercommunicate, operating on dedicated highway lanes. That's decades away.

Other areas of commerce will automate much faster, throwing multitudes of workers out of their jobs. The result will be an economy where almost everyone is an Uber driver, since welfare programs will have work requirements, so it will be drive or starve.

So, no FSD. We're headed for FCD, Fully Chauffeur Driven.

 

[lunitiks]

Germany passed a law which is similar to Telsa Autopilot principle:

A human driver in the seat at all times,

there must be controls for driver to take over (Michigan allows cars with no steering wheel and no braking pedal),

However, with few improvements:

Human is only responsible during manual operations

Manufacturers are responsible if there's an accident during an autonomous session.

This law will only be applicable to L3 (or higher), so it's not similar to Tesla Autopilot in my opinion

 

[cwerdna]

For reference, https://www.sae.org/misc/pdfs/automated_driving.pdf has a table of definitions of each level.

I'd say for level 5, I'm going with the option "2021 and Beyond..."

 

[lunitiks]

Notice at the beginning of that gif how they managed to capture the bird dropping as it hit the lidar!!
Man I'm curious how long they were filming.

Amazing

 

[HankLloydRight]

We tolerate 30,000 deaths a year in the US in human-driven cars. If FSD is 100 times safer but not perfect, there will be a media outcry about the "killer robots" prowling the streets. Manufacturers will drown in civil suits and criminal complaints.

I don't think things will take off until we have cars that actively intercommunicate, operating on dedicated highway lanes. That's decades away.

I completely agree with this. FSD like in that Tesla dazzle video is years or even a decade away.