Where do you think Tesla's FSD will be 12 months from now (April 2020)?

[diplomat33]

I know we talk a lot about about whether Tesla can do FSD with some arguing yes and some arguing no. But let's examine a more pragmatic question. Putting aside the question of whether the cars will be L4, L5 or whatever, what do you think Tesla will achieve in the near term? Especially in light of the recent Autonomy Investor Day, I thought this would be a good topic.

Where do you think Tesla's FSD will be 12 months from now (April 2020)?

I do expect Tesla to make quick progress thanks to AP3's faster processing power.

Personally, in 12 months, I think FSD will be really good. I think NOA on the highway will be fully hands free by then. And I think "automatic city driving" will still be driver supervised but be able to handle most situations really well.

Thoughts?

 

[Daniel in SD]

I think you should define “good” and “ok”.
Miles between disengagements in downtown San Francisco?

 

[diplomat33]

I think you should define “good” and “ok”.
Miles between disengagements in downtown San Francisco?

You define it. What say you?

 

[Daniel in SD]

You define it. What say you?

I actually don’t know. I’m not sure I trust the disengagement numbers that get reported to the California DMV.
Miles between disengagements does seem like a good way to measure but you’ve got to somehow get a good sample of driving and not cheat by preemptively disengaging.
Cruise claims they’re way ahead of Waymo because they test in San Francisco instead of suburbia.
I’m going to guess that Tesla is at 10 miles between disengagements in urban driving in a year.

 

[diplomat33]

I actually don’t know. I’m not sure I trust the disengagement numbers that get reported to the California DMV.
Miles between disengagements does seem like a good way to measure but you’ve got to somehow get a good sample of driving and not cheat by preemptively disengaging.
Cruise claims they’re way ahead of Waymo because they test in San Francisco instead of suburbia.
I’m going to guess that Tesla is at 10 miles between disengagements in urban driving in a year.

The thing about miles per disengagement is that you need a big sample but also a diverse sample in order for the numbers to be meaningful. Too small a sample and your error bars will be too high. But if the driving is not diverse enough, then the numbers will only reflect the state of your self-driving in those specific situations.

I don't think Tesla releases the numbers of miles between autopilot disengagements. Considering how big the fleet is and how spread out it is across the US, the miles between AP disengagements would be interesting to see. And if they included the reasons for the AP disengagements, that would be fascinating. Maybe when Tesla does release feature complete FSD with driver supervision, they could publish the fleet's miles per disengagements?

 

[Daniel in SD]

It will be interesting to see if Tesla tests FSD in California or if they do it in a less restrictive state. I’m not guessing what their reported numbers will be, I’m guessing what the actual number will be if you count preemptive disengagements.

 

[1375mlm]

The way I look at it:
If Autopilot was only half as capable as it is right now, it would STILL be twice as good as any other semi-autonomous system I could buy today.
If Tesla wasn't doing it, we would still be driving our old, dumb, stinker mobiles without any mention of level 5 autonomy in the next few years.
But here we are, even being able to have the discussion of the possibility of having level 5 autonomy within the next year is already a victory for Tesla.

So if they take twice as long to do it, they're still leagues ahead of literally every other company out there, and the gap is only widening.

As for what I think their timeline will actually look like, it seems to me that by the end of the year 90% of all driving CAN be done by autopilot, in the same way that it can be NOA works right now. Works great in most situations, as in 9/10 of them, but the 1/10 is scary and needs you to be watching for it.

 

[Kant.Ing]

I predict they will remove radar and GPS or any positioning system from the car or pretend they're still using. I am a firm believer that vision is all we need!

Added safety or redundancies are not necessary even if they're cheap and other cars have them.

 

[GeoX750]

I think you should define “good” and “ok”.
Miles between disengagements in downtown San Francisco?

I would like a youtube channel with videos showing people driving for miles and miles, like an entire 45 min-1 hr commute, with AP on, without disengagements. I get that it would be boring to watch, but don't want it sped up, need sound and everything. I have seen one TMC poster post his entire commute (I think when he got NOAP without confirmations) that I need to find again. Seeing how others are using AP would really be helpful to me.

Unless I am in stop-and-go traffic or low traffic conditions, I can't use EAP (2019.12) for more than a couple minutes without having to disengage and manually take over for a maneuver. Now, I admit I am a picky driver (what lane I drive in, where I am in the lane, how close I am to other cars, how fast I go past traffic in a slower, how I smoothly change lanes and merge, when I change lanes to exit, avoiding potholes and road hazards, no close calls, etc).

Most telling that it is AP and just not my driving, however, is whenever my kids are in the car and they hear the AP chimes they groan because they know the ride is about to get choppy, due to, sudden phantom braking, the car suddenly speeding up and right away slowing down as cars leave our lane and after the gap created traffic is already slowed down ahead, our car slowing down as cars cut into our lane (because we are not following close enough), reacting to lane merges and exits, jerky auto lane changes, etc.

It would be useful for me to see videos of others using AP for long periods of time on an actual drive, like an entire long commute, no test drives or demos.

 

[RDoc]

With the possible addition of some Easter Eggs, my guess is that it will be pretty much what's out now. To be clear, I don't think the actual released for everyone version will do what the investor's day demo did.

 

[Tam]

...need sound...

No disengagement with plain AP2 V8.1 for 200 miles in 3.5 hours with various speed from 70 MPH to stop and go traffic.

No disengagement is easily achievable if you pick the correct routes.

Youtube refused my video with sound due to Arabic music that I listened from Slacker while driving. I had Youtube deal with the issue but it didn't go anywhere so I had to mute the sound myself and re-uploaded a current silent video.

I'll try the commute again but this time will be with NoA in June 2019. I'll make sure to shut the radio off this time so I can keep the sound portion for video on. Stay tune!

 

[GeoX750]

Thanks!! I have only watched 20 mins, but I can tell its going to be very helpful and revealing. The two lane, straight divided highway with little traffic and no merges just does not exist where I drive. I can't wait to get on a road trip.

Kudos for excellent driving and lane discipline and great technique to avoid nags.

 

[diplomat33]

With the possible addition of some Easter Eggs, my guess is that it will be pretty much what's out now. To be clear, I don't think the actual released for everyone version will do what the investor's day demo did.

I highly doubt that. It seems very unlikely to me that a whole year from now, Tesla would still be testing FSD in beta and not release a single feature to the public. Especially, since Elon has already said that we can expect FSD features and the FSD computer upgrade in about 2-3 months. Now, even taking into account "Elon time", I still doubt that they would go a whole year without being able to release even a single FSD feature.

 

[diplomat33]

I predict they will remove radar and GPS or any positioning system from the car or pretend they're still using. I am a firm believer that vision is all we need!

Added safety or redundancies are not necessary even if they're cheap and other cars have them.

Sorry but I have to disagree. Even if vision works 100%, redundancy is still needed. What if a camera gets dirty or goes out?

 

[boonedocks]

I highly doubt that. It seems very unlikely to me that a whole year from now, Tesla would still be testing FSD in beta and not release a single feature to the public. Especially, since Elon has already said that we can expect FSD features and the FSD computer upgrade in about 2-3 months. Now, even taking into account "Elon time", I still doubt that they would go a whole year without being able to release even a single FSD feature.

I can’t help but cringe every time someone says “but Elon said”

 

[electronblue]

Q2/2020 then. On the 22nd Elon Musk said Level 5 no geofence is feature complete (Q4/2019) and should be live somewhere in the U.S. (Q2/2020), so I picked that — Tesla’s FSD is L5 autonomous! — solely based on Elon Musk’s information who should know best.

 

[diplomat33]

I can’t help but cringe every time someone says “but Elon said”

I am well aware of Elon's track record when it comes to promises. But his words still carry some meaning in terms of intent at least. Even if they are late or unable to release features in 2-3 months, Musk's words still signal an intent to try to do it. So we can at least know that Tesla wants to release features in 2-3 months. My point is merely that I highly doubt that nothing will change in 12 months since we know Tesla wants to at least try to release something in 2-3 months.

 

[electronblue]

I can’t believe I’m the only one here who believes Elon here so far!

 

[jerry33]

I can’t believe I’m the only one here who believes Elon here so far!

Easy, Elon time--usually a year later, occasionally longer.

 

[boonedocks]

I am well aware of Elon's track record when it comes to promises. But his words still carry some meaning in terms of intent at least. Even if they are late or unable to release features in 2-3 months, Musk's words still signal an intent to try to do it. So we can at least know that Tesla wants to release features in 2-3 months. My point is merely that I highly doubt that nothing will change in 12 months since we know Tesla wants to at least try to release something in 2-3 months.

It was more of a joke than anything else. I was one of those guys that quoted Elon's 3 months maybe 6 moths definitely defending him to the naysayers. Well.....2+ years later, and still waiting, and I'm just a bit jaded at quoting Elon anymore. (That was just one of many examples though), No harm or ill will just 1/2 kidding 1/2 not