Musk: Tesla Full Self-Driving Capable By End Of 2020


Believes automaker could sell 3 million units by 2023

Elon Musk has lots to say about autonomous vehicle capability and Tesla. Luckily for those who’d like to hear his thoughts, he made an appearance on a podcast (audio below) by ARK Invest, where he spoke with its founder and CEO/CIO Cathie Wood, along with analyst Tasha Keeney. Published today, it sheds some light on the automaker’s projected timeline and its approach to the problem of cars driving us to our destinations without intervention.

Before the half-hour discussion broached the main topic, it began with talk about electric vehicle sales estimates five years out. Basing their figure solely on projected battery prices from their analyst Sam Korus, they suggested a 20-fold increase in EV sales globally by 2023. This bullish figure would mean a third of global sales would be electric by then. Musk said this sounded reasonable, though added it could be off by a year or two.

When asked about his own outfit’s output in that same period, he first prefaced his estimate by saying it would involve a lot of guesswork and that it’s very difficult to predict things “on an exponential curve” because a year or two can make a huge difference. Having said that, though, he then proceeded to suggest Tesla could produce as many as 1.5 million cars by 2021 and up to 3 million vehicles in 2023.

When the conversation turned to vehicle autonomy, things got a lot more interesting and featured more bold predictions. Musk says that by the end of this year — 2019 — full-self-driving optioned Tesla vehicles will be feature complete. He explains thusly,

“…meaning the car will be able to find you in a parking lot, pick you up, take you all the way to your destination without an intervention…”

He goes on to state he is very certain of that progress — he explains that he directly manages the Autopilot and engineering every week in detail — but clarifies that doesn’t mean it will be perfect or not need human oversight. There is also the issue of gaining approval from regulators to be taken into account.

Musk goes on to say that he believes the vehicles will be able to self-drive without human intervention by the end of next year — 2020. This is a lot sooner than most autonomous vehicle experts believe the automaker can accomplish the feat, and we have to be somewhat skeptical of the prediction in light of the fact that we’re still waiting for the previously touted self-driving coast-to-coast trip (with hands-free charging) that was supposed to happen in 2017.

In any case, the next two years promise to be exciting ones for fans of the California automaker and (possibly) autonomous driving. You can hear the entire podcast below, which has a lot more about why Musk is confident in his approach over others. There’s even a bit about cryptocurrency there at the end. Enjoy!


Source: ARK Invest

Categories: Tesla

Tags: , , ,

Leave a Reply

113 Comments on "Musk: Tesla Full Self-Driving Capable By End Of 2020"

newest oldest most voted

The Coast to coast Tesla trip, with Level 5 autonomy, is finally not far off (24 months or less)!

2 years ago!!!!

Didn’t happen.

That’s not what he said. “Without intervention”. but you are going to need to stay alert and ready to take over. So really good L2. Not L5.

He said you could fall asleep in your car and it would drive you to your destination about the end of next year. That translates from Elon time into roughly mid- 2022.

> He said you could fall asleep in your car and it would drive you to your destination about the end of next year.

I’ve seen the videos — there are some people who seem to think it can do that already.

Good point! 🙂

> So really good L2. Not L5.

That kinda shows up how badly thought out the five levels are. Levels three and four may as well not exist.

Elon said it will be Feature complete by End of 2019 for certainty. Meaning it will be level 4 for certain in less than 11 months.
Usually Elon doesn’t use for certain, usually a “Maybe” here and there. This is like gold to investors.

Think about the premium people will pay for any level 4 or 5 AP cars, especially ones that can make you passive income. A 50k Tesla might just be worth 100k in the 2nd hand market to taxi fleet or Uber. Like… I cant even imagine how it would be like a Tesla doubles it’s resale value overnight kind of scenario…

He’s not talking about level 4 this year. And “end of next year” is beyond his event horizon.

True. By the end of next year, autonomous levels will be meaningless. Maybe that is why he never mentions them

Only if that was going to happen…

It is pretty clear he is saying Level 4 autonomy, which I doubt very much you could use as an automated taxi. It would be interesting to know what constraints to the driving envelope would be.

It’s not clear at all. I wish the interviewers would actually ask him to define what a feature complete FSD is in comparison to the SAE levels.

I think the problem is that the SAE levels are too loosely defined. They sound like they’re clear and unambiguous, but they’re not. They fail to consider different approaches to building up to full autonomy. For example, Waymo are working within geofenced areas, trying to solve the problem entirely but only within that area, whereas Tesla are working without a fixed area but solving the problem in stages by getting it working in certain conditions, and then introducing more complexity. Waymo’s approach could attain absolute full autonomy within its geofence but could conceivably be useless outside that area. Whereas Tesla could attain full autonomy that can go almost anywhere but needs the driver to take over, for example, on certain types of roads or junctions. Which level would they both be at in this scenario? Which one is closer to level 5? Which one is more useful? Note, I’m not trying to predict how things will actually pan out (we’ve had lots of ambitious promises from all parties that haven’t materialised); just showing how the SAE levels don’t really fit with the differing development models that are being taken. Even if you ignore all of that, it’s been proven pretty… Read more »

Autonomy has too many possibilities to boil down to a few categories, but SAE levels are not as vague as you say. For example, you do not have to pay attention with Level 4.

The last time he said for certain, cast iron guarantee, was 10k model 3 by the end of 2018. They are going to miss that by at least a year.
So I would add 12 months at least to anything he says that is for certain going on history.

No. He said target production was 10k cars per week (S, 3, X).

Are you sure? I’m sure he said that about model3 production!

Not Level 4. Really good Level 2.

Erm… sure?

Level 2: In level 2, at least one driver assistance system of “both steering and acceleration/ deceleration using information about the driving environment” is automated, like cruise control and lane-centering.

Level 4 vehicles are “designed to perform all safety-critical driving functions and monitor roadway conditions for an entire trip.” However … it does not cover every driving Scenario.

Call it 2+ if you like, but it’s not Level 3.

Level 4 means driver can zone out or sleep and not monitor the environment while the car is doing it. But the car might not be able to handle all driving scenarios. So, like it could be L4 on the freeway or in a parking lot, but not in city traffic or bad weather.

L2 require the human to monitoring the environment and ready to take over. Even if the car can manage to navigate and automate most driving situations.

I don’t understand the logic of this. When you say it might be worth $100,000 second hand do you mean they can earn that much or they would pay that much? Why would they pay $100,000 for a used car when they could buy a new one for half of that?

“Feature complete” by the end of this year is so nondescript.

WTF is ” feature complete”, It means nothing!

The main take away from the podcast for me was when Elon said the last ten percent of FSD was “really hard”.

“Really hard” is a term with meaning!

“Feature complete” = “it mostly works” in computerese

It means the software side has matured to the point where all it needs is to be fed with data and it becomes more and more accurate at accomplishing the task of independent full self driving. If you listen to the podcast, he did a good job explaining it as the “march of nines” where you keep feeding it and that reliability record goes up, basically tacking a nine at the end of that fraction.

Hypothetically speaking, If the AI is 99.99 percent reliable this year, it can become 99.999 percent reliable the next, 99.9999 percent after that and so on. That’s the basic gist of the explanation.

He also said Tesla will sell 1.5 mil cars in 2021 and 3mil in 2023. All full self driving ready by end of 2020…
I think Tesla will just displaces 50% of the cars on the road in a few years time.

If you look at Tesla Asset growth history over the last 9 years, it’s a believable statement.

That is what happened during the dot com bubble… Every company started drawing lines after the first few points…

9 years isn’t a few points and it’s not like Tesla has a huge overall marketshare. There is a lot of room to grow. Many people would love to be able to afford a Tesla.

9 points…. 1 for each year.

True enough, I didn’t build any car as last year so if I build to this year that is huge exponential growth

Exponential growth for 9 years is a trend line. — Regression Analysis.

Add 12 -18 months to his predictions and you get an accurate picture.
That is just how he rolls.

And I think I’ll keep getting better looking as I get older

Amazon looks better after 10 years.

At about 90 million units per year – 3 million units would only be a 3.3% market share. Very reasonable (BTW – that means they would be selling more than BMW)

(⌐■_■) Trollnonymous

Full autonomous……..meh.

I don’t want or need it.

That fact that you do not think through an issue tells me that you need it more than an average person. Remember when you did not need a smart phone, or the ability to have fire. It will be other people who will determine what you need.

Technically you don’t need a smartphone – it’s just a convenience.

Technically you don’t need a house, clothes, limbs, eyes, ears, or a tongue and those are just convenient, too.

Seriously? You equate a smartphone to ” a house, clothes, limbs, eyes, ears, or a tongue “? Are you 14 or younger?

American Indians had farms and lived off the land with none of those.

Yet the Indians sold Manhattan to the Dutch for a bunch of glass beads.

Native Americans bro not American Indians

Thanks for the correction.

I was wondering the same oh, don’t get me wrong I have a smartphone and it is very convenient, I’m on one right now as a matter of fact but at my work we get Summer help better usually College in high school age and I think most of them would equate a smartphone to having a set of eyes. You walk into a room with six of them sitting there and they almost never talk to each other each one is just staring at their phone, it’s almost creepy

If you compare having eyes and a house to having a smartphone then that is about as ridiculous an exaggeration as one could make

I don’t need my phone. Every human I see is on their phones. I am never on it. And I am young. Not brainwashed like every other human my age. Always in present time.

Well, you seem to be using the internet anyway so I guess what you’re saying is that you only use big computers and not small computers?

Good for you, it might make me a hypocrite because I spend some time on mine but I’m in my forties and when I get home I put it away in my kitchen and just check it every so many hours when I’m out there. My son just turned 12 and I got him his first phone but his time on it is limited but I see plenty of his friends who are out as a group and they spend almost the entire time looking at their phone.

I would argue that technically you don’t need a smartphone today. Granted there are things you can’t do if you didn’t have one but it is not a need

Maybe not yet, but you will get old some day. My 94 year old grandfather finally had to give up his keys and it was a gut punch. As evidenced by so many oldsters on the roads that wad up their cars because they drive when they shouldn’t. That final loss of freedom, the inability to travel around with impunity is the final nail in the coffin for many geriatric folks. Autonomy will save them from the ignominious experience of no longer feeling like they can participate in the world around them.

That’s why my mom has a smart phone an Lyft.

That will be a very good purpose for autonomy. We do have services like uber and Lyft which would probably be cheaper depending on how much they leave the house but people who grew up having their own car will likely want to have one as long as possible. I say that to my buddy at work who’s the nostalgic type despite being only 39 years old. His mother recently stopped driving and I told him imagine when you’re older or even if you want to go to the bar and have a few without worrying about it how nice it would be to have your car just drop you off and then pick you up.

Older generation, blind, handicap, etc, etc.

Whether or not you need it, Tesla is basically announcing that they are about to outcompete the entire surface transportation industry and no one seems to really appreciate that. Well no matter, you’ll get it when everything depends on autonomy.

Yes because there is literally no other companies on the planet working on autonomy. They’re still trying to master the synchromesh manual transmission

I’ve seen google product, they’re never number 1 in their field.
They’ve even destroyed their search results for profit.

So, stock market performance isn’t just about asymmetric information, those who have early access outPerform$. But, asymmetric PROCESSING of the information. In other words Tesla is an under appreciated GOLD MINE.

If cars become fully autonomous there is no need to own one. You can use it like a taxi without the driver and it should be cheaper and more convenient than owing a car.

You might own one but where you use to have multi-car families you might only have one.

If I want to go to the store to pick up a loaf of bread or whatever, I don’t want to wait around for a taxi. The business model for a taxi service I don’t think will change much whether the car is autonomous or has a driver.. I can see car sharing working under limited circumstances but I don’t see car sharing replacing the family car.

I used to think the same thing about electric cars, that if my electric car couldn’t make the trip because of limited range or whatever then I could just rent an electric car that could. What a joke that turned out to be. If you try to rent a Tesla it’s a couple hundred dollars a day and the mileage that’s allowed per day is usually very limited.

You’ve never used Lyft or Uber.

You can buy a Leaf for $11,000 today, that’s not a joke.

In some areas you can buy TWO Leaf for $11,000 today, that’s not a joke.

Wouldn’t that be “Leaves”?

You don’t wait. An autonomous cars turns up in your driveway faster than you can find your keys, get in your car, open the garage door, and back out your driveway.

Oh the joys of hitting critical mass.

Well that would clearly depend on having a large storage of cars very close to people’s neighborhoods which is not going to be the case everywhere especially not in the early days

I agree to an extent. Whoever these autonomous taxis are stored would have to be very close to people’s neighborhoods. If you could summon them by voice and have them there in less than 5 minutes and it’s significantly cheaper than owning your own I think it would take off more but like many I grew up having my own car and if the difference in cause wasn’t much I would prefer that. Having said that I could see a scenario where I have a car and on the occasion I need a truck an autonomous one could be summoned for that purpose. People would then be free to stop buying a vehicle that meets 100% of their possible needs

“You can use it like a taxi without the driver…” What’s the big deal with the driver?? You can use taxis, or Uber/Lyft right now and not own a car. Really, the guy making a living is the barrier for you? If you really believe that these new high tech self driving cars will be cheaper to ride in than the low tech Uber car with driver, you’re not paying attention to how these companies work.

Exactly I can summon an Uber or Lyft using a smart speaker at home or my phone.

Cost. Drivers cost a lot. The idea is that a fully autonomous robotaxi will be significantly less expensive to operate. Additionally, it will be the automakers running their own robotaxi networks, which will be more profitable for them then selling to individuals.

Shared self driving cars sound like a wonderful idea, until you consider how inconsiderate and unhygienic people can be. Taxis, Lyft and Uber have a human driver to ensure the car isn’t soiled and trashed by the passengers. Who will do that in self driving cars? When people use self driving cars as brothels on wheels, who will clean the bodily fluids off the seats before the next passenger hops in?

When you get perma banned from the service because you were a messy twat, you’ll soon find this a non issue.

That is easier said than done. What will likely happen is that a messy car will get used by many passengers before someone bothers to call the home office and complain about it. Then what do you do? Which one of the many passengers who used the car since the last inspection should be banned?

That’s the other thing people don’t take into account. A lot of people are not going to want to load their young kids into the back of a car that has had 5,000 people in the back of it doing god-knows-what over the last month or so. Also people with kids tend to store stuff in their car like chairs they sit in at a baseball game, stuff like that that you don’t want to load in and out every single time you go somewhere

Corner case for 100% dominance. Ok, so initially just 50% market penetration.

Let’s hope recreational vehicles persist. Sometimes there’s value beyond point A to point B.

Exactly I could see the benefits of autonomous taxi services especially with older people or even with kids. If they were safe enough a 12 year old to get a ride to school in one but it seems many people I think we are a race of Vulcans instead of humans

I think I’d still own one, but imagine how it will be? The car can take me or my partner to work, then drive home to take the other to work. Same at the end of the day. Share one car rather than having two. Know this wouldn’t work for everyone, but really opens up possibilities.

A legal fully autonomous system will probably add at least $10k on to a car. That’s fine if you are displacing a the cost of a driver, like in a Taxi, but for the average driver it’s probably going out of reach for a while. Around town I really don’t see much point but for long trips it would be nice to relax or sleep while the car drives itself down the road.

10k for full autonomy or another 35k for a second car? It’s a no brainier.

I think the cost is already set at $5k.

$5k for EAP and another $5k for FSD. So, yeah $10k.

Student loan debt payments are around $500 a month.
That’s the Car Payment.

I’m sure many are dropping the car option and car insurance until the debt is paid, along with first time home ownership which is also dropping. That’s helping Lyft.

People don’t understand exponentials, ok. But InsideEVs don’t even understand decimal numbers, or comprehend what Elon said about millions of corner cases that Autopilot needs to solve before full autonomy. He also made very clear that Autopilot was not full autonomy, and he made reservations vs regulatory approval etc. Elon has the alfa version of Autopilot in his own car, with full self driving features enabled. Hard to think of anyone better equipped to make estimations about its future developement speed than him. #exponential curve

Didn’t the article talk about level 4 autonomy not level 5?

Okay so if everybody else is stupid according to you and he really knows because he’s got the Alpha version then why did he say a Tesla drive itself across the country in 2017?

Maybe autonomous vehicle capability will be ready in 2 years. To be implemented everywhere though will take a lot longer, say 10-15 years or more. Legal liability will be the highest hurdle.

Elon, the neck is one gimbal, but the eye is mounted on horizontal, vertical, and torsional orthagonal gimbals. Integrated cameras give depth perception, up and down regulate based on lighting and have an aperture control system, to name a few features. I’m a huge Tesla fan but I wouldn’t say you can match the human visual system and brain as a vehicle guidance system so glibly.

He’s just using the most advanced, expensive, car AI system on the planet, “glibly”, I guess.

“we have to be somewhat skeptical of the prediction in light of the fact that we’re still waiting for the previously touted self-driving coast-to-coast trip (with hands-free charging) that was supposed to happen in 2017.”

Yup! Credibility check.

Elon usually delivers, but on self driving, he hasn’t delivered nearly as much as other area as far as the promise or dates are concerned..

Similar for Canadian Supercharger’s on the Trans Canada….!

Or solar powered Superchargers.

(yes, a couple sites have some token panels, but years later the grand vision has not even partially materialized)

It’s pretty sad that they made a point of
concentrating on autopilot and then didn’t ask any penetrating questions, like-
What exactly do you mean by FSD?
Why is your definition different to the international agreed standard?
Why don’t you call it level3 or level4?
When will your car be able to drive me all around the city while I sleep?
When is your coast to coast drive?
Why do you make promises about autopilot and then it doesn’t happen?
Why should we believe you this time?

They didn’t ask any penetrating questions because if they did he wouldn’t let them interviewing another day

The podcast did allot of explanations and has technical details addressing allot of those questions, have you seen the full podcast? It was pretty good.

Try around 9:52 and 10:13 for example, and later he talks about capability vs regulatory constraints later on etc.

Loved the crypto section.
When they bring up crypto he is like.- “Seriously, Crypto!?”
It’s kind of like he is saying -I thought this was a serious interview.

I mean the guy did do Paypal, so he kinda understands “internet money” per se. Plus there are plenty of Model 3s on the road today no doubt purchased with Bitcoin profits. Contrary to the stereotypes, they didn’t all go buy lambos.

If Tesla is making 3mil cars in 2025(quite possible) who’s making the other 20+ million?

USA sales are around 17 million, the same for Europe, in China they’ll be probably closer to 30 million by that time, add Japan, Latin America, Africa and the rest of the world. We are taking close to 80 million cars a year.
Those remaining cars will be sold by Toyota, Renault, VW, Mercedes, Ford, Honda, …

All the other manufacturers. Tesla is here and now, yet BMW, MB, Audi all still sell so many more cars than you can imagine. Why would you think that will change? Some people just buy the same brand because of this thing called brand loyalty, it’s a real thing believe it or not.

BMW and MB and Audi don’t sell more than 3 million cars a piece. Musk is declaring that he’ll outsell all of them.

Audi is simply a division of Volkswagen who has a group sold 11 million cars last year

How is that relevant? No one mentioned VW.

They should have asked what “Feature Complete” means, and what is in the list of “Features”. It would be helpful to measure their progress. Sign recognition? Intersections? Automatic connection at Superchargers? Ability to park at multi-level parkings? I’m sure we would also be surprised what is not on the list al all.

This seems like a good practice that every manufacturer should be reporting:

“Over the past quarter, we’ve registered one accident or crash-like event for every 3.34 million miles driven in which drivers had Autopilot engaged.

For those driving without Autopilot, we registered one accident or crash-like event for every 1.92 million miles driven. By comparison, the National Highway Traffic Safety Administration’s (NHTSA) most recent data shows that in the United States, there is an automobile crash every 492,000 miles. While NHTSA’s data includes accidents that have occurred, our records include accidents as well as near misses (what we are calling crash-like events).”

“In the 4th quarter, we registered one accident for every 2.91 million miles driven in which drivers had Autopilot engaged. For those driving without Autopilot, we registered one accident for every 1.58 million miles driven. By comparison, NHTSA’s most recent data shows that in the United States there is an automobile crash every 436,000 miles.”

So you still have to stay alert the whole time which I would imagine means you can’t summon your car to come and get you because then by definition you’re not inside ready to take control obviously. Also it won’t be legal so even if it isn’t really ready for prime time and you can say it is what but it’s the Regulators fault

Do Not Read Between The Lines

22 months maybe, 44 months definitely.

I guess most people have no clue that FSD capability does NOT equal L4 or L5 regulatory approval …. this is like almost apples and oranges.

Autopilot is still plowing into fire trucks and accelerating into gore dividers. Full self-driving by 2020 will not happen, even if restricted to divided highways.

A 2 year old child is feature complete, but that does not mean it is Albert Einstein already.
Not the Implementation of all features, but a perfect quality of the features is the big step to take.