Timestamped Guide To Part 2 Of Elon Musk Interview By Third Row Tesla
Credit to Author: Paul Fosse| Date: Fri, 31 Jan 2020 21:49:28 +0000
Published on January 31st, 2020 | by Paul Fosse
January 31st, 2020 by Paul Fosse
This is your guide to part 2 of the Elon Musk interview done by Third Row Tesla. There isn’t really any need to read about or watch part 1 before part 2, but if you want to, here is the link to my article on part 1 of the interview. I think they both stand on their own, so you can read, listen to, and/or watch them in either order.
As with my previous work, this is not a complete transcript of the content, only the highlights, so I’m sure there are some things that I miss. If you have the time, I encourage you to listen for yourself.
0:00 — Third Row Crew going on about how great Autopilot is.
1:10 — Elon states that cars will be capable of “crazy maneuvers like a high-speed chase.”
1:20 — Elon drops the bombshell that a rewrite of the Autopilot software is almost complete. This rewrite combines the planning, perception, and image recognition functions. It sounds like it is moving more function to being “learned” by the neural net and less function being hand coded. My take: Unlike the conclusion drawn in this article, this is a major delay to the Full Self Driving program. This means the demo we saw at Autonomy Day last year was judged to be not good enough to put into production and they had to go back to the drawing board. As we learned from listening to Andrej Karpathy talk about the tradeoffs between combining things and keeping them independent, putting 3 functions together will make if more efficient, but that will have 2 other effects. The interfaces to all the other parts of the system need to at a minimum be retested and maybe redesigned. Also, the ability to just quickly hand code around a small known issue in planning, perception, and image recognition is removed. Instead, you have to train the neural network around each known issue. I think this is a good change, but it probably sets the system back close to a year.
2:30 — 3D labeling allows 100 to 1000 times the efficiency, since instead of labeling each frame of each image from each camera, they are labeling a few moments of video from all 8 cameras with a single label. 1000 images labeled together. Also mentioned: the Dojo training supercomputer should get up and running late this year or early next year.
5:20 — At some point in the future (maybe a decade) autonomy will be easy (like GPS technology today). My take: Even companies like Apple and Google will be able to figure it out. (Having worked for companies that were perceived as leaders in technology, I realize how quickly they get fat, dumb, and happy after their initial success.)
6:00 — Still not a car that matches the 2012 Tesla Model S.
6:50 — “It’s not going to be that hard to put an electric motor in a nice lightweight car,” they thought long ago. Tesla’s idea was completely wrong — they had to redesign both the motor and the car.
8:20 — Key is to adapt quickly when you find you’ve made a mistake.
9:20 — Omar makes a great comment: “You need to be naive, because had you known how hard it was, you might not have done it.”
11:30 — Elon tells us that the Roadster was built in an old Fold dealership in Menlo Park they had rented cheaply.
14:30 — Shifting to Model S and the Fremont factory: GM just left us the junkiest equipment but we made it work. Top suppliers either wouldn’t work with us or would give us their worst team.
16:00 — “Vertically integrate or die.” We were going to have battery modules made at the same place that made BBQ grills. We brought it in-house.
16:30 — Massive changes from the prototype to production means you need a tight feedback loop to make it work.
18:10 — Don’t really need battery modules anymore (since quality control is high), could just put cells in a pack. Model 3 has them, but they serve no purpose.
22:00 — Errors in the organization manifest in the product. Since they had a module and a pack team, they had both. Combine the team and the product combines.
23:00 — Box in a box. Model 3 battery is enclosed and the body is enclosed. Should just be one covering in the future.
24:00 — Tesla was first American company in 80 to 90 years to reach volume production of a car. Regulatory requirements so much steeper than 80 to 90 years ago.
25:15 — Point Elon hasn’t heard from short sellers, but which is valid: Existing manufactures make most or all of their money selling high-margin parts to their existing fleet. Like printers and cartridges, razors and blades, game consoles and games, the cars are loss leaders. The money is in the service. Since 90% of Tesla cars are under warranty, they get little revenue from service. This means their cars have to be so good that people are willing to pay a premium for them. They can’t be subsidized by the service profits of the existing fleet.
27:00 — Dealerships get 20% of their revenue but 50% of their profit from service (according to Edmunds).
28:50 — 2 things new car companies must do to be successful: #1 Electrification. #2 Autonomy.
31:00 — Tesla’s goal is to minimize service cost. The dealer’s business is to make money on service, which creates a conflict of interest between the dealer and the customer.
32:20 — Schumpeter’s Creative Destruction: Innovation tends to come from new entrants.
34:20 — What do you think of Rocketlab’s approach of using a helicopter to retrieve the first stage? A lot can go wrong. Overall, Elon is impressed with the company.
36:10 — City to city travel in 30 minutes. Noise is biggest concern.
38:10 — How do you make all the Raptor engines?
39:30 — Rockets have to be made in the US (because of restrictions on weapons).
40:00 — If we have city-to-city travel in 30 minutes, do we need electric airplanes? Yes, but Tesla can’t do everything.
42:30 — Gali suggests that company is financially better off, so it can spend more on research and development. Elon retorts, “That’s not how it works.” The shortage is in talented engineers. You can’t just ramp with money. There just aren’t that many exceptional engineers.
44:00 — So many mistakes on the Model 3 — the whole company had to work to correct them. Took talent off solar, Powerpacks, and other car projects.
46:00 — Biggest problem Tesla is solving now is making cars on each continent. Solves transport, tariffs, and damages in transport.
48:00 — First 6 weeks of production for Europe and Asia, next 3 weeks for the East Coast and last 3 weeks for the West Coast. All the cars show up at the last minute and have to be delivered by end of quarter. This is called the wave. Painful, but if they don’t do it, financials will suffer.
49:40 — Local production will break the wave.
51:00 — Cars stuck in Shanghai because of the wrong sticker.
52:45 — Vincent asked, “How does Chinese public perceive EVs?” Elon answered they are pro EV and the largest EV market in the world.
54:00 — Elon explains all the advantages of local production in China.
57:00 — Local production on the same continent really helps the financials because you get paid for the car before you have to pay your suppliers.
58:40 — Why Berlin? It has the best nightclubs! We needed to move quickly and BMW had already done a lot of the environmental work on the site (saves Tesla a year). Local and state government very supportive. A lot of talent in the area.
1:02:00 — Roller coaster still coming at Fremont?
1:03:40 — Did you expect so many Cybertruck orders? They discuss the development and release of the product.
1:07:30 — “We want to inspire people by building something different.”
1:08:00 — Travis Scott music video discussed. Then Elon talks about how it will be a real engineering challenge.
1:10:00 — Closing thoughts by the Third Row Tesla team.
The biggest news is the redesign of the Autopilot and Full Self Driving software. I realize my take that this is a one-year delay is more pessimist than most, but delivering software is my day job and I have been doing it longer than even Elon has. It is very tough for him (or anyone) to admit to a delay all at once. On the earnings call, we saw a combination of a 3 month delay in “functionally complete” FSD and a significant lowering of the bar of what that means. Previously, it meant it had a better than 50% chance of getting you to your destination without an intervention. The new definition is it might get you there without an intervention. I interpret that as we won’t get to the original goal till the end of 2020, about a year of total delay.
Other than that bad news, the rest of the news on Tesla and the fact that it’s currently learning to scale up to 3 continents is all very positive!
If you decide to order a Tesla, use a friend’s referral code to get 1,000 miles of free Supercharging on a Tesla Model S, Model X, or Model 3 (you can’t use it on the Model Y or Cybertruck yet). Now good for $250 off on solar, too! If you don’t have any friends with a Tesla, use mine: https://ts.la/paul92237
Follow CleanTechnica on Google News.
It will make you happy & help you live in peace for the rest of your life.
Paul Fosse A Software engineer for over 30 years, first developing EDI software, then developing data warehouse systems. Along the way, I’ve also had the chance to help start a software consulting firm and do portfolio management. In 2010, I took an interest in electric cars because gas was getting expensive. In 2015, I started reading CleanTechnica and took an interest in solar, mainly because it was a threat to my oil and gas investments. Follow me on Twitter @atj721 Tesla investor. Tesla referral code: https://ts.la/paul92237