Elon Musk's answers in 2020
Q1 How dangerous is AI and how do we make it go well?
They are very different, but the most fundamentally transformative will be AI. link
He says he has been banging the AI drum for a decade. link
Q3 Where do you stand politically?
He condemned orders that stopped people leaving their homes as fascist and undemocratic. link
He calls the extended shelter-in-place orders forcibly imprisoning people against their constitutional rights. link
He says people who want to stay home should be allowed to, but should not be compelled to. link
Q5 Why go to Mars / what is SpaceX's mission?
Establishing a base on the Moon or Mars takes a tremendous amount of work to make it self-sustaining. Once the transportation problem is solved, many things are possible, like the Union Pacific Railroad opening the West Coast. link
The whole purpose of SpaceX is to help make life multiplanetary. Satellite-launch revenue is a way to fund that; broadband is a much larger market. link
He said designing a house in obsessive detail is less important than allocating that time to getting people to Mars. link
He said this is hopefully the first step on a journey toward a civilization on Mars, life becoming multiplanetary for the first time in Earth's 4.5 billion year history. SpaceX's fundamental goal is the technologies to make life sustainably multiplanetary. link
He said human spaceflight was always the fundamental goal of SpaceX. link
He said humanity should be on track to become a multi-planetary species and a space-faring civilization, to find out what the universe is about and keep consciousness going. link
Says Starship is designed to make life multiplanetary. link
He is mostly concerned with technology that enables a lot of people to go to Mars and make life multi-planetary: a base on the Moon and a self-sustaining city on Mars, as life insurance for life as a whole. link
Q6 When will AI surpass humans?
He says the people most wrong about AI are very smart people who cannot imagine a computer being way smarter than them. link
AI will be able to do everything better than humans over time, everything. link
Q8 Why are your timelines so often late?
He says production almost always follows a herky-jerky S-curve. link
He says the intermediate portion of the production S-curve is very difficult to predict. link
He says he gives the 50th-percentile estimate, so about half his predictions will be wrong. link
He says absent a massive force majeure event, 40% growth is the likely number. link
He says his estimates are the likely midpoint, not a figure with lots of margin. link
He says his predictions may come true late, but they do come true. link
He is optimistic in general, and that extends to schedules. He would wish they could do it in two years, which might mean four; he would be surprised if it took more than four. link
He says manufacturing capacity is underestimated at first and later sometimes overestimated, because ramps follow an S-curve. link
He says manufacturing follows the S-curve, which people who have not spent time in manufacturing often miss. link
He says new factories typically take 12 to 18 months to reach capacity. link
Q11 What drives you / what is the meaning of life?
Human spaceflight was always the fundamental goal. We must make life sustainably multiplanetary — not one planet to the exclusion of another — while the window of opportunity is open. link
The point is to find out what the universe is about and ensure the continuance of consciousness as we know it. link
Take a moment to appreciate the little things, the big things, the things that move your heart; that is as close a definition as he can think. link
Q12 What's the biggest remaining Starship problem?
He said getting the domes right on the propellant tanks is one of the most difficult things. link
It is absolutely fundamental to achieve full reusability in access to space — the holy grail of space. Falcon 9 is only partially reusable; Starship has the potential for full reusability. link
There is really just one thing that matters: a fully and rapidly reusable rocket, reasonably big. Without that, we're nowhere. link
After immense effort Falcon 9 achieved booster reusability and is mostly reusing the fairing, but he called that architecture close to a local maximum. link
Four essential elements of making life multiplanetary: rapid complete reuse, low-cost propellant, orbital refilling, and propellant production at the destination. link
Q14 What do you think of OpenAI?
Says he has no control and very limited insight into OpenAI. link
Says OpenAI is essentially captured by Microsoft. link
Q15 How competitive is China?
He told the team they achieved an incredible outcome that hardly anyone expected. link
He says making cars in California and shipping them halfway around the world was pretty silly. link
China has a lot of really smart, hard-working people. Independent of government policy, the Chinese economy is going to be at least twice as big as the US, maybe three times. Without radical innovation the US will be voluntarily second. link
He says the US should invest in infrastructure, and that China's infrastructure looks much better than America's. link
He promises miners a giant long-term contract if they mine nickel efficiently and in an environmentally friendly way. link
He asks mining companies anywhere in the world to mine more nickel. link
He says Tesla's long-term competitive strength will be primarily manufacturing. link
He says the Tesla China team's progress is beyond all reasonable expectations. link
Q16 How good is Tesla self-driving?
In January 2020 he admitted he had hoped for FSD feature-complete by the end of 2019, and said it might come in a few months. link
He says Tesla has a real-time information advantage over insurance companies on how cars are driven. link
He says labeling video from all eight cameras at once is roughly a three-order-of-magnitude improvement in labeling efficiency. link
He explains that FSD "feature complete" only means some chance of driving home to work with no interventions. link
He says Tesla is only beginning to take full advantage of its Autopilot and FSD hardware. link
He says drivers effectively label reality as they drive, which keeps improving the neural net. link
He compares Tesla's fleet data to users training Google every time they click a search result. link
He says Tesla has quite literally orders of magnitude more data than everyone else combined. link
He says there is a tremendous amount of untapped potential in the fleet that could upgrade to Autopilot or FSD. link
He describes releasing supervised FSD first, then improving reliability until no driver supervision is needed. link
He says Tesla would soon collect data from over a billion intersections per month. link
In July 2020 he said he was very confident FSD functionality would be complete by the end of 2020, because he was driving it himself. link
He says he personally tests the latest alpha FSD build for city streets and finds it profoundly better than people realize. link
In July 2020 he said his own drives were close to going from home to work with no interventions. link
He says it is hard to convey how much better a fully 4D system, correlated in time, works than one built on isolated 2D images. link
He expects people in fully self-driving cars to spend the time on productivity and entertainment. link
He says each order of magnitude of reliability takes a massive amount of work. link
In 2020 he said Autopilot was moving from 2.5D image recognition to a new architecture that matters for full self-driving. link
He says FSD is a general neural-net approach that needs no high-definition maps or cell connection. link
He compares Tesla's million-car fleet feedback to how Google search improves. link
He says weird corner cases are what is really valuable, and only reality can provide them. link
He says the car should drive like a person even with no connectivity in a place no Tesla has been. link
He says FSD should drive like a person even with no connectivity, somewhere no Tesla has been. link
Q19 Is population decline a problem?
Calls population collapse the 2nd biggest danger to civilization after AI. link
Q22 When will Starship go to Mars?
Load the Mars fleet into Earth orbit, then about 1000 ships depart over ~30 days every 26 months. link
He hopes he is not dead by the time people go to Mars. If they do not improve the pace of progress, he will definitely be dead before Mars. 18 years just to get people to orbit; they had better improve the rate of innovation. link
If progress is exponential, an uncrewed mission could go in maybe four years, aiming at the Mars transfer window four years out. link
He feels fairly confident about six years from now, timed to Earth-Mars synods roughly every 26 months. link
Q26 What do you think of tariffs?
He says Tesla gains a big efficiency by making affordable cars on the continent where the customers are. link
Q30 How do you hire and manage people?
You don't need college to learn; everything is available for free. The value of college is seeing whether someone can work hard through annoying homework, and hanging around people your own age. He does not consider going to college evidence of exceptional ability. link
He says designing new production systems is extremely exciting and fulfilling. link
He says Tesla probably contains more than a dozen startups, with every major product line being a startup. link
He says there are more than a dozen effective startups inside Tesla, every major product line. link
He says Tesla is not dependent on enterprise software, which is a very big deal. link
He wants leaders on the front lines: nobody bleeds for the prince in the palace; troops fight harder if the general is there. link
Q37 How should AI labs manage safety?
Says all orgs developing advanced AI should be regulated, including Tesla. link
Q44 What is Neuralink's near-term goal?
Says high-bandwidth neural interfaces are underappreciated; Neuralink may be in a human this year. link
Not testing on people yet; he thought they may be able to implant a Neuralink in a person in less than a year, with low rejection risk like other implants. link
Says Neuralink will transfer processing to a supercomputer extension of the brain. link
Short to medium term Neuralink is to help cure brain and spine injuries. First implanted devices in humans will be for quadriplegics. link
Q45 Is there hidden alien or advanced propulsion technology?
People think there are aliens, but he has not seen any sign of them; as far as we know we could be the only life. link
Q46 Do regulations hold you back?
He says low labor mobility in Europe, including months of garden leave, worries him for Berlin hiring. link
The US is still great for entrepreneurship, but the gradual creep of immortal rules will eventually let us do nothing. Default: government should get out of the way and be a referee, not a player picking winners. link
Q47 Has Tesla come close to failing?
In April 2020 he said Tesla was doing the opposite of other companies and going pedal to the metal on new products and expansion. link
He says the crisis made Tesla look closely at its cost structure, as one always must in a crisis. link
He says in the 2020 crisis Tesla decided to keep expanding rapidly and investing in new technology, even though risky. link
When starting SpaceX he thought they maybe had a 10% chance of reaching orbit. He spent the PayPal money on SpaceX and Tesla; almost both companies went bankrupt. 2008 was a tough year. Four Falcon 1 attempts, then the NASA cargo contract saved SpaceX. link
Notes Ford and Tesla are the only US carmakers that avoided bankruptcy. link
Q49 Why is Tesla building a chip fab?
He says supply chains send materials around the world many times, and vertical integration can cut that. link
Q53 What's your vision for energy abundance?
He says buyers will choose between a roof that is alive with power or one that is dead without it. link
He says the real problem is scaling battery production to levels people cannot fathom. link
He says the solar glass roof faces challenges but will be a major Tesla product line. link
Tesla open-sourced its patents years ago. Creating a patent portfolio that discouraged others from making electric cars would be inconsistent with accelerating sustainable energy. The real protection is innovating fast enough. link
In April 2020 he said Solar Roof installs should reach 1,000 a week within about a year, maybe by the end of 2020. link
He says Tesla wants at least 1,000 solar roof install teams, for about 1,000 installs a week. link
He pitches Tesla Solar with a Powerwall so homeowners can be energy-independent and act as their own utility. link
He lists the three elements of a sustainable energy future: wind and solar generation, battery storage and electric transport. link
He says battery and solar will both be enormous, as they must be for a sustainable future. link
He said roughly 10 TWh a year of battery production is needed to transition the global vehicle fleet to electric. link
He said roughly 20 to 25 TWh per year sustained for 15 to 25 years would transition the world to renewable energy. link
He said battery production would move from gigawatt hours to terawatt hours. link
He imagines an integrated home system for generation, storage, heating, cooling, air filtration and water purification. link
He expects batteries in a few years that cost half as much with a third or less of the capex. link
Range over 600 km is already shown; the fundamental impediment is battery cost so everyone can afford an EV, with a long-term cell-cost goal around 50-55 cents per kWh. Berlin-Brandenburg is meant to become possibly the world's largest cell plant. link
Q57 Will Starlink work as a phone?
He says the Starlink antenna is about the size of a medium pizza, likely more bandwidth than a car needs. link
Q59 Will Tesla own the robotaxi fleet?
He says Tesla would likely enable car sharing before the robotaxi fleet, because car sharing does not need regulatory approval for full self-driving. link
He says upgrading the fleet to full self-driving over the air may be the biggest asset value increase in history. link
He says the step change to full self-driving is probably worth at least $100,000 per car. link
He says full self-driving would raise a car's use from about 12 hours a week to about 60, five times more. link
He says an Uber or Lyft driver could manage a fleet of 10 cars. link
Q60 Will Starlink go public?
Says Starlink will probably IPO, but only once revenue growth is smooth and predictable, with priority for small retail investors. link
Says Starlink will most likely go public once its revenue growth is reasonably predictable. link
Q70 When will Starship fly hourly? (asked on X)
Starship is being designed to relaunch an hour after landing with zero nominal work. He wants a capability of three flights a day for the ship, most of that waiting for the ground track to come over the launch site. link
Q79 When does Tesla stock have its NVIDIA moment? (asked on X)
He says many retail investors have deeper and more accurate insights than big institutions and analysts. link
He says the mind-blowing financial picture is high volume and high margin because of autonomy. link
He says that with autonomy Tesla can go way beyond industry margins while making cars more affordable. link
Q100 What will Tesla's Robotaxi look like? (asked on X)
He notes average car occupancy is about 1.2 and may only rise to about 1.4 with autonomy. link
He says Tesla will probably make a minivan or Sprinter-like vehicle at some point, once batteries allow. link
Q104 Why is Tesla putting Starlink in its vehicles?
In January 2020 he said there were no plans to put Starlink in cars that year, since Starlink was aimed at homes, businesses, aircraft and boats. link
He says cars would mostly use cellular, but in the countryside could connect via a Starlink antenna. link
Q106 When will Starship demonstrate orbital refilling?
Starship can go Mars surface to Earth surface by itself, but needs a massive booster plus orbital refilling to get to Mars. link
No, apart from orbital refilling, which he thinks is very important. Besides a fully and rapidly reusable rocket you need orbital refilling; that is fundamental. link
Sequence: fully reusable orbital rocket, then orbital refilling, then Mars. link
Docking with station is very hard; orbital refilling should in theory be easier because Starships dock with themselves and the first flights will be uncrewed. link
Q107 When will a city on Mars be self-sustaining?
Back-of-the-envelope, you need on the order of a million tons of useful payload to the surface of Mars — hopefully not 10 million, probably not 100,000. That means something like a megaton per year to Earth orbit or it is not relevant. link
About a dozen Earth–Mars transfer windows, so ~25 years, for the self-sustaining-city cadence. link
Building ~1000 Starships to create a self-sustaining city on Mars is the mission; production is hard, prototypes are easy. link
Q116 Should people work remotely?
Recruiting for Tesla's AI and chip teams, he said the main engineering locations are Palo Alto, Austin and Seattle, but remote work is acceptable for exceptional candidates. link
Q117 What's the long-term vision for Starlink?
Launching satellites taps out around $3 billion a year; providing broadband is more like an order of magnitude more, roughly $30 billion. Starlink will serve the 3–4% hardest-to-reach customers for telcos, not compete in high-density 5G cities. link
Q118 What's your algorithm for designing products?
He says the aim is to take cost out of the car without making it worse, by making it lighter and simpler. link
He says the rear third of the Model Y body is cast as a single piece, never done before at that size. link
He says the goal is to take cost out of the car while making the product better, not worse. link
He says Tesla puts far more engineering into the machine that makes the machine than into the car itself. link
He says manufacturers should be able to tell designers that a product is dumb to manufacture. link
He says Tesla loves manufacturing and more smart people should work on it. link
He says there is far more opportunity for innovation in manufacturing than in the product itself. link
He describes Tesla designing the machine that makes the product, and then making that machine itself. link
He asks what the physics limit or platonic ideal of a part is rather than comparing to the market. link
Q121 Should governments put a price on carbon?
Says a carbon tax is needed to correct the unpriced externality, and the market will do the rest. link
Q124 When will Dragon carry astronauts?
He called it hopefully the first step toward a civilization on Mars. link
Q125 What is Tesla's mission?
He notes Tesla has the highest-demand electric vehicle in the world with no advertising spend. link
He restates that Tesla's overarching goal is to help accelerate the advent of sustainable energy. link
He says Tesla's overall goal is to maximize the area under the curve of customer happiness. link
He says people would rather go to the dentist than buy a new car, so Tesla makes ordering easy. link
He says Tesla has no market share goals; accelerating sustainable energy is the fundamental good. link
He said physics favors electric transport, batteries for stationary storage and solar/wind for generation. link
He says Tesla will not succeed in its mission without making cars affordable. link
He says what bugs him most is that Tesla cars are not affordable enough. link
He says Tesla must not go bankrupt, but is not trying to be super profitable either. link
He says the 20 million vehicles a year goal comes from replacing 1% of the global fleet of about 2 billion vehicles. link
He says the best company will be the one that makes great products at an affordable price. link
He says 20 million vehicles a year is a good goal because it would replace 1% of the global fleet each year. link
Q128 Why did Tesla build the Cybertruck?
He says the Cybertruck goal was the most badass futuristic armored personnel carrier, one that beats any pickup truck. link
He says Cybertruck demand is unlike anything Tesla has seen and it will make as many as it can sell for years. link
He says the Cybertruck is better than people realize. link
Q129 Can the Boring Company solve traffic?
He says we should invest in the transportation of the future, not of the past. link
Says tunnels are the solution to traffic: you can have as many levels as you want, even in New York or Beijing. link
Q130 What is the Tesla Semi?
He says nickel-based cells have higher energy density and are needed for something like the Semi. link
Q131 Why does rocket reusability matter?
He said you want them reusable so you simply reload propellant and fly again, like aircraft. link
Q132 How fast is Tesla expanding Superchargers?
He said Tesla would put Superchargers at airports and wherever there is a need for them. link
Q134 Why does Tesla use cameras instead of lidar?
He says once you solve passive optical, you have solved self-driving. link
Q136 When will the Tesla Semi reach volume production?
In January 2020 he said the Semi uses so many battery cells that speeding up its production would mean making fewer Model 3s and Model Ys. link
Says the Battery Day technology affects long-term production, especially Semi, but will not reach serious high-volume production until 2022. link
In October 2020 he said battery cell supply was the only real constraint and had to be solved before the Semi could ramp to significant volume. link