Elon Musk's answers in 2019
Q1 How dangerous is AI and how do we make it go well?
He said there are things that will affect the future but he is not sure if they are good or bad, like AI. That one worries him. link
He said you could think of humanity as a biological bootloader for digital superintelligence: a tiny piece of code without which the computer cannot start. You could not evolve silicon circuits; they needed biology to get there. link
AI has potential to be dangerous to the public and should have a regulatory agency like other public dangers, though government usually waits until something terrible happens. link
Q5 Why go to Mars / what is SpaceX's mission?
The critical point for Starship/BFR: more payload than the biggest rocket ever made, for less money per launch than the smallest orbital rocket; without ~100X lower orbital cost, humanity will never be multiplanet. link
He said originally he was not going to start a rocket company. He really wanted us to become a spacefaring civilization. He wanted a philanthropic greenhouse on Mars to inspire the public so they would tell Congress to appropriate more money for NASA. His initial goal was to increase NASA's budget. link
He said reusability is fundamental. A fully reusable vehicle will cost a hundred times less per flight than an expendable one. Every other mode of transport is reusable; the weird one is space. It needs to be fully and rapidly reusable — that is the breakthrough for humanity to become multiplanetary. link
He said he actually doesn't care at all about money. He does care about us becoming a spacefaring civilization, and if we don't achieve full and rapid reusability it will not happen. That is the only reason he wants money at all. link
He thanked NASA: SpaceX would not be here without NASA, without the work done before SpaceX started and the support after it did. link
He said he really believes in the future of space and it is important that we become a spacefaring civilization. He wants the things in science fiction novels and movies to be real one day. link
The whole goal of SpaceX was crewed spaceflight. The company's full name is Space Exploration Technologies: improve space exploration technology to help make humanity a spacefaring civilization. After 17 years they still hadn't launched anyone; he hoped they would later that year. link
He said humanity landing on the Moon was maybe the greatest thing ever, and he could not believe they were launching from that pad. link
He said we should have a permanently occupied human base on the Moon and send people to Mars and build a city on Mars. link
After sustainable energy, he said we want to be a multi-planet species — rockets. link
He said it is important to take the set of actions most likely to continue consciousness into the future. One of those is to become a multi-planet species, not as an escape hatch or because Earth is doomed, but because there is an irreducible probability something happens to Earth. link
He said he is very pro-Earth. From a resource standpoint, less than 1% of Earth's resources should be dedicated to making life multiplanetary — somewhere between lipstick and healthcare, maybe half a percent to 1% of Earth GDP. link
Travel to Mars is only just possible given Earth's gravity; a little more g and it would not work. He also read Carl Sagan's pale blue dot passage. link
With respect to space he said there is really one problem: a fully and rapidly reusable orbital rocket. That is the holy grail. Bicycles, aircraft, cars and horses are all reusable; rockets still mostly are not. link
Q6 When will AI surpass humans?
He said people underestimate the capability of AI. They think it is a smart human, but it will be much more than that — much smarter than the smartest human, more like a chimpanzee trying to understand humans. link
He said computers are already much smarter than people on many dimensions; we keep moving the goalposts. There is a smaller and smaller corner of intellectual pursuits where humans are better, and soon we will be far surpassed in every single way — guaranteed, or civilizational end. link
Q7 Will people still need to work?
He said over time AI will make jobs kind of pointless. Probably the last job that will remain will be writing AI software, and then eventually the AI will write its own software. He would recommend studying engineering or physics, or working on something where people want to interact with other people. link
Q8 Why are your timelines so often late?
He said with exponentials, small changes in the calendar breakpoint make enormous percentage differences. link
Explains the middle of the manufacturing S-curve is extremely difficult to predict. link
He thought very quickly, maybe even toward the end of 2019 and he'd be shocked if not 2020 at the latest, having a human intervene would decrease safety, like elevator operators. link
He said that next year, for sure, there would be over a million robotaxis on the road. link
He said if you are going to try something innovative you are in unexplored territory, so the odds that something will go wrong are pretty high. It is only if you try something well understood that there is little chance of failure — but then it will not be interesting. Uncharted territory will result in failures. link
He acknowledged people were puzzled by his million robotaxis by end of next year prediction. link
He said if a design is taking too long, the design is wrong, and therefore the design must be modified to accelerate progress. One of the most fundamental errors in advanced development is sticking to a complicated design and not striving to delete parts and processes. link
Q11 What drives you / what is the meaning of life?
He cited Hitchhiker's Guide to the Galaxy: the Earth is a giant computer that answers 42. The really hard part is the question; the answer is the easy part. At the point you can properly frame the question, the answer is comparatively easy. link
He does not think consciousness permeates all matter; if you cannot test the hypothesis you cannot conclude it is true. link
He said he is mostly trying to figure out the set of actions that increase the probability that the future is good, and take those actions. link
Q12 What's the biggest remaining Starship problem?
He said the switch to steel was because the advanced carbon fiber was taking too long. The test for a design is whether it is in the future or not; if it is not, who cares. link
For the Starship program he said the cost of fuel and oxygen, pressurized without helium, is about $900,000 a flight. If the vehicle is fully reusable, even with other operating costs maybe around $2 million a flight, that is much less than even a tiny rocket. link
Starship fleet designed to fly 3 times per day, about 1000 flights per year, as the mind-blowing payload-to-orbit metric. link
Q14 What do you think of OpenAI?
Explains he parted ways with OpenAI partly due to Tesla competing for the same people and disagreements. link
He said OpenAI was intended to mitigate the risk of artificial general intelligence. He hoped that is what it does; there is some chance it may amplify the risk. link
Q15 How competitive is China?
He said the Tesla China team has done an amazing job, mind-blowing progress. He has never seen anything built so fast. He really thinks China is the future. link
He said China is actually the world leader in environmental sustainability, and he is not sure how well that is known outside China. He thought about half of all the electric cars in the world were made in China the previous year. link
Q16 How good is Tesla self-driving?
He guessed it would be safe to fall asleep and wake up at the destination probably toward the end of next year. link
He said Tesla expected to be feature complete with self-driving later that year. link
After about three years Tesla had its full self-driving computer, an order of magnitude more processing than the NVIDIA unit, redundant, with hardware already capable of full self-driving once software and regulators catch up. link
He said LiDAR is a fool's errand and anyone relying on it is doomed. link
He said he felt very confident predicting autonomous Tesla robotaxis the next year. link
Says when the LA/NY Autopilot journey happens this year, every FSD owner will be able to do it. link
He said the reason self-driving is hard is all the corner cases: the road should be a certain way but isn't — painted lines missing or offset. If you had clearly painted lines, self-driving would be trivial. The real trick is figuring out where you should be on a road that is not the way it is supposed to be. link
He said probably sometime next year the car could be autonomous without supervision. link
He can tell immediately: Autopilot has little nuances of movement and stays dead center in the lane, unlike human bumper-car weaving. link
Q18 Are we in a simulation?
He said one or two things are going to happen: civilization will end, or games will be so realistic you can't tell the difference between them and reality. We could be somebody's video game right now. link
He said AI will get to the point where it can completely simulate a person in every way, many people simultaneously. There is a strong argument we are in a simulation right now. If life were a video game the review would be: graphics incredible, plot confusing, respawn takes a long time. link
Q19 Is population decline a problem?
He is worried about the birth rate. Contrary to the view that there are too many people, he said that is outdated. Assuming a benevolent AI future, the biggest issue in 20 years will be population collapse, not explosion. Humans have a 20-year boot sequence. link
Q20 How hard should founders work?
He said he wants to get the most amount done possible; he takes breaks only to get more done. He would not recommend running two companies. Asked when he last had a work-free week, he said the year 2000, when he had malaria. link
Q22 When will Starship go to Mars?
He said there was probably a 50% chance of sending ships to Mars at the end of next year. link
Q30 How do you hire and manage people?
He said someone could join as a junior engineer at Tesla Shanghai and ultimately maybe have his job one day. link
A friend asked who the chief engineer of SpaceX is; he said it is him. What he tells the team is everyone is a chief engineer: everyone must understand, broadly, all the systems in the vehicle so you do not get subsystem optimization. link
He said if you want to advance technology you have to recruit the world's best engineers and create an environment that lets them be as innovative as possible — reward innovation and punish lack of it. link
He looks for evidence of exceptional ability, not college or high-school graduation. Give three bullet points of exceptional ability: an impressive device, a tough competition, a great idea. A 4.0 GPA can even be a contra-indicator. He worries whether Nikola Tesla alive today would get an interview. link
Q38 What is the limiting factor for AI scaling?
He said all you needed a GPU for was to help the game look better, and those same GPUs ended up being very useful — a huge boost to AI. Better graphics to AI, and that same path is key to self-driving. link
Q44 What is Neuralink's near-term goal?
He said the goal is to solve important spine and brain problems with a seamlessly implanted device. link
He said if you can't beat them, join them: that is what Neuralink is about — can we go along for the ride with AI. There should be other companies creating a high-bandwidth interface to the brain. link
Q45 Is there hidden alien or advanced propulsion technology?
He said we should not take it for granted that consciousness will continue because we have not encountered any aliens. Where are the aliens? People think we have found them; trust him, he would know — we have not. link
Q47 Has Tesla come close to failing?
When he started SpaceX he only thought there was maybe a ten percent chance of getting Falcon 1 to orbit. He did not think this would happen. He had been chief engineer since day one and did not really know anything about rockets, which is why the first three failed. He only had money for three flights. link
He said he always thought they would fail, so it was all upside. Starting out he thought maybe they had a 10% chance of reaching orbit. link
The first three Falcon 1 launches did not work. They scraped together parts for the fourth, and if that one hadn't worked, that would have been it for SpaceX. link
He said they blew up three rockets in the beginning. He put the PayPal money into SpaceX, Tesla and SolarCity, then had to give them all the money and borrow for rent in 2008. They scraped spare parts together for a fourth launch; if it had not worked, SpaceX would have died, and probably Tesla with it. link
Q53 What's your vision for energy abundance?
He said for the future to be good we have to have sustainable energy: sustainable production and consumption of energy, meaning low-cost solar panels and appealing electric cars. Then become a multi-planet species. link
He said humans will solve environmental sustainability. That is not a reason for complacency: we must take immediate and dramatic action and continue the momentum. link
Q59 Will Tesla own the robotaxi fleet?
Says Tesla will charge a percentage fee for cars in its robotaxi fleet. link
Says owners of unsupervised-FSD cars could earn more than their lease cost by letting others use them, and managing a small robotaxi fleet will be a career for many. link
Q91 What are your post-AGI plans? (asked on X)
Says neuro-silicon symbiosis isn't currently on track to arrive in time for AGI, and Neuralink's goal is to raise that probability above zero. link
Q101 Will Neuralink enable telepathy between humans? (asked on X)
He said achieving good AI symbiosis is important from an existential threat standpoint. link
He described an AI extension of yourself, a tertiary layer above the limbic system and cortex. link
Says a Neuralink could increase communication rate by several orders of magnitude. link
Q106 When will Starship demonstrate orbital refilling?
Orbital refilling is critical for high payload to Moon or Mars: Starship-to-Starship first, later dedicated tankers. link
Orbital refilling is vital; more likely spacecraft-to-spacecraft like aerial refueling than a dedicated depot at first. link
Optimistically toward the end of the next year (2020); he would be surprised if it took longer than two years. link
Q107 When will a city on Mars be self-sustaining?
Replying to a question that assumed 2022 cargo and 2024 cargo-and-crew tests: about 10 orbit-synch windows every 26 months, so self-sufficiency around 2050. link
He said this is a good step, getting people back in orbit. What he really wants to see is a permanently occupied human base on the Moon and us building a city on Mars. He can see the beginning of that; before he dies he will die happy. link
Possible to make a self-sustaining city on Mars by 2050 if they start in 5 years (from Mar 2019 ≈ 2024) and take 10 orbital synchronizations. link
Roughly a million tons on Mars; if any ingredient is missing and ships from Earth stop, the city dies out. link
The economics of Starship’s payload-to-orbit rate have to be in that league to build a self-sustaining city on Mars. link
Q112 Where do you get your news?
He said he generally tries not to read the daily news because newspapers seem to answer the question of what is the worst thing that happened on Earth today. He finds science and technology periodicals interesting, and Twitter enlightening at times. link
Q116 Should people work remotely?
He said remote work carries a big communication penalty, so the talent gained from people who won't move has to be very high. link
Q117 What's the long-term vision for Starlink?
He said one of the criteria for Starlink is that it is important for real-time gaming, and later second-generation constellations are aiming further. This would be really great for any kind of real-time gaming, anywhere on Earth. link
Q118 What's your algorithm for designing products?
He said take the approach that the constraints you are given are guaranteed to be some degree wrong. Question your constraints. It does not matter if the person handing them to you won a Nobel Prize. link
He said one of the biggest traps for smart engineers is optimizing a thing that shouldn't exist. link
He said he tries to propagate that everyone should be chief engineer: everyone should have at least a cursory understanding of the whole rocket or the whole car so they optimize the product as a whole. link
From physics: you should always assume that you are some degree wrong, and your goal is to be less wrong. People tend to assume they are right and must prove it. The question is how wrong, and can you be less wrong tomorrow. link
Q119 Should you love the product you're making?
He said if you are trying to figure out what others love but you don't love it, it is really hard to make that great. When you are working on something, if you fall in love with it, that's a good sign. link
He said people often try to make products they think others would love but they do not love themselves. If you don't love the product, you should not expect that others will. link
Q125 What is Tesla's mission?
He said Tesla always designs with safety as the number one goal. link
He said it is basically mad to buy a fossil fuel car at this point. link
He said his interest in electric cars went back to high school, from a physics view that we need something long-term sustainable. link
Q126 What did Elon promise for the Model 3?
Says the exact Model 3 in his profile pic costs $35k and takes a minute to buy online. link
Q128 Why did Tesla build the Cybertruck?
He said Tesla needed something manufacturable in volume with basically no corrosion and no need for paint. link
He said it is very rare that a product comes along that experts said was impossible. link
Q129 Can the Boring Company solve traffic?
Says efforts to solve big-city traffic haven’t worked; without trying something new, traffic hell forever. link
Q131 Why does rocket reusability matter?
Contrasts Falcon (~80% reusable, days between flights, needs boats) with Starship, planned to be fully reusable with rapid reflight. link
Says humanity will never be multiplanet without a fully and rapidly reusable orbital rocket. link
Q136 When will the Tesla Semi reach volume production?
Says Tesla has been mired in production and logistics and he is looking forward to getting Semi into production. link