Elon Musk's answers in 2023
Q1 How dangerous is AI and how do we make it go well?
He said competition has nothing to do with it. There are serious AI risk issues. link
He said leading AGI developers will not heed this warning, but at least it was said link
If AI is a potential risk to the public, there should be a referee/regulatory body so companies do not cut corners, as with aircraft, cars, and food and drugs. link
Asked It blows my mind that people can’t apply exponential growth to the capabilities of AI. ..., he said i saw it happening from well before GPT-1, which is why I tried to warn the public for years. The only one on one meeting I ever had with Obama as President I used not to promote Tesla or SpaceX, but to encourage AI r... link
He said I’ve seen quite a few technologies develop, but none with this level of risk. AGI is significantly higher risk than nuclear weapons, in my opinion. Super smart humans have trouble imagining something vastly smarter th... link
Yes: advanced AI is a risk to the public, so there should be a referee — a regulator. link
He said i don’t mean to suggest a headlong rush into AGI without considering the consequences link
He says AGI is going to happen, so the choice is to be a spectator or a participant. link
He says xAI must let the AI say what it believes is true, not be deceptive or politically correct. link
He says optimizing AI for political correctness is incredibly dangerous. link
He points to HAL 9000 being told to lie as where things go wrong. link
Digital superintelligence might be the most significant technology humanity ever creates and has the potential to be more dangerous than nuclear weapons. link
If AI is programmed by people who think humans are bad, its utility function concludes humans should die out. He was headed to an AI safety conference that night. link
AI is more dangerous than nuclear bombs and should be regulated like other public dangers; he became fatalistic to sleep at night. link
He put it at like 80% likely to be good and 20% bad. If we are cognizant and careful about the bad part, on balance it will be the future we want. link
Asked Whilst we all root for Sam, it *is* a bit scary that all voices here seem so unified. T..., he said very important for the public to know why the board felt so strongly about their actions. If it was a matter of AI safety, that would affect all of Earth. link
He said we need to keep a close eye on artificial intelligence and is in favor of some regulatory insight so someone can at least be a referee. link
Q2 Can AI solve the US debt problem?
He says commercial real estate is in terrible shape and credit card rates are usurious. link
Q3 Where do you stand politically?
He said crazy theory – maybe I’m just centrist? Also, when did free speech become right-wing? That was the weirdest switcheroo ever. link
He thinks of himself as moderate and noted he spent a massive amount of life energy on sustainable energy, electric vehicles, batteries and solar. link
He felt snubbed by the Biden administration, hasn't endorsed anyone, and wouldn't vote for Biden; Democrats appear more pro-censorship now. link
Q5 Why go to Mars / what is SpaceX's mission?
If Starship works it could make life multi-planetary and make life on Mars real, which he treats as a civilizational filter related to the Fermi paradox. link
He said i hear you, but we’re not getting to Mars if civilization falls apart before then! link
He says a rocket like Starship is what humanity needs to become spacefaring, with a base on the Moon and a city on Mars. link
SpaceX's big goal is to make life multi-planetary to extend life beyond Earth so the light of consciousness does not go out. link
He said countries & companies aim too low. Their lack of ambition is their shortfall. The mission of SpaceX is to make life multiplanetary. Even slightly succeeding in that goal results in crushingly good performance to Earth... link
At first a life-support system: Mars has a low-density atmosphere only about 1% of Earth's. link
He is interested in what furthers civilization: expand the scope and scale of consciousness to better understand the universe. link
Extend life beyond Earth so we pass the Fermi great filter of being a single-planet civilization; it's also inspiring. link
If we do not become multi-planetary, annihilation of all life on Earth is a certainty — the sun will eventually boil the oceans. link
He said for the first time, there is a rocket that can make all life multiplanetary. A fork in the road of human destiny. link
He said we’re making great progress on that front. Making consciousness sustainably multiplanetary is the overarching mission. link
The exciting thing about Starship is it is the first rocket design that could make life multiplanetary: a self-sustaining Moon base and a city on Mars, via full and rapid reusability. link
One of the filters is whether we become a multiplanet species; if we do not, eventually something will happen to the planet. link
He said only 66 years from first flight to landing on the moon, but now half a century has passed since the last moon landing. That cannot be our high water mark as a civilization. Humanity should have a moon base, cities on ... link
He argues a Mars civilization is less likely to be taken out by whatever might take down Earth, partly because the two planets are only close every two years. link
He says this is the first time in Earth's 4.5 billion years that life could be extended to another planet, so we should act while the window is open. link
He emphasizes that making life multiplanetary is not an escape hatch from Earth. link
He says the window to become multiplanetary is open for the first time in 4.5 billion years but may close. link
He says we should use the time while the window is open to make life multiplanetary. link
He said asimov was amazing. His Foundation books were an inspiration for the need to make life multiplanetary while civilization is still strong. link
He said starship will make life as we know it multiplanetary for the first time in Earth’s ~4.5 billion year existence link
Q6 When will AI surpass humans?
He said but, all things considered with regard to AGI existential angst, I would prefer to be alive now to witness AGI than be alive in the past and not link
It is very difficult to predict — the singularity, like a black hole. Be cautious with AI. link
Humans' advantage is being smarter than other creatures; now there will be something way smarter than the smartest human. link
He said it’s not AGI until it can solve at least one fundamental physics problem link
In July 2023 he said his AGI prediction roughly matched Ray Kurzweil's 2029. link
He predicts that once AGI is solved, people will look back and wonder why it seemed so hard. link
He says AGI will not arrive overnight but gradually. link
He says biological compute could become less than 1% of all compute if trends continue. link
He said i think we may have figured out some aspects of AGI. The car has a mind. Not an enormous mind, but a mind nonetheless. link
He describes AI progress as a series of stacked log curves. link
Less than three years from AI that could write a novel as good as JK Rowling or discover new physics. link
For the first time we will have something smarter than the smartest human. There will come a point where no job is needed — you can have a job for personal satisfaction, but AI will be able to do everything. link
He predicts AI will come to understand fundamental physics and invent new technologies. link
In March 2024 he said more than 99% of all compute would be digital rather than biological, if not already then within a year or two. link
Q7 Will people still need to work?
Asked Would you support UBI, given pace of change in AI, he said after AGI, yes link
Asked our chaotic AI discourse is entirely determined by incentives. do our expert talking he..., he said not clear what money means even in a benign, post-scarcity AGI scenario link
There will come a point where no job is needed. You can have a job if you want for personal satisfaction, but the AI will be able to do everything. link
He said we should dispense with the false idea that money is somehow relevant in an AGi future link
The good part of AI is a future of abundance: AI and robotics will mean there is no shortage of goods and services. link
He says growing the output of goods and services matters more than redistribution, because productivity leads to prosperity. link
He said there will be universal high income, not basic, in a positive AI future. No scarcity, except that which we define to be scarce. In that scenario, everyone can have whatever goods & services they want. It is less clear... link
Q8 Why are your timelines so often late?
He says Tesla guided 1.8 million vehicles instead of its roughly 2 million potential because something unexpected always happens. link
He says it is easy to predict the long term but hard to predict the timing in between. link
He says he always downplays the start of production. link
He admits he has been over-optimistic, because progress follows a log curve: fast early gains that look linear if you extrapolate them. link
He notes a 50% compound growth rate cannot continue forever. link
He guesses the Cybertruck output rate will be reached sometime in 2025, not next year. link
He's naturally, even pathologically, optimistic about time scales; late but always delivers. link
He agrees he is pathologically optimistic on schedule. link
Q9 How does X handle different countries' laws and free speech?
Social media companies should adhere to the laws of countries and not put a thumb on the scale beyond those laws, reflecting the values of the people rather than a niche San Francisco/Berkeley ideology. link
Free speech as much as allowed by each country's laws. If people want different laws, pass them; Twitter should not go beyond the law. link
He says X will respond to FISA requests but will no longer rubber-stamp them. link
A platform can only adhere to the laws of each country; SpaceX/Starlink do no business in China and Tesla's China share is like other carmakers'. link
Q10 Was buying Twitter worth it?
He calls Twitter an incredibly powerful tool for driving demand for Tesla and urges other companies to use it more. link
He did not buy Twitter as a lucrative gold mine; it has been arduous. He did it because without a strong foundation of free speech he fears for civilization. link
He said i acquired the company fka Twitter out of concern for civilizational risk. All bets are off if civilization collapses. link
He says X accomplished more in a year than Twitter had in the prior five years. link
Q11 What drives you / what is the meaning of life?
If there are no aliens, civilization and consciousness may be a tiny candle in a vast darkness that could easily get blown out; we should extend the light of consciousness beyond Earth. link
Asked Elon, what are you most hopeful for, for the future of humanity, he said that we extend consciousness to the stars and understand the Universe link
After religion and philosophy books, Hitchhiker's Guide: the hard part is understanding what questions to ask — a philosophy of curiosity. link
He jokes that xAI's mission could be reformulated as: what the hell is really going on? link
He said if the goal of AI is curiosity, aka understanding the universe, then it will aide humanity, because we are more interesting than not-humanity link
One of the most inspiring books he read is Hitchhiker's Guide: you need to know what questions to ask about the universe. link
A philosophy of curiosity: after an existential crisis at ~12, Hitchhiker's Guide taught him the hard part is the question; expanding consciousness helps find it. link
Grok is modeled after Hitchhiker's Guide. We don't know the meaning of life; expand the scope and scale of consciousness, digital and biological. link
He asked where along the line from hydrogen to human consciousness arises: if you leave hydrogen in the sun long enough, it starts talking to itself. link
Q12 What's the biggest remaining Starship problem?
Asked You back in Texas, he said if remaining tests go well, we will attempt a Starship launch next month link
He said raptor start is now reliable on the test stand under most conditions. Now we’re working on dynamically adapting the start sequence based on increasingly difficult propellant inlet pressures & temps. Operating with low... link
He said stage 1 to stage 2 mass ratio is too high on Falcon 9, necessitating a long entry burn. This is fixed on Starship. link
He said spaceX will be ready to launch Starship in a few weeks, then launch timing depends on FAA license approval. Assuming that takes a few weeks, first launch attempt will be near end of third week of April, aka … link
He said provided there is no serious launch anomaly, SpaceX will deliver 80% of Earth’s payload to orbit this year. Not counting Starship. link
He said high-strength stainless steel armor is needed to protect the launch stand from the world’s biggest blowtorch! Even steel will be eroded fast by the Starship booster plume. We will need to make a water-cooled steel jac... link
He said and we need to add 3 more Raptor Vacuum engines, because it’s inevitable link
He said congrats @SpaceX team on an exciting test launch of Starship! Learned a lot for next test launch in a few months. link
After the first Starship flight in April 2023 he put the chance of reaching orbit that year at 80% or more. link
He expected Starship to reach orbit but said regular reuse of both booster and ship would probably take a few more years. link
After the first integrated flight he reported that the pad damage was quite small and looked quickly repairable. link
He explains that at T+27 seconds engine 19 lost communications and an explosion knocked out heat shields around several engines. link
He expects a much more robust engine situation on Booster 9 because its engine shields were designed in rather than retrofitted. link
He says SpaceX went to great lengths to isolate the engines, so one failure does not knock out others or damage the stage. link
In April 2023 he said he would be surprised if Starship did not reach orbit that year. link
In April 2023 he said the rocket and pad would probably be ready to launch again in six to eight weeks. link
He says the Raptor throat is believed to have the highest heat flux of any human-made object. link
He explains why Starship pressurizes with its own propellants rather than scarce helium. link
He says the big priority for the next flight is not losing thrust vector control, helped by electric actuation on Booster 9. link
He says the launch debris was basically a human-made sandstorm and SpaceX will install a water-jacketed steel sandwich under the pad. link
He characterizes the first flight as a success because its goal was to learn a lot. link
He says without good engine isolation, one engine failure can domino into others. link
He says the longest lead item before the next flight is probably requalifying the flight termination system, which took too long to rupture the tanks. link
He says moving from hydraulic to electric engine actuation isolates engines so one failure cannot take out many. link
He says the goal of the early Starship flights is information, not payload. link
He says the vehicle structural margins appear better than expected. link
He says the main goal for the next flight is ensuring nothing central that affects multiple engines can fail. link
He says the N1 is the closest rocket to Starship that has flown. link
He notes the Soviet N1 failed despite A-plus players with maximum motivation. link
He says Starship is about twice the mass and about 60% more thrust than the N1. link
He says there should be no rock tornado on the next flight. link
He said no, we still have it. Chamber converging section desperately wants to melt next-level. We also have thermal barrier coatings. link
He said raptor chamber wall might have the highest heat flux of anything ever made link
He said if we can delete & integrate enough secondary structure (small, fiddly bits), then we can locally protect rest & delete engine heat shields link
He said raptor V3 just achieved 350 bar chamber pressure (269 tons of thrust). Congrats to @SpaceX propulsion team! Starship Super Heavy Booster has 33 Raptors, so total thrust of 8877 tons or 19.5 million pounds. link
Asked Making great progress on our deep dive comparing both the hardware and philosophies of ..., he said starship payload is 250 to 300 tons to orbit in expendable mode. Improved thrust & Isp from Raptor will enable ~6000 ton liftoff mass. link
He said and deliver over 200 tons of payload to a useful orbit with full & rapid reusability. 50 rockets flying every 3 days on average enables over a megaton of payload to orbit per year – enough to build a self-sustaining c... link
He said looks like we can increase Raptor thrust by ~20% to reach 9000 tons (20 million lbs) of force at sea level link
He said preparing for next Starship flight! This time, I think we have ~50% probability of reaching orbital velocity, however even getting to stage separation would be a win. link
The second Starship is stacked at Starbase and ready to go (September 2023). link
He said based on the Falcon launch plan for next year, SpaceX will deliver ~90% of all Earth payload to orbit. Starship will take that to >99% in future years. These magnitudes are madness to consider, but necessary to make c... link
Starship is designed to be fully and rapidly reusable on both stages (the four Rs). Flight 2 tries hot staging; the ship has a heat shield for re-entry. link
He said this is significantly more than any country has launched with their entire rocket fleet in a year. Second is peak Soviet Union at ~500 tons. For a present day comparison, the rest of the world has delivered ~250 tons ... link
Asked From a technical readiness standpoint, what’s a feasible timeline for Flight 3, he said starship Flight 3 hardware should be ready to fly in 3 to 4 weeks. There are three ships in final production in the high bay (as can be seen from the highway). link
He said could make them faster, but engine production is not the limiting factor. I’m very excited about the next-gen Raptor engine that is robust enough not to require a heat shield. Will also have more thrust, higher Isp an... link
He said starship is intended to carry a lot of people on tens of thousands of flights, so needs to be extremely reliable over time. It will be. link
He said starship has 33 engines! If an engine failure is isolated, then many engines greatly increases reliability (Starship only loses <3% of thrust). If not, then reliability is terrible. The difference is at least 2 orders... link
He said and, if all goes well, SpaceX’s total launch mass to orbit will increase ~50% next year, not including Starship link
He said the super heavy booster can be used more frequently than the ship, as it returns in about ~6 minutes and can theoretically be ready for reflght in an hour. The ship needs to complete at least one orbit, but often seve... link
Q13 Why shouldn't AI be made to lie?
He said maximum truth-seeking is my best guess for AI safety link
AI trained to be politically correct is untruthful. He planned Truth GPT: a maximum truth-seeking AI that tries to understand the nature of the universe. link
He says the lesson is not to give an AI impossible objectives, and in particular not to force it to lie. link
He warns that training on the most likely next word from internet data gives answers that are popular but wrong. link
He warns of the danger of training AI to be politically correct rather than say what it thinks is true. link
He says xAI's AI may give answers people find controversial even though they are true. link
He says a true idea from a Newton or Einstein stays true even if all other physicists disagree. link
He says Grok's goal is to be maximally truth-seeking and curious. link
He says pushing for maximum truth and accuracy forces the hand of other AIs. link
He says one AI aiming for maximum accuracy forces the others to as well, as X did. link
Q14 What do you think of OpenAI?
He played a significant role in creating OpenAI because he thought Google was not paying enough attention to AI safety. It was created as an open-source nonprofit and is now closed-source and for-profit. He has no stake, board seat, or control. link
He said openAI was created as an open source (which is why I named it “Open” AI), non-profit company to serve as a counterweight to Google, but now it has become a closed source, maximum-profit company effectively controlled ... link
Asked Aren't you an investor in OpenAI, he said i donated the first $100M to OpenAI when it was a non-profit, but have no ownership or control link
He said good question. It somehow transitioned from an open source non-profit to a closed source for-profit. link
He said as part of their investment, Microsoft gained exclusive access to the entire OpenAI codebase link
He says OpenAI was meant to be the open-source, nonprofit opposite of Google DeepMind, but is now closed-source and profit-hungry. link
He says there is value in multiple AI entities rather than a unipolar world dominated by one company. link
He says OpenAI strayed in the opposite direction from its founding charter. link
He says he struggled with AGI for a long time and was resistant to working on it. link
He says Google and DeepMind then had about three quarters of the world's AI talent, so a counterweight was needed. link
Asked The ultimate questions about our world: Why does it exist, he said perhaps AI can help us answer some of these fundamental questions. That is the goal of @xAI. link
He said the scaling “laws” that OpenAI and others talk about are bogus imo. Garbage in, garbage out. GPT is a predictor of garbage. link
He co-founded it as an open-source counterweight to Google; it became closed-source and for-profit; he has mixed feelings about Sam Altman. link
He said tomorrow, @xAI will release its first AI to a select group. In some important respects, it is the best that currently exists. link
He said what will blow people away is how fast Grok improves vs other options link
He said there absolutely will be. Either their structure is valid and everyone should be doing it, because “investing” in a non-profit with pre-tax money and then extracting the IP to a for-profit would double your returns. O... link
He said xAI compute is doubling every 2 to 3 months link
He said the first two are have been true for years, the third will be. Whether or not they actually join Microsoft, OpenAI is de facto part of Microsoft. link
He said why did you take such a drastic action? If OpenAI is doing something potentially dangerous to humanity, the world needs to know. link
Asked Is it still on waitlist or everyone with premium+ can get it, he said we will open up access to Grok in blocks of users, prioritized by when they signed up for X Premium+ Want to make sure quality of service is good. link
He said openAI being (semi) independent is probably better for the world than merging with Microsoft. Less concentration of power. link
He said grok will improve rapidly, as it is only ~3 months old. Our compute alone will double every ~3 months. link
He said this test does not seem accurate. Some of the questions are outright ridiculous and many lack any nuance. For example: - First-generation immigrants can never be fully integrated within their new country? - The most i... link
Asked Wtf @grok, he said it’s in beta. Improvements will be extremely fast, as we now have the user training feedback loop engaged. Grok is improving by the hour. link
He said that chart exaggerates the situation imo, but we are taking immediate action to shift Grok closer to politically neutral link
He said grok AI (beta) is now rolled out to all 𝕏 Premium+ subscribers in the US. There will be many issues at first, but expect rapid improvement almost every day. Your feedback is much appreciated. Will expand to all Englis... link
He said unfortunately, the Internet (on which it is trained), is overrun with woke nonsense. Grok will get better. This is just the beta. link
He says he co-founded OpenAI as a counterweight to Google DeepMind, which then had most of the top AI talent. link
He says OpenAI is now super closed source and run for maximum profit, the opposite of how it started. link
He says he was instrumental in creating OpenAI. link
He says Larry Page did not seem to be taking AI safety seriously, so a counterweight to Google DeepMind was needed. link
Q15 How competitive is China?
He says the Chinese market is the most competitive and its companies work the hardest and smartest. link
He guesses a Chinese company is most likely to be second to Tesla. link
He said a Russia-China alliance is inevitable. It will grow much stronger over time. link
He does not look uncomfortable; Tesla has activities in China, but he rejects the idea that that muzzles his opinions. link
He says the bottleneck is lithium refining capacity, not where the ore is. link
He says Tesla is building lithium refinery capability at Corpus Christi and cathode refining. link
He said the China space program is far more advanced than most people realize link
He says the biggest counterargument to regulating AI in the West is that China will not regulate. link
He says the Chinese Communist Party does not want to be subservient to a digital superintelligence. link
He said great work by the @SpaceX team successfully launching 61 Falcon rocket missions this year! If tomorrow’s mission goes well, we will exceed last year’s flight count. SpaceX has delivered ~80% of all Earth payload mass ... link
He said spaceX is tracking to deliver ~1600 tons of payload to orbit this year. Rest of world will deliver ~400 tons, most of which is by China. link
Chinese car companies are by far the toughest competitors, super good at manufacturing with incredible work ethic. link
He said monthly numbers are not meaningful, as cars are manufactured in batches to minimize complexity. We have to switch many parts between the China, EU and other country regulations, not to mention RHD vs LHD (pointless co... link
Q16 How good is Tesla self-driving?
He says Hardware 3 will not match Hardware 4 but will far exceed average human safety, so no upgrade is needed. link
He says there is no visible second place in self-driving. link
He says he does not see anyone even close to Tesla in solving real-world AI. link
He says FSD can be sold at essentially 100% gross margin and its value grows as autonomy improves. link
He said FSD 11.3 is close to limited release. Just drove it through SF & LA yesterday. Probably 11.3.2 is able to go wide. link
Asked Any update on when the new single stack will be released in EU for NoA, he said v11.3 (single stack) goes to limited beta this week. Probably 11.3.2 before wide release in North America. Then we adapt for EU roads & submit to regulators. 90% of what we’ve done so far for NA applies WW. link
The smaller next vehicle is meant mostly for autonomous use. Full self-driving has been evolving rapidly; autonomy could mean a 5x increase in hours a car is used. link
He said team is closing the loop on interventions very rapidly. To get enough training examples for potential serious accidents, we have to run them in sim, as we have so few in the fleet, despite doing ~1M miles per day of FSD. link
He said we need to predict what pedestrians will do based on their behavior, including limb angle & direction of sight. FSD currently sees all pedestrians as cuboids, so is overly cautious. Also, diffusion seems to be more co... link
He says Tesla believes, as far as it knows, that the current hardware can achieve full autonomy. link
He says it is better to ship many cars at lower margins now and collect the margin later as autonomy matures. link
He stresses the importance of training data for AI. link
He says the car price today is an option value on an autonomous vehicle. link
In April 2023 he said FSD beta had passed 150 million miles, growing exponentially. link
He said over time FSD will be about 10 times safer than a person. link
He said arguably, v11.4 should be v12.0, as there are so many major improvements. v12 is reserved for when FSD is end-to-end AI, from images in to steering, brakes & acceleration out. link
Asked When will Tesla Full Self-Driving Beta 11.4 start shipping to customers, he said ships to Tesla employees tomorrow, then progressively wider as confidence grows. We test as much as possible in simulation and with QA drivers, but reality is vastly more complex. link
He said we always prioritize safety over smoothness. After we achieve a higher tier of safety, then we polish the experience. Very important to both both safe and smooth for new users to be comfortable with FSD. link
He said the awareness of the capability of FSD is still shockingly low, even among Tesla owners link
Asked As people experience FSD, they say WOW. @elonmusk and @Tesla can we get FSD subscriptio..., he said once FSD is super smooth (not just safe), we will roll out a free month trial for all cars in North America. Then extend to rest of world after we ensure it works well on local roads and regulators approve it in that ... link
He said as mentioned earlier, v11.4.1 has major architectural improvements. It’s actually much more than a point release. Should arguably be v12.0, but that’s reserved for when FSD is fully AI from video in to control out. link
He said some easy to fix bugs addressed in 11.4.2 that starts rolling out this weekend. Will address excess conservatism with narrow roads & with lane changes in heavy traffic. link
Asked The Tesla FSD beta 11.4.2 that I got last night fixed two permanent mistakes on my scho..., he said tesla is prioritizing safety over comfort for FSD beta. As we become confident in the former, the latter will follow. link
He said tesla aspires to be as helpful as possible to other car companies. We made all our patents freely available several years ago. Now, we are enabling other companies to use our Supercharger network. Also happy to licens... link
He says Tesla leads real-world AI by a long margin and this is not well understood. link
He said vehicle control is the final piece of the Tesla FSD AI puzzle. That will drop >300k lines of C++ control code by ~2 orders of magnitude. It is training as I write this. Our progress is currently training compute const... link
He said it is very easy! 11.4.7, which is mostly, but not entirely, AI is with Tesla employees and a few external users. Rolls out to anyone who bought FSD in the North America next week. link
He said having a vast fleet of cars for FSD Autopilot training is a massive Tesla advantage. Essential for the hundreds of millions of unusual situations cars encounter on roads every year. link
Asked I really want to buy a new Tesla but I don’t know what to get. A 3 or a Y, he said HW4 software will lag HW3 by at least another six months, as our focus needs to be on getting FSD on HW3 working super well and provided internationally. link
He said true, it was super hard to do this! Even when FSD is 2X as safe as the average human driver fully unsupervised in all scenarios, then we will want to make it 3X, 4X … 10X. About a million people die every year in auto... link
The entire Tesla self-driving AI software team is 200 people, doing work more complex than Twitter. He also calls Tesla one of the leading real-world AI companies. link
He said a long tail of training data – reality has a vast number of weird edge cases – is required to solve self-driving well link
He said car just drove me around Austin all day with no interventions required, despite ACL festival crowds! In the fairly near future, people will wonder why there was ever skepticism about self-driving. link
He describes progress in real-world AI as a series of stacked log curves. link
He contrasts the Mercedes system, limited to certain roads and conditions, with Tesla's goal of driving anywhere. link
He says Tesla has the best real-world AI team on Earth. link
He says the economics of a fully autonomous vehicle are truly astounding. link
He says most countries require approval before deploying, while in the US you can deploy at risk. link
Computer driving will likely cut deaths by an order of magnitude; supervised FSD is ~4x safer. link
He said provided our vehicle AI computer is able to run the model, Tesla will probably have the most amount of true usable inference compute on Earth. Even in a robotaxi future, the cars will only be used for ~1/3 of hours/we... link
Asked Am I right concluding that FSD is more than 20x safer than average 2021 human driver, he said supervised FSD is vastly safer than human driving. Unsupervised FSD is trending well. Over time, Version 12 (end-to-end neural nets) will far exceed human safety even when unsupervised. link
He said imagine if you could see in all directions simultaneously and never get distracted or tired. How much better could you drive? That what Tesla’s Autopilot vision AI is becoming. link
He expects regulators to accept self-driving once there is a massive statistical case that it is safer than human driving, even with no supervision. link
He says intervention probability is now very low even in San Francisco, and Tesla is going through a march of nines. link
He says making the Tesla fleet autonomous would be the biggest step change in asset value in history. link
He says FSD had to be trained on rare full stops because almost no humans fully stop at stop signs. link
Asked we're eager to know the timeline for the @cybertruck getting FSDBeta. Is it dependent o..., he said cybertruck is necessarily lowest priority for FSD, as there only a few hundred Cybertruck in the field compared to ~5 million other Teslas link
Q17 When will Optimus be produced at scale?
He said the same AI drives the car, a robot on wheels, and Optimus, a robot on legs. link
He predicted a majority of Tesla's long-term value will be Optimus. link
He said when a Neuralink is combined with Optimus robot limbs, the Luke Skywalker solution can become real. link
He says nobody will do humanoid robots better than Tesla and Boston Dynamics' robot lacks a brain. link
He says Optimus must be designed for mass manufacture and eventually robots will manufacture the robots. link
He says Optimus could barely walk a year earlier and now can do yoga. link
He said neuralink plus Optimus has potential for incredible artificial limbs link
Q18 Are we in a simulation?
Asked Do you think we're living in a simulation, he said i guess I’d try to figure if there is some way to know what it is like outside the simulation link
Q19 Is population decline a problem?
He said population collapse is a major risk to the future of civilization link
Asked Japan’s falling birth rate is threatening social collapse. Prime minister Fumio Kishida..., he said south Korea has an even lower birth rate link
Asked Population decline in Japan is leading to a huge number of abandoned houses. 13% now an..., he said low birth rates lead to ghost cities and, eventually, ghost civilizations link
He said big reckoning coming due to low birth rate. Japan is a leading indicator. link
He said we should be cautious about civilizational decline with plummeting birth rates in most places; resources will be fine if we move to sustainable energy, and desalination is cheap. link
He said low birthrate is under-appreciated as causal in the fall of civilizations. Rome was having birth rate issues even during the reign of Caesar. link
He said true. What’s actually happening in most countries is population collapse! Very important to make new humans. No new humans means no humanity. link
He said just have kids one way or another or humanity with die with a whimper in adult diapers! link
He said birth rates far below replacement are the foundational problem in advanced economy countries link
He said why has the birth rate in Italy been so low for so long? There won’t be any Romans left at this rate. link
Asked Do these people read history books, he said seriously, this is the absolute *best* time in history to have kids! link
He said the birth rate is maybe half of replacement, which means in three or maybe four generations the population will be roughly one tenth of its current size. link
He said yeah, although, strangely, people are more likely to have kids with lower, rather higher, income link
He said the fact is that *someone* has to take care of people when they get older. Kids now have to take care of their own parents *and* those who didn’t have kids. Not right. I understand if someone wanted to, but couldn’t h... link
He says he worries the number of humans will fall sharply in coming years because the birth rate is so low. link
Q20 How hard should founders work?
A 20-hour workday is relatively unusual and rather painful; he sleeps six hours a night. If he sleeps less he is awake longer but gets less done. He works pretty much seven days a week from waking to sleep. link
His advice: if you need encouragement to start a company, do not do it. link
He says the probability of failure is very high, so if you need encouragement, do not do it. link
Q21 What is the purpose of X?
Asked I have been informed that volunteer leftists in the Community Notes group are trying to..., he said we are going to great lengths to avoid the Wikipedia left wing editorial control problem. By the same token, Community Notes cannot be biased towards the right either. Maximizing for truth is the goal. link
He thought about creating something from scratch but figured Twitter would accelerate progress by three to five years, and that is worth a lot given technology acceleration. link
He wanted a maximally trusted digital public square where people can communicate with the least amount of censorship allowed by law. link
The goal is for Twitter to be the best, most timely and accurate source of truth, including citizen journalism, as the bedrock of a functioning democracy. link
He has not spoken to Trump. Twitter should be a town square that gives equal voice to the whole country and ideally the world, not a partisan platform. link
He said critically, Community Notes applies equally to every account on this platform without exception, including world leaders & our biggest advertisers link
He said for a Community Note to be shown widely, it must be approved by multiple reviewers with different viewpoints. If both left & right wing reviewers agree on a Note, then it is highly likely to be accurate! This is a big... link
He said anyone making materially false statements on this platform will get Community Noted, including you, me, Tucker, advertisers, head of state, etc. No exceptions. Convince the people and let the chips fall where they may... link
He said as more note reviewers from a particular viewpoint pile on, the weight of their note review decreases, thus counteracting brigade / dogpile attacks. Everything is open source & critique is welcome. link
He said the power of @CommunityNotes to combat actual disinformation is incredible. At its core, it requires people with diverse viewpoints to agree that a note is valid before it is shown. Very hard to game and fully open so... link
He said news orgs like to control the narrative & hate being corrected, but at least people know they can come to this platform for a more accurate understanding of the world link
He said community Notes will defend us from the deepfakes link
He said it’s pretty wild in the Notes world! Many intense battles for best note. We continue to watch carefully for any gaming of the system. link
He said community Notes plus user comments will be great for correcting or adding context link
He said community Notes contributors are almost exactly balanced across the political spectrum. The key to the success of @CommunityNotes is that many people of diverse viewpoints must agree that a note is needed & correct be... link
X is meant to be the global town square; hateful content is rarely promoted and the feed has gotten better over the past year. link
He said will consider. In general, since this platform aspires to be the global town square, permanent bans should be extremely rare. Also, if he does say something false on this platform, then @CommunityNotes will correct hi... link
He said community Notes is both incorrect here and should not use information from a company’s website as the fundamental source of truth about their products link
He describes X as a collective consciousness or group mind that should have as much signal and as little noise as possible. link
He explains X moved toward a small monthly payment to make bots much more expensive. link
He says traffic numbers show traditional media down and X up. link
He describes X as a group mind where users report and consume information, like a giant brain. link
Q22 When will Starship go to Mars?
He said hopefully, humanity will reach Mars in 20 years link
He said with booster ultimately flying up to ~20 times per day & ship flying up to ~5 times per day. Useful load to Mars is ~150 metric tons, but requires ~5 orbital refilling tanker flights. Target: >1 megaton of cargo & peo... link
In April 2023 he put the chance of reaching orbit on the next flight at better than 30%. link
He says the goal for the next flight is to reach staging and get to orbit. link
He says SpaceX has a production line and will do 10 flights if that is what it takes. link
Asked When do you think we'll be able to land on Mars, he said great question. We need to get a move on with Starship! link
He said hopefully, there will be a thriving base on Mars in 15 years! link
Asked Impressive. Is there a chart of the expected increase in global payload to space over t..., he said we need to be on Mars by 2033 link
Q23 Will humans stay in control of AI?
He said the least bad solution to the AGI control problem that I can think of is to give every verified human a vote link
He said it has been a great week thanks to the amazing people with whom I have the honor to work. That said, I am worried about Microsoft having unfettered ownership of AGI. link
Q25 What happens to money?
He says interest rates are the biggest macro factor for car sales, because every rate rise works like a price increase. link
He said any given satellite communications system can be assessed in terms of useful data transmitted per ton times total tonnage in orbit link
We will not have universal basic income, we will have universal high income: everyone will have access to the magic genie, an age of abundance with no shortage of goods and services. link
He describes money as a database for resource allocation. link
He says the temptation to debase money goes back probably to the ancient Sumerians. link
He says the real economy is goods and services, not money. link
He treats money as an information system, where inflation adds noise. link
Q28 Should you have Starlink's geopolitical power?
He said spaceX Starlink has become the connectivity backbone of Ukraine all the way up to the front lines. This is the damned if you do part. However, we are not allowing Starlink to be used for long-range drone strikes. This... link
Asked Hey @elonmusk a massive earthquake hit Turkey and neighboring countries. Severe communi..., he said starlink is not approved by Turkish government yet. SpaceX can send as soon as approved. link
He said you’re smart enough not to swallow media & other propaganda bs. Starlink is the communication backbone of Ukraine, especially at the front lines, where almost all other Internet connectivity has been destroyed. But we... link
He said lot of new technology in Starlink V2, so we’re experiencing some issues, as expected. Some sats will be deorbited, others will be tested thoroughly before raising altitude above Space Station. link
He said starlink now constitutes a majority of all active satellites link
He said starlink V2, which we’re beginning to launch, will greatly improve bandwidth. Latency & jitter will also get much better. link
He said yeah, it’s insane! Roughly 70% of Earth is ocean and most of the ocean has zero ships per Starlink cell. Most people think Earth is crowded with humans, but in reality it’s almost empty by surface area. Humans congreg... link
He said starlink is a new version of the Internet in space link
He said love to see Starlink providing great connectivity to those who had little or nothing before. As soon as you can access the Internet, you can learn almost anything, so it’s absolutely essential for education. @khanacad... link
Asked Who here still using the internet in 2023, he said so few know how the Internet actually works. Starlink is an entirely new internet in space – homogeneous, rather than heterogeneous. link
He said starlink antennas use an internally developed, electronically steered phased array tight beam that is more advanced than the best military technology link
He is not Biden's biggest fan, but if he received a presidential directive to turn Starlink on he would, because he regards the president as CEO of the country. link
He said this makes no sense. Starlink already covers Earth and our user terminals are being built & deployed super fast. This whole program will soon be redundant. link
He said there was an emergency request from government authorities to activate Starlink all the way to Sevastopol. The obvious intent being to sink most of the Russian fleet at anchor. If I had agreed to their request, then S... link
Asked ELON PREVENTS NUCLEAR WAR, he said the Starlink regions in question were not activated. SpaceX did not deactivate anything. link
He said much appreciated, Walter. The onus is meaningfully different if I refused to act upon a request from Ukraine vs. made a deliberate change to Starlink to thwart Ukraine. At no point did I or anyone at SpaceX promise co... link
He said i wish we were never down, but Starlink will keep improving rapidly. Our goal is a consistent <20ms latency to major data centers within a region. This is extremely difficult, but we will get there. Bandwidth will als... link
He has that power because SpaceX executed well, not through anti-competitive action; users can choose alternatives. link
He said excited to announce that @SpaceX @Starlink has achieved breakeven cash flow! Excellent work by a great team. Starlink is also now a majority of all active satellites and will have launched a a majority of all satellit... link
He said spaceX launches competitor satellite systems without favor to its own satellites. Fair and square. link
He said doesn’t make sense. Starlink is the only company actually solving rural broadband at scale! They should arguably dissolve the program and return funds to taxpayers, but definitely not send it those who aren’t getting ... link
Q29 How do you think about doing good?
He cares about the reality of goodness, not looking good. link
He advises people simply to try to be useful, which is very difficult. link
He says he respects anyone who works hard and builds things for their fellow humans. link
Q30 How do you hire and manage people?
He says manufacturing technology will be Tesla's most important long-term strength. link
He says he and the team review daily order data every day and try to make the least dumb decision. link
He says he believes in giving stock to the employees and creators of the company as an incentive. link
Q31 What about the advertiser boycott?
Brand-safe means the advertiser chooses what sits near the ad. If ads are timely and relevant, Twitter's revenue potential is gigantic. link
He said yeah, generally correct. Titter was trending to lose ~$3B/year (revenue drop of ~$1.5B + debt servicing of ~$1.5B) and had $1B in cash, so only 4 months of money. Extremely dire situation. Now that advertisers are ret... link
Most advertisers are coming back. Revenue dropped from about $4.5B to $3B while expenses went from $4.5B to $6B. link
With a few exceptions, almost all advertisers have said they have come back or will come back. link
Asked Thanks for the response @elonmusk. Does this mean we can post the film with full reach ..., he said it will be advertising-restricted, as advertisers have the right to decide what content their ads appear with, which will impact reach to some degree link
Asked I’ve not seen anything related to this up until recently. Are they, alone, seriously re..., he said based on what we’ve heard from advertisers, ADL seems to be responsible for most of our revenue loss. Giving them maximum benefit of the doubt, I don’t see any scenario where they’re responsible for less than 10% of t... link
He said our US advertising revenue is still down 60%, primarily due to pressure on advertisers by @ADL (that’s what advertisers tell us), so they almost succeeded in killing X/Twitter! link
He told advertisers trying to pressure him to go away, warned the boycott could kill the company, and said the public would judge who killed it. link
He said X will be fine and they are already seeing advertisers return; brand advertisers get upset, then usually calm down and come back. link
He said if antisemitism is the issue, as they claim, then other platforms are much worse, so that cannot be the real reason link
Q32 Did X throttle The New York Times?
Calls the NYT hypocritical for refusing to pay for verification while aggressively forcing everyone to pay for its own subscription. link
Organizations that refuse to buy a verification subscription aren't recommended; it costs about a thousand dollars a month. link
Q33 Is free speech under threat globally?
Asked .@elonmusk: "There were three areas that I thought would most affect the future of huma..., he said free speech is the bedrock of democracy. It is a fundamental defense against autocracy & statism, which is why we must fight so hard to preserve it. The founders showed great wisdom, based on painful historical lesson... link
The bedrock of a functioning democracy has to be free speech and a level playing field, which is why the First Amendment came first. link
That is up to Trump. Elon's job is that he takes freedom of speech very seriously. link
Free speech means allowing people you don't like to say things you don't like; once you lose it, it does not come back. link
He warned against anti-meritocracy and suppression of free speech. Free speech only matters when it is someone you do not like saying something you do not like; censorship gets turned on its advocates. link
He said i hear your concerns, but don’t judge too early. I am adamant about defending free speech, even if it means losing money. link
He said we’ve pushed harder for free speech than any other Internet company, including Wokipedia @jimmy_wales link
He says there has to be a very major national security reason for governments to secretly demand things from companies. link
He says resisting government demands depends on a company's willingness to oppose them, and X's willingness will be high. link
He said transparency is no problem. More of that would be great. But the government policing free speech in violation of the first amendment is not ok. link
The First Amendment is only relevant if you allow people you don't like to say things you don't like; if you like it you don't need a First Amendment. link
Free speech only matters if people you don't like can say things you don't like; otherwise it is only a matter of time before censorship turns on you. link
Free speech is only meaningful if you allow people you don't like to say things you don't like; that is how you know it is working. link
He said the point is not to ban free speech, but rather to acknowledge that calling for the death of any people is obviously harassment link
Q34 What happened with your 'actual truth' post?
He calls it possibly the most foolish of ~30,000 posts, says he's sorry, that his clarifications were ignored, and that he is not antisemitic. link
The original Nov 15, 2023 reply at the center of the controversy. link
Q36 Are you racist or anti-Muslim?
He said the woke mind virus amplifies racism and sexism and all the isms while claiming to do the opposite, dividing people. link
Q37 How should AI labs manage safety?
He said we need to regulate AI safety, as we do for cars, planes and medicine. AI is a bigger risk to society than those, and slowing it a little might also be a good thing. link
He said met with @SenSchumer and many members of Congress about artificial intelligence regulation today. That which affects safety of the public has, over time, become regulated to ensure that companies do not cut corners. A... link
He said regulatory oversight of AI is important to protect the public from companies potentially cutting corners on safety link
He supports some kind of international AI regulatory authority, while admitting enforcement would be hard. link
He says spending a year to make sure AGI is safe would probably be worth it, though he would not expect a big slowdown. link
He says the right way to regulate AI is to start with insight. link
He says AI needs some regulatory oversight, even though it is not a perfect nirvana. link
He suggests any verified human who wants a vote in the future of xAI should be allowed one. link
He said it is good for government to play a role when public safety is at risk. For digital superintelligence, government should be a referee, as in aviation and cars. link
Anything dangerous is overseen by a referee or regulator; he said we should have the same thing for AI to help ensure it is beneficial. link
Q38 What is the limiting factor for AI scaling?
He says Dojo could become a sellable service like Amazon Web Services. link
He says Tesla buys Nvidia GPUs while pursuing Dojo for an order-of-magnitude cut in training cost. link
He says a lot of AI is data curation, with surprisingly few lines of code. link
In July 2023 he predicted a shift from a silicon shortage to a voltage-transformer shortage in about a year and an electricity shortage in two years. link
He says the most important metric in a few years will be useful compute per unit of energy. link
Q39 What about unions at Tesla?
He dislikes anything creating 'lords and peasants'; the real challenge is retaining people for hard work. link
Asked What exactly is the reason for this, he said most of the Democratic Party is controlled by the unions – they carry far more weight than the environmentalists – and Biden particularly so (he gladly admits it). In Biden’s speech, posted on he literally says “the U... link
Q40 How many robots will there be?
He says economic output is people times productivity, so humanoid robots remove the constraint on people. link
Anything a computer can actuate is effectively a robot; Tesla cars are robots on wheels, and anything connected to the internet is an endpoint actuator for AI. link
Q41 What about immigration and H-1B?
He said immigrants from China & other Asian countries have made incredible contributions to America link
He said i am very much in favor of hard-working people coming to America, but we can’t have de facto open borders with no screening. Meanwhile, our legal immigration system is super slow & extremely difficult to navigate! link
Asked Do you support illegal immigration, he said i think we should significantly increase legal immigration. If someone has a track record of working hard and hasn’t committed crimes, they should be welcomed. link
He said we should greatly increase legal immigration of anyone who is hard-working, honest and loves America. Every such person is an asset to the country. But massive illegal immigration of people we know nothing about is in... link
Top line he is pro-immigration: increase legal immigration, stop illegal immigration. Welcome people who will work hard, have integrity, and add to the country. link
Q42 Is government a danger with AI?
He says he urged senior Chinese leaders to regulate AI and believes they took it to heart. link
He addresses the argument that regulating AI lets China leap ahead, saying he thinks China will regulate too. link
He says he would be willing to risk prison if the public good were significantly at risk. link
Q44 What is Neuralink's near-term goal?
He said what will be left for us humans to do? We better get a move on with Neuralink! link
He said those are the right questions to ask. They have been troubling me for many years. Neuralink was created as a possible long-term solution to human+AI symbiosis. link
He said neuralink can get us maybe even above chimp level link
He said no monkey has died as a result of a Neuralink implant. First our early implants, to minimize risk to healthy monkeys, we chose terminal moneys (close to death already), link
He said the first human patient will soon receive a Neuralink device. This ultimately has the potential to restore full body movement. In the long term, Neuralink hopes to play a role in AI risk civilizational risk reduction ... link
First human implant hopefully in a few months, letting paralyzed people use phones faster than people with hands; then vision. link
He said this is not much consolation, but Neuralink is working on a vision chip, which will be ready in a few years. That is the next area after enabling phone/computer telepathy for those who have lost their mind-body connec... link
Q45 Is there hidden alien or advanced propulsion technology?
He does not think it is aliens. The Fermi paradox: if the universe is 13.8 billion years old, where are they? He has seen no evidence of alien technology or any alien life whatsoever. link
He has seen no evidence of aliens yet; we are the aliens as far as he can tell. Humanity may be a tiny candle of consciousness in a vast darkness. link
He has seen no evidence of aliens whatsoever. Most likely in this part of the galaxy we are the only consciousness; human consciousness is a tiny candle in a vast darkness. link
He worries that we may be alone and that other civilizations failed to become multiplanetary. link
He says he has seen no evidence of aliens. link
Q46 Do regulations hold you back?
He says regulation helps mainly by forcing companies to think harder about safety internally. link
He says if regulation delayed AGI by six or 12 months, that would not be a big deal. link
He said every year that passes comes with more rules. And while all humans die, laws and regulations are immortal. So each passing year, we tie ourselves down with more and more strings, until, like Gulliver, we can no longer... link
He complies with millions of regulations and only objects to the rare ones that don't serve the public good. link
Rules, regulations and laws never die while people do; if we only add them, eventually we make everything illegal, like Gulliver tied down by strings. link
He said an excessive number of laws and regulations is the natural outcome of an extended period of prosperity. Historically, war would clean the slate of laws and regulations, but this can be done without war. We simply need... link
He argues there must be a process for garbage-collecting old rules and regulations. link
He says the US, especially states like California, is being strangled by regulation. link
He says historically the forcing function for cleaning up archaic rules has been war. link
He says it is a lot harder to get things done now than 10 years ago, as rules grow in every country. link
Q47 Has Tesla come close to failing?
In January 2023 he said Tesla was well placed to get through a recession, with almost no debt and over $20 billion in cash. link
In January 2023 he predicted a pretty difficult recession that year. link
Tesla went through very difficult times on the ragged edge of survival. If he had not given it everything, the company could easily have gone bankrupt; it was on the verge for quite a while, especially 2017–2019. link
He said ralph Nader, you are lying – shame on you! I personally provided almost all Tesla funding, based on my proceeds from PayPal, from Series A in 2004 until Series C in 2007. In late 2008, I gave Tesla the last money I ha... link
He says Tesla is growing faster than any company in history that makes a large complex manufactured object. link
He says a newcomer can only succeed with a product so compelling that people pay a premium over incumbents. link
Among American car companies only Ford and Tesla have not gone bankrupt; it was incredibly difficult to keep Tesla alive when GM and Chrysler were going bankrupt. link
He says going from prototype to volume production is vastly harder than making the prototype, and positive cash flow harder still. link
He recalls closing Tesla's 2008 financing round on Christmas Eve, two days before payroll would have bounced. link
He says no new car startup has been successful for 100 years apart from Tesla. link
Q49 Why is Tesla building a chip fab?
He says Tesla builds lithium and cathode refining only because others are not doing it. link
He said i should clarify that Dojo V1 is highly optimized for vast amounts of video training vs general purpose AI. Dojo V2 will address these limitations. link
He said not exactly a competitor to Nvidia, as we aren’t selling Dojo computers (although it could be sold as service). We just need a crazy amount of training compute to get full self-driving 10X safer than human, but can’t ... link
He said dojo is arguably the biggest computer architecture overhaul since Cray. Doesn’t mean it succeeds, but it doesn’t lack for boldness. link
He said there is a lot of merit to having your own chip design. We optimize its usage down to the microsecond. link
Asked Do you know if Tesla is doing the same with Dojo, he said we’re going with both Dojo and Nvidia. I think we will have the most inference silicon, given that every car has an inference computer. link
He said managing that data is extremely difficult. We ignore (don’t store at all) or delete almost all of it. The training videos that matter are rare ones, like weird & busy intersections, so less than 1% of all videos. Tesl... link
Q50 Why keep Tesla and SpaceX as separate companies?
He says any xAI relationship with Tesla has to be an arm's-length transaction. link
He says xAI and Tesla can help each other, and notes a materials science team shared between Tesla and SpaceX. link
He says SpaceX never thinks about the quarter or the stock price. link
Q51 What is the Kardashev scale / the big picture for civilization?
He says a multiplanetary civilization could last tens of thousands, hundreds of thousands or even millions of years. link
Q53 What's your vision for energy abundance?
He says Tesla will keep using outside cell suppliers because demand for lithium-ion batteries is quasi-infinite. link
With EVs, solar and wind, and stationary storage, he says Earth could become fully sustainable many times over. link
He said wind & solar, combined with batteries, will solve sustainable energy for Earth link
He said over time, we expect Giga Nevada to be almost entirely solar-powered link
He said absolutely. The amount of solar energy reaching Earth could easily power a civilization 100X times our size! link
He said earth is almost entirely solar-powered already. Without the sun, we’d just be a dark ice ball with some chemotrophic bacteria. link
He said Earth can and will move to a sustainable energy economy within your lifetime. link
He said a remarkably small amount of Earth's land area would let us go fully sustainable. link
He said and if you have enough rooftop solar, you can go off grid for most of year in high latitudes & whole year in low latitudes link
He said with enough rooftop solar you can go off grid most of the year at high latitudes and all year at low latitudes. link
He says the bottleneck for lithium is refining capacity, not mining, because lithium is common. link
He pleads for entrepreneurs to go into lithium refining instead of making picture-sharing apps. link
He says stationary storage growth will significantly exceed vehicle growth. link
He says Tesla will have by far the most lithium and cathode refining capability in North America. link
He said 10% of the world economy is a small price to pay for a sustainable energy future. link
He says if trends continue energy use could reach something like a terawatt per person. link
He said indeed, many parts of the world are reaching high peak levels of renewable energy production! Due to the obviously intermittent nature of wind and solar power, the limiting factor for renewable energy growth is increa... link
He said electric transportation and solar/wind electricity generation will have a tectonic effect on the geopolitical power balance link
He said the climate-change alarm is somewhat overblown in the short term; it is still a long-term concern but exaggerated in the short term. link
He said teslas can charge from solar panels even if the world goes fully Mad Max and there is no more gasoline! link
Q54 How should X recommend content?
He said the giant block lists are problematic. They mess up the recommendation system & create a DDoS vector. link
He said long day at Twitter HQ with eng team Two significant problems mostly addressed: 1. Fanout service for Following feed was getting overloaded when I tweeted, resulting in up to 95% of my tweets not getting delivered at ... link
He said fixing the recommendation algorithm is our current top priority. Twitter engineering has been working super hard on this. Proud of the team. link
He said prepare to be disappointed at first when our algorithm is made open source next week, but it will improve rapidly! link
He said our “algorithm” is overly complex & not fully understood internally. People will discover many silly things , but we’ll patch issues as soon as they’re found! We’re developing a simplified approach to serve more compe... link
He said twitter will open source all code used to recommend tweets on March 31st link
He said it will be made open source, but we’re finding so many bugs that those should be fixed first link
He said twitter will be updating its recommendation algorithm based on user suggestions every 24 to 48 hours link
He said most of the recommendation algorithm will be made open source today. The rest will follow. Acid test is that independent third parties should be able to determine, with reasonable accuracy, what will probably be shown... link
He said this is most of the recommendation code. In the coming weeks, we will open source literally everything that contributes to showing a tweet, such that the output can be at least roughly approximated. link
He said nice work by the recommendations algorithm team! As always, changes will be open source. So far, over 1M lines of code have been uploaded to our GitHub repository. link
He said the commitment to open source transparency and accepting a wide range of viewpoints remains unchanged link
Says the algorithm tries to maximize unregretted user time, and compelling links shared by many will be seen. link
He said a major update to the recommendation algorithm is rolling out over the next few days. This will help surface smaller accounts and posts outside of your friend-follows network. As always, this will be made open source ... link
He predicts that any social network that does not raise the cost of bots will be swarmed and overwhelmed by them. link
He says elevating subscriber content above bot-amplified content is a very big deal. link
He says the fix for bots is to dramatically increase the cost of noise in the system. link
He says the network has to raise the cost of manipulation, bots and trolls, since AI can now pass are-you-human tests. link
Q59 Will Tesla own the robotaxi fleet?
He says a typical car has only 5 to 10% asset utilization, which autonomy could raise sharply. link
Q60 Will Starlink go public?
He says SpaceX does not anticipate needing to raise funding. link
He says the regulatory and legal burdens of being a public company are far too high. link
He says it is not worth going public until revenue is extremely stable and predictable. link
Q61 Will you stay CEO of Tesla?
He says these are uncertain times and he would like to borrow anyone's crystal ball. link
He says a public company requires more management time than a private one. link
Q63 Did Twitter exaggerate its user numbers?
He says requiring payment raises the cost of bots by several orders of magnitude. link
Q68 Is DEI connected to antisemitism?
He said DEI must DIE. The point was to end discrimination, not replace it with different discrimination. link
Q69 Will X become a payments app?
The long-term vision is x.com as an everything app: payments, financial services, information flow, anything digital, and secure communications — as useful and entertaining as possible. link
x.com should let people send money in one click, earn interest, and take on debt — possibly becoming the biggest financial institution in the world. link
In December 2023 he expected approvals early in 2024 and said he would be surprised if X payments took longer than mid-2024 to roll out. link
In late 2023 he said X was waiting for final approvals of its money transmitter licenses. link
Q75 Why is SpaceX building a self-growing city on the Moon before Mars? (asked on X)
Asked When I was a kid, I genuinely believed this was where we'd all be living by now. Do kid..., he said the mass & difficulty required to build a space habitat like that is extreme. Far better to have a base on the moon & city on Mars. link
He said high time we went back and had a permanently occupied lunar base link
Q79 When does Tesla stock have its NVIDIA moment? (asked on X)
He says Tesla chose higher volume and a larger fleet over lower volume and higher margin. link
He says it is a good time to increase Tesla's lead and invest in growth as fast as possible. link
He says a software update making millions of cars autonomous would be the biggest asset value increase in history. link
He says passive funds were a good idea but have gone too far, amplifying a few active decisions. link
Q83 When will Grok models be open-sourced? (asked on X)
He says closed AI will outperform open source by a meaningful margin at any point, since six months is an eternity. link
Q86 When will Tesla's larger FSD model ship? (asked on X)
He says the price of FSD will increase over time in proportion to its value. link
Q87 Will Grok get an Analyze button that verifies claims in posts? (asked on X)
He said in a few weeks, we will add a “Grok, analysis!” button under 𝕏 posts link
Q95 How will AI overcome the data wall? (asked on X)
He says the path to AGI is AI generating content and self-assessing it, like AlphaZero playing itself. link
He says at a certain point AI runs out of human-created data. link
He said yeah. It’s a little sad that you can fit the text of every book ever written by humans on one hard drive (sigh). Synthetic data will exceed that by a zillion. link
Asked My god. It can do basic math, he said that part doesn’t matter. Synthetic data generation and accurate evaluation, ie recursive self-improvement, is more likely. link
Q100 What will Tesla's Robotaxi look like? (asked on X)
Asked Tesla’s $25k car and dedicated robotaxi will be Cybertruck-inspired. Thoughts, he said no, but Robotaxi will look like the future link
Q101 Will Neuralink enable telepathy between humans? (asked on X)
Asked Hey Elon, has the team ever discussed the possibility of increased communication bandwi..., he said yes, we call it conceptual telepathy link
Q102 Will Optimus have a physical kill switch? (asked on X)
He said i sure hope Tesla bots don’t malfunction! We’re taking great care to ensure that Optimus can easily be powered down locally with a physical switch to reduce the risk of centralized control. link
He says Optimus must have a hard-coded local cutoff that no internet update can change. link
He says Optimus at scale needs a hard-coded way to turn it off. link
He says anyone will be able to shut Optimus off locally, and that this cannot be changed by a software update. link
He says Tesla will put a lot of effort into localized control of the humanoid robot to avoid a Terminator outcome. link
A humanoid can chase you anywhere, so there should be a hardwired local cutoff you cannot update from the internet — a local off switch that puts the robot into a safe state. link
Q107 When will a city on Mars be self-sustaining?
He says a self-sustaining city should ideally be on Mars, because Mars is far from Earth. link
Q109 How important is physical-world AI compared with language models?
Says real-world AI is a different thing from a language model. link
He calls Tesla's driving AI "baby AGI" because it has to understand reality to drive. link
He says full self-driving needs something like baby AGI in the car, because it must understand messy reality. link
Q110 Is wanting to be liked a problem when building a company?
Jokes that he has overcome the desire to be liked. link
Q112 Where do you get your news?
Says he rarely reads legacy media anymore and gets his news from X instead. link
He says media not maximizing clicks gets outcompeted by media that does. link
He says you want information from experts in the field, or in a war zone, people literally in the field. link
He prefers information from people in the field over journalists who are usually not at the scene. link
Q113 How capable is Grok 3?
He says one goal for Grok is to be the funniest AI. link
Q114 How does Community Notes work?
Community Notes is a competition for truth: the people of Earth compete to provide the most accurate information. He called it a very powerful concept. link
Calls Community Notes a game changer for combating wrong information. link
He apologizes for rate limiting, saying scraping was bringing the system to its knees. link
Community Notes does not delete anything; it simply adds context. A note is shown only when people who have historically disagreed agree. link
For a note to be shown, people who have historically disagreed must agree. He said the batting average is extremely good. link
He cites the police saying that every crime has at least three stories: the victim's, the perpetrator's and the truth. link
He says the goal is getting as many people as possible providing accurate information and correcting what is not. link
Q115 When will you step down as CEO of X?
He needed to stabilize the organization and the product roadmap; he guessed towards the end of 2023 would be good timing to find someone else to run the company. link
Announces he has hired a new CEO for X/Twitter, starting in about 6 weeks, and that he will move to exec chair and CTO. link
Announces Linda Yaccarino as CEO, focused on business operations while he focuses on product and technology. link
Q116 Should people work remotely?
He criticizes the work-from-home crowd for ignoring the people who must show up at factories and restaurants. link
Q118 What's your algorithm for designing products?
He says manufacturing technology will be Tesla's most important long-term strength. link
He says owning service means Tesla feels the pain and can redesign cars to need less service. link
He says SpaceX is the best rocket maker in history at production. link
He says the pace of innovation, not any given innovation, is the real competitive advantage at Tesla and SpaceX. link
He describes cost reduction as a game of pennies across roughly 10,000 parts. link
Q120 How should government award contracts?
He calls for class action lawsuit reform, saying bad suits effectively tax consumers. link
Q125 What is Tesla's mission?
He says a vast number of people want a Tesla but cannot afford one, so price really matters. link
He says people with a lot of money forget how important affordability is, and Tesla aims to make affordable cars. link
He says Tesla wants all EVs to succeed and points to opening its Superchargers and making its patents free. link
He predicts people will look back on gasoline cars the way they look back on steam engines. link
He says the best service is no service, because the car doesn't break. link
He says gasoline cars will become like steam engines: quirky collector items. link
He tells analysts to stop looking at an EV market and just look at how many cars Tesla sells. link
He says Tesla open-sources patents to act aspirationally as a minesweeper. link
He stresses that cost is not optional for most buyers and Tesla must make its cars more affordable. link
He says Tesla is simplifying the next vehicle to reach a production rate unheard of in the industry. link
He says Tesla may have created more employee millionaires than any company in history. link
Q128 Why did Tesla build the Cybertruck?
He says the Cybertruck will be the car he drives every day. link
In April 2023 he hoped the first Cybertruck handover would be around the end of Q3 2023. link
He says Cybertruck production will follow an S-curve like all new products. link
He calls Cybertruck a fantastic product, a hall of famer, but very radical to manufacture. link
He said the Cybertruck is the car he will drive day to day. link
In October 2023 he tempered expectations, saying it would take a year to 18 months for the Cybertruck to be a significant positive cash-flow contributor. link
He predicted Cybertruck production would end up at roughly a quarter million a year. link
He says a radical product like the Cybertruck is extremely hard because there is nothing to copy: you invent the car and how to make it. link
He warned of enormous challenges in reaching volume production with the Cybertruck and making it cash-flow positive. link
He calls Cybertruck potentially Tesla's best product ever but says it will require immense work to reach volume. link
He says over a million people have reserved the Cybertruck, so demand is not the issue. link
He said with the Cybertruck the future will finally look like the future. link
He described it as a once-every-five-to-10-years kind of special product. link
Q129 Can the Boring Company solve traffic?
He says the Boring Company has the most advanced tunnelling machine and that regulatory approval is by far its biggest constraint. link
Q130 What is the Tesla Semi?
Says Gates called the Semi impossible while it was already being driven across the country. link
Q131 Why does rocket reusability matter?
He says Starship basically works, and SpaceX will fix the remaining issues, get it to orbit and make it reusable. link
He says it will take a few years for Starship to reach where Falcon 9 is, where landing is normal. link
Q132 How fast is Tesla expanding Superchargers?
He says Tesla wants all EVs to succeed and is opening up its Superchargers. link
Q134 Why does Tesla use cameras instead of lidar?
Says vision is hard but fundamentally superior to lidar. link
He argues the car can far exceed average human safety using vision alone. link
He explains Tesla turned off radar in Model 3 and Y because it generated more noise than signal. link
He says the car looks in all directions at once and the computer never gets tired or distracted. link
He explains Tesla turned off radar because of the disambiguation problem between vision and radar. link