How good is Tesla self-driving?

"90% of miles driven being autopilot"

"So what we're doing is similar to maybe what you've experienced with Waze where you have the cars operating collaboratively as a network"

"I think we'll probably be able to do it in five or six years."

"And the standard for fully autonomous driving is going to be much greater than for a person"

"much easier than people think"

"autopilot version one is really intended as a load alleviation for the driver"

"my guess for when we will have full autonomy is about three years"

"the two most profound innovations in automotive since the moving production line are electrification and autonomy."

"Essentially the network of vehicles is going to be constantly learning"

"I will be surprised if a majority of them are not self-driving in let's say 10 to 15 years"

"nobody will buy a car unless it is autonomous"

"rather that it's a workload reduction"

"The basic news is that all Tesla vehicles exiting the factory have the hardware necessary for level five autonomy."

"if in writing some article that's negative, you effectively dissuade people from using an autonomous vehicle, you're killing people."

"it will be very unusual for cars to be built that are not fully autonomous"

"You can absolutely be superhuman with just cameras"

"a fleet white hack is basically impossible"

"We have the most advanced AI neural net of any consumer product by far, so it’s going through exhaustive testing."

"the hardest thing to predict about the timing is regulatory approval."

"The statistics are unequivocal that autopilot improves safety. No question."

"even though we think autonomy, even the current autonomy, reduces the probability of a death by 50%"

"Like, they get too used to it. That tends to be more of an issue."

"It certainly needs to be better & we work to improve it every day, but perfect is enemy of good. A system that, on balance, saves lives & reduces injuries should be released."

"To date, Autopilot resources have rightly focused entirely on safety. With V9, we will begin to enable full self-driving features."

"Reality is that it will be pretty easy to bully a self-driving car, as it will always yield. Will prob have a manual override that requires continuous press for hardcore lane changes."

"My guess as to when we would think it's safe for somebody to essentially fall asleep and wake up the destination probably towards the end of next year."

"we expect to be feature complete with self driving sometime later this year."

"now we've finally developed our full self-driving computer"

"LIDAR is a fool's errand and anyone relying on LIDAR is doomed"

"I feel very confident predicting autonomous robotaxis for Tesla next year"

"when we do it this year, everyone with Tesla Full Self-Driving will be able to do it too"

"the reason self-driving is hard is because you have all these corner cases"

"I think probably sometime next year you'll be able to have the car be autonomous without supervision."

"I can tell immediately if a car is on Tesla autopilot"

"I said I was hoping would be future computer with FSD by the end of last year. We got pretty close. You know, it's looking like we might be feature complete in a few months."

"It's a fundamental information advantage that insurance companies don't have."

"arguably like a three order of magnitude improvement in labeling efficiency."

"feature complete just means like it has some chance of going from your home to work, let's say with no interventions."

"we're really only beginning to take full advantage of the Autopilot hardware, the FSD hardware."

"the driver when driving and taking action is effectively labeling reality as they drive and making neural net better and better."

"So when you search on something and you click on a link, you're training Google every time you do that."

"This is an advantage that no one else has. We're quite literally orders of magnitude more than everyone else combined."

"There's also a tremendous amount of untapped potential in the fleet out there that could upgrade to turn on autopilot"

"And then we keep improving the reliability to the point where it no longer needs to be supervised by the driver."

"we will be collecting data from over a billion intersections per month."

"this is why I'm very confident about full self driving functionality being complete by the end of this year, because I'm literally driving it."

"So I personally test the latest alpha build of full self driving software when I drive my car and it is really, I think, profoundly better than people realize."

"it's almost getting to the point where I can go from my house to work with no interventions"

"it's just hard to convey just how much better a fully 4D system would work."

"when things do become full self driving, so what are people going to do in the car? Well, I guess they're probably going to do productivity and entertainment of some kind."

"There's just a massive amount of work with each kind of order of magnitude of reliability,"

"The actual major milestone that's happening right now is really a transition of the Autonomy system"

"this is a generalized neural net based approach. There is no need for high definition maps or a cell phone connection."

"It's the same thing for autonomy, having on the order of a million cars that are providing feedback"

"weird corner cases. Only reality can give you that."

"the car should still be able to drive just like a person."

"even if you have no connectivity whatsoever and you're in a place that you have never been to before and no Tesla has ever been there, the car should still be able to drive just like a person."

"Tesla Full Self-Driving will work at a safety level well above that of the average driver this year, of that I am confident."

"this in itself is good for the environment because you need fewer cars to get the same thing done."

"It will just keep minimizing the amount of input that you need to do until the car just reads your mind."

"I think it's looking highly likely for, to be in Europe next year."

"I think Tesla's open to licensing Autonomy because I think Autonomy will be such a significant lifesaver and preventer of injuries that it is not a technology we want to keep to ourselves."

"it's looking quite likely that it will be next"

"So, and so if we can have autonomy that would potentially solve, you know, save on the order of a million lives per year."

"I actually am confident that we will solve it this year"

"the best real world ai team"

"we have never said ever that Tesla uh the Tesla autopilot does not require attention"

"leading real world AI company"

"Self-driving electric cars will be all that matters."

"basically FSD Beta will be available to anyone who requests it by the end of this year."

"But eventually all cars will be self driving."

"five-fold increase in the utility of a car"

"Hardware 3 will not be as good as hardware 4. But I'm confident that hardware 3 will so far exceed the average, the safety of the average human."

"I don't think you could see second place with a telescope"

"we don't see anyone even close to us in, in achieving this"

"And the value of FSD grows as the autonomous capability grows."

"FSD 11.3 is close to limited release. Just drove it through SF & LA yesterday. Probably 11.3.2 is able to go wide."

"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."

"full self-driving capability has been evolving"

"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."

"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 compute-efficient than transformers for vision."

"that's the best of our knowledge that we believe the current hardware can achieve full autonomy."

"it's better to ship a large number of cars at a lower margin and subsequently harvest that margin in the future as we perfect autonomy."

"Those who understand AI will understand the importance of data, of training data"

"the price right now is an option value on an autonomous vehicle"

"we've now crossed over 150 million miles driven by full self driving beta and this number is growing exponentially."

"It will be much, much better, like a lot, like over time, 10 times more, 10 times safer than a person."

"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."

"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."

"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."

"The awareness of the capability of FSD is still shockingly low, even among Tesla owners"

"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 country."

"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."

"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."

"Tesla is prioritizing safety over comfort for FSD beta. As we become confident in the former, the latter will follow."

"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 license Autopilot/FSD or other Tesla technology."

"Tesla is the leader I think by pretty long margin in real world AI."

"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 constrained, not engineer constrained."

"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."

"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."

"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."

"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 automotive accidents and ~10 million are seriously injured."

"the entire self-driving AI software team is 200 people"

"A long tail of training data – reality has a vast number of weird edge cases – is required to solve self-driving well"

"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."

"I would characterize our progress in real world AI as a series of stacked log curves."

"Our system is meant to be holistic and drive in any conditions."

"Tesla has the best real world AI team on Earth, period."

"the economics of a fully autonomous vehicle are truly astounding in a positive way."

"for most parts of the world you have to get approval before deploying things, whereas in the US you can deploy things at risk or take liability for what you deploy."

"humans are terrible drivers"

"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/week, leaving 2/3 for distributed inference, like SETI."

"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."

"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."

"when we can show that even completely without supervision, there is a massive statistical case for self driving being actually safer than if someone had their hands on the wheel"

"So we're really just going through a march of nines."

"it will be the biggest step change in asset value in history"

"Almost no humans actually come to a full stop."

"Cybertruck is necessarily lowest priority for FSD, as there only a few hundred Cybertruck in the field compared to ~5 million other Teslas"

"When unsupervised FSD significantly exceeds human safety levels and is approved by regulators, the numbers will get really crazy!"

"He was not on FSD. The software had unfortunately never been downloaded. I say “unfortunately”, because the accident probably would not have happened if FSD had been engaged."

"We have been short on training compute for FSD, so haven’t trained with other video, but certainly could. Will do it later this year when we have some spare capacity."

"I’ve wanted to do that for a long time :) Our real-world simulation and video generation is the best in the world, but unfortunately making a game can only come after we release unsupervised FSD that is far safer than even supervised FSD."

"Where Tesla video generation exceeds OpenAI is that it predicts extremely accurate physics. That is essential for self-driving."

"It will take at least a decade before a majority of cars are self-driving, but this is a legitimate concern in that time horizon. Of more immediate concern is that it is already possible to severely ostracize someone simply by freezing their credit cards and bank account, as happened, for example, in Canada with the trucker protest."

"Three significant improvements to FSD will roll out roughly every two weeks. Should be really shining bright by late April or early May."

"V12.3.1 addresses several small annoyances and should start rolling out this weekend"

"Yeah, 99% of people have no idea. And improvement will accelerate dramatically now that we are no longer AI training compute constrained."

"All US cars that are capable of FSD will be enabled for a one month trial this week"

"Most people still have no idea how crushingly good Tesla FSD will get. It will be superhuman to such a degree that it will seem strange in the future that humans drove cars, even while exhausted and drunk! Cars will take you where you want automatically, just like getting in an elevator and pressing a button, something that also used to be manual."

"Won’t be long before Tesla exceeds 10B miles of FSD"

"Pretty much. It has been staggeringly difficult to make generalized self-driving work, requiring all that you describe above and more. The investment in training compute, gigantic data pipelines and vast video storage will be well over $10B cumulatively this year. But that is nothing compared to the ~quarter trillion dollars in cars on the road with Tesla-designed AI inference computers being trained by their drivers."

"it's actually been pretty helpful that other autonomous car companies have been cutting a path through the regulatory jungle."

"Every time something goes wrong that gets added to the training data and you get this training flywheel happening in the same way that Google search has this sort of flywheel."

"Tesla's inference AI Inference efficiency is very vastly better than anyone any other company."

"Elevators used to be operated by a guy with a relay switch, but sometimes that guy would get tired or drunk or just make a mistake"

"We believe version 12 is ready for supervised FSD in LHD countries. RHD will take a bit longer."

"Accurate analysis. Only about 1/10,000 of distance driven is useful for training. The limiting factor for FSD progress was AI training compute, but now it is validation, as the interventions are so rare."

"We must focus on FSD for the other 99% of the fleet. Cybertruck FSD is a few months away."

"Should be very rare after the 12.4 rollout. We have to sync up vehicle software (way too many hardware variants) with FSD."

"FSD 12.4.1 releases today to Tesla employees. If that goes well, then it will be released to a limited number of external customers this weekend. There are a massive number of changes to this build. It should arguably be called v13, but we’re sticking to 12 😂 Two other versions are in earlier stages of testing: 12.5 and 12.6, which could be called v14 and v15. We are starting to get to the point where, once known bugs are fixed, it will take over a year of driving to get even one intervention."

"the overwhelmingly important thing is is achieving unsupervised full self-driving"

"Yeah, it’s awesome. Once you use Tesla Autopilot/FSD for commuting, you won’t use any other car."

"All that matters is improving Tesla self-driving AI such that miles between interventions is >> human"

"The amount of testing time it takes to figure out if the new AI is better than the existing AI as measured by miles between interventions is the limiting factor on progress. The better FSD gets the longer it takes to find interventions."

"Yes, Cybertruck FSD releases with 12.5, which also finally merges the city and highway code."

"Tesla has done so much that would be considered “AI research breakthroughs”, but we are laser-focused on making it work, rather than crowing about breakthroughs. Having a large fleet is only one piece of the puzzle. Plenty of other car companies produce way more cars than Tesla, but are nowhere on generalized self-driving."

"you could do unsupervised possibly by the end of this year. I would be shocked if we cannot do it next year."

"pretty soon we will ask for regulatory approval of the Tesla supervised FSD in Europe, China and other countries. And I think we're likely to receive that before the end of the year."

"that trend exceeds humans for sure next year."

"our solution is a general solution that works anywhere."

"So if you rendered a new Earth, it would work on a new Earth."

"By far the biggest differentiator for Tesla is autonomy."

"There are a few major OEMs that have expressed interest in licensing Tesla full stop driving"

"Just did a walkthrough of the Tesla supercompute cluster at Giga Texas (aka Cortex). This will be ~100k H100/H200 with massive storage for video training of FSD & Optimus. Great work by the Tesla team!"

"Any car company that fails to solve self-driving will die"

"Haven’t read the article, but the above is not accurate. Tesla has learned a lot from discussions with engineers at xAI that have helped accelerate achieving unsupervised FSD, but there is no need to license anything from xAI. The xAI models are gigantic, containing, in compressed form, most of human knowledge, and couldn’t possibly run on the Tesla vehicle inference computer, nor would we want them to. The Tesla AI models have incredibly “dense” (in a good way lol) intelligence, as they compress video of reality into driving commands, but must operate on a ~300W computer with memory size and bandwidth far lower than say an H100 GPU. Tesla real-world AI also has a vastly larger context size than an LLM, as the combined video history from all cameras is several gigabytes in size."

"We should change that so the yoke is stationary, as it is not mechanically coupled to the wheels. Same goes for when it is on Autopilot."

"Then the battery packs of the busses are too small. Teslas can operate well even above the Arctic Circle. Also, busses will not be needed when Tesla rolls out unsupervised full self-driving, as they will take people point to point for a similar cost to a bus ticket."

"If I were another car company, I would license FSD"

"So, we do expect actually to start fully autonomous, unsupervised FSD in Texas and California next year."

"I think we'll see autonomous cars become ten times safer than a human."

"FSD will soon exceed 10,000 miles between critical interventions, which is a year of driving for most people"

"Our internal estimate is Q2 next year to be safer than human and then to continue with rapid improvements thereafter."

"we think that we'll be able to have driverless Teslas doing paid rides next year. Sometime next year."

"we'll definitely have available in Texas and probably have it available in California subject to regulatory approval"

"there's no need to wait for robot taxi or Cyber cab for to experience full autonomy. We expect to achieve that next year with the, with our existing vehicle line."

"it's currently showing a 10x safety improvement relative to the US average."

"should be frankly, blindingly obvious at this point that that is the future."

"there are significant improvements in the miles between interventions."

"Therefore the right, imo only, general solution is artificial neural nets and cameras."

"Not Teslas, once unsupervised FSD starts to happen next year"

"Your Tesla can now drive you from your house in one complex city to another house in a different complex city"

"Strange, as Google was in the best position to solve for an AI self-driving solution"

"There will be some buying of cars for individual ownership, but self-driving cars will be >90% of miles driven"

"Autopilot will not function unless it detects an attentive person in the driver’s seat"

"Unsupervised full self-driving begins"

"unsupervised FSD in the US this year, in many cities, but nationwide next year."

"We feel confident in being able to do an initial launch of Unsupervised no one in the car, Full self driving in Austin in June."

"my number one recommendation for anyone who doubts is simply try it."

"What we're seeing is at this point significant interest from a number of major car companies about licensing Tesla full self driving technology."

"who's in second place for real world AI? I would need a very big telescope to see them."

"people will actually go to manual driving to check their text messages so the computer doesn't yell at them"

"I think this will be the largest asset value increase in human history."

"we're looking for a safety level that is significantly above the average human driver."

"So thousands of cars every day are driving with no one in them."

"Now I'm kind of glad that not that many people bought the FSD package."

"it's like meeting someone when they're like a toddler and thinking that they're going to be a toddler forever."

"Within a few years, almost no one will want a car without full self-driving"

"The Tesla factory IS the product. The Cybercab production line is like nothing else in the automotive industry."

"We just used publicly available video on the Internet of roads and signs in China and used that to train in sim"

"in 10 years probably 90% of miles driven will be uh autonomous"

"It will be ultimately 10 times safer than a human because it never gets tired."

"And I think in five years probably we'll, we'll have regulatory approval, I think globally."

"People don’t shoot lasers out of their eyes to drive. Just try Tesla self-driving today, which just uses cameras and AI, and you will understand."

"Exactly. We could have done a neighborhood by neighborhood handcrafted solution with high resolution maps and expensive sensors much sooner, but that doesn’t scale well and gets destroyed by real-world AI and cameras."

"A crazy amount of execution is needed to get there, but I do see a path to world’s biggest company"

"💯 The Tesla factory, especially our next gen Cybercab line, is the product. That, autonomy and Optimus, are what matter."

"For the first time, there will soon be a generalized, pure AI solution to full self-driving. Just cameras and the Tesla AI chip with Tesla AI software."

"Tesla full self-driving will amaze you"

"Tesla self-driving will be far safer than human driving"

"this is the advantage of having a generalized solution using artificial intelligence and an AI chip that Tesla design specifically for this purpose"

"can you go to sleep in your car and wake up at your destination? And I'm confident that will be available in many cities in the US by the end of this year."

"I'm confident also that later this year the first Model Y will drive itself all the way to the customer."

"what we're solving for is a general solution to autonomy, not a city specific solution for autonomy"

"buying a gasoline car that is not autonomous will be like riding a horse while using a flip phone."

"It needs to be meaningfully safer than if it's cars manually driven."

"we're at Tesla, we're absolutely hardcore about safety."

"I feel confident in predicting large scale autonomy around the middle of next year, certainly the second half next year, meaning I predict that there will be millions of Teslas operating autonomously"

"we'll be launching unsupervised autonomy, basically self-driving cars with no one in them in Austin next month"

"unsupervised full self driving"

"Cybercab will become ubiquitous"

"Robot cars (aka FSD) should have a noticeable effect in 1 to 2 years"

"Tesla self-driving massively improves your quality of life and safety for the thousands of life hours you’re in a car"

"Yes And unsupervised FSD in Austin"

"For the past several days, Tesla has been testing self-driving Model Y cars (no one in driver’s seat) on Austin public streets with no incidents. A month ahead of schedule. Next month, first self-delivery from factory to customer."

"This tweet 9 years ago was how I started the Autopilot, now AI, team at Tesla. Ashok, who now leads the team, was the first person I interviewed. Milan, who leads Optimus, also joined very early. Many of the key people in Tesla AI have been there from the beginning."

"Tesla AI/Autopilot could probably beat the best human around a track already. Eventually, it really won’t be a contest at all. Human reflexes cannot match machines."

"The automakers keep being told that this isn’t real or that just buying some hardware from Nvidia will solve it. As Tesla robotaxis become widespread and their other solutions don’t work, they will naturally turn to us."

"These are unmodified Tesla cars coming straight from the factory, meaning that every Tesla coming out of our factories is capable of unsupervised self-driving!"

"Public support is much appreciated. FSD in Europe will save lives of people in the car, as well as vulnerable road users like pedestrians, cyclists and pets. Please let your government know that you want approval."

"Tesla FSD can operate in all conditions"

"Your team has done a great job maximizing the performance of the Tesla AI computer. Robotaxi wouldn’t be possible yet without that."

"The @teslasi robotaxi launch begins in Austin this afternoon with customers paying a $4.20 flat fee!"

"We need to validate that improvements for Austin don’t cause regressions elsewhere. For Europe and China, we are awaiting regulatory approval. Hopefully, soon🤞"

"Your Tesla self-driving capability will see a step change improvement as we integrate upgrades for the Austin robotaxi build into the general production release"

"Not sure what the status of Illinois is. A pretty good guide is that if Waymo is operating somewhere, then regulations allow robotaxi in that city or state. Tesla will operate there as soon as we reaffirm safety testing in that locale and receive a license to operate."

"Tesla is training a new FSD model with ~10X params and a big improvement to video compression loss. Probably ready for public release end of next month if testing goes well."

"Tesla already does that for self-driving! For real-time video generation with interaction like a good video game, I’d say next year."

"This will substantially reduce the need for driver attention, but some complex intersections, heavy weather or unusual events will still require attention. Note, the Austin robotaxi FSD build is ~6 months more advanced than what is available in cars in America and there are some additional breakthroughs in Tesla AI that will make the car feel eerily human."

"The FSD software update next month will be a major step-change improvement for rare conditions"

"The FSD release in about 6 weeks will be a dramatic gain with a 10X higher parameter count and many other improvements. It’s going through training & testing now. Once we confirm real-world safety of FSD 14, which we think will be amazing, the car will nag you much less."

"The Tesla Robotaxi service area is already larger than any competitors in Austin and the Bay Area"

"They will switch from claiming that Tesla self-driving is fake to Tesla self-driving is unfairly good"

"This variant of the Model Y doesn’t start production in the US until the end of next year. Might not ever, given the advent of self-driving in America."

"(He was just on version 13). Version 14 is the second biggest update to Tesla AI/Autopilot ever after V12. It feels alive."

"Version 14 of Tesla self-driving feels sentient"

"Lidar and radar reduce safety due to sensor contention. If lidars/radars disagree with cameras, which one wins?"

"your car is going to feel like it is sentient by the end of the year"

"Robotaxi app now available for download"

"Unify in architecture, not in details. The parameters/activations will be different."

"Many more Tesla self-driving V14 improvements still to come, but wide rollout of V14.1 has begun. 14.2 rolls out in a few weeks and then 14.3 a few weeks later, depending on safety testing. There is so much change that we are carefully confirming each one."

"You can now have a self-driving Tesla take you to SJC airport!"

"Cool! FSD is super convenient and increases safety significantly."

"It’s easy to get to pretty good self-driving, but insanely hard to get to great"

"Tesla autonomous driving might spread faster than any technology ever. The hardware foundations have been laid for such a long time that a software update enables self-driving for millions of pre-existing cars in a short period of time."

"“Robotaxi” is the same Model Y that everyone who buys the car receives. It just has a slightly different software build."

"Tesla Model Y robotaxi service now available in the greater Austin area!"

"Tesla is the world leader in real world AI"

"it's four times safer on full self-driving than than than not"

"Not sure if it’s super new, as we’ve been doing something along these lines at Tesla for a while, but it does make sense. The single biggest technical challenge of Tesla self-driving AI is context compression of 1.5GB/s of video to ~2kB/s of control outputs using a puny inference computer without making any mistakes. Super hard to avoid overly lossy compression at any given step."

"Yeah, this is actually a safety issue! We need to allow people to text and drive while on Autopilot, rather than turn off Autopilot to do so."

"Exactly. The reason Tesla hasn’t yet spent more on AI training is that it wasn’t yet the Limiting Factor for FSD. It will be that for Optimus, so we’re building Cortex 2 for training."

"I’ve tried to warn them and even offered to license Tesla FSD, but they don’t want it! Crazy … When legacy auto does occasionally reach out, they tepidly discuss implementing FSD for a tiny program in 5 years with unworkable requirements for Tesla, so pointless. 🤷‍♂️ 🦕 🦕"

"The Tesla Robotaxi fleet in Austin should roughly double next month"

"Over 6000 consecutive miles using Tesla full self-driving!"

"car is is essentially a solved problem at this point"

"Nvidia is providing helpful tools to the automotive industry, but the automotive industry is doing very little on their own By the end of this year, Tesla will have spent ~$10B cumulatively just on Nvidia hardware for training. We combine that with our own AI4 chips for processing vast amounts of video or we would probably need to spend double that amount. Then we are making ~2M cars/year and rising, all with our dual SoC AI4, 8 cameras, redundancy in steering actuation and others systems, high bandwidth communication, etc."

"You’re right. The actual time from when FSD sort of works to where it is much safer than a human is several years. The legacy car companies won’t design the cameras and AI computers into their cars at scale until several years after that. So this is maybe a competitive pressure on Tesla in 5 or 6 years, but probably longer."

"Roughly 10 billion miles of training data is needed to achieve safe unsupervised self-driving. Reality has a super long tail of complexity."

"Tesla will stop selling FSD after Feb 14. FSD will only be available as a monthly subscription thereafter."

"AI4 by itself will achieve self-driving safety levels very far above human. AI5 will make the cars almost perfect and greatly enhance Optimus. AI6 will be for Optimus and data centers. AI7/Dojo3 will be space-based AI compute."

"With the important caveat that initial production is always very slow and follows an S-curve. The speed of the production ramp is inversely proportionate to how many new parts and steps there are. For Cybercab and Optimus, almost everything is new, so the early production rate will be agonizingly slow, but eventually end up being insanely fast."

"Insurance is half price when Tesla self-driving is activated, because it increases safety so much"

"Just started Tesla Robotaxi drives in Austin with no safety monitor in the car. Congrats to the @teslasi team! If you’re interested in solving real-world AI, which is likely to lead to AGI imo, join Tesla AI. Solving real-world AI for Optimus will be 100X harder than cars."

"Cybercab production starts in April"

"I should also mention that the $99/month for supervised FSD will rise as FSD’s capabilities improve. The massive value jump is when you can be on your phone or sleeping for the entire ride (unsupervised FSD)."

"it just increasingly feels sentient"

"Try @Tesla self-driving. If you haven’t done so or you tried an earlier version, try it again now. It’s awesome 😎 Great work by the @teslasi team!"

"Cybercab, which has no pedals or steering wheel, starts production in April"

"This is before supervised FSD is approved in China. Limiting factor is production output in Shanghai."

"order of magnitude safer than um than humans driving"

"they're actually buying Tesla full self driving with a car attached as opposed to a car with fsd."

"FSD Supervised approved in Slovenia 🇸🇮"

"We redesigned Cybercab production for production over 5 times faster than conventional automotive manufacturing!"

"we want to make sure that we do not have an accident"

"Robotaxi operating hours moved from 10pm to 11pm. The main thing we’re trying to solve is making sure that we don’t run over pets when they’re hard to see at night. Literally trying to avoid grey kittens on grey tarmac in the dark."

"Tesla FSD feels like magic"

"Tesla self-driving is magical"

"Cybercab is scaling"

"Same thing happened with Tesla FSD. The model ate more and more of the C++ code, until there was nothing left."

"Robotaxi improvements"

"The cybercab fleet in Austin, for example, has grown by a factor of 10 since launch about a month ago."

"It really is an extremely fast week over week growth in autonomous miles."