Tesla's Lead Is Unbeatable.
ฝัง
- เผยแพร่เมื่อ 27 ก.ย. 2024
- The Robots Are Coming www.farzadmesb...
FREE One Year Supply of Vitamin d3+k2 and 5 AG1 Travel Packs ➡ drinkAG1.com/f...
Get fit! Mention FARZAD when asking for a quote for a custom-built nutrition and fitness plan: www.teamevolve...
Join this channel to get access to perks:
/ @farzyness
Use my referral link to purchase a Tesla product ts.la/farzad69506
I worked at Tesla starting in July of 2017 as an Operations Analyst out of the Bethlehem, PA distribution center facility and left as a Program Manager based out of Austin, TX as of September 2021. I spent most of my time in the distribution and supply chain organizations.
Before Tesla, I was a Director of Business Intelligence and Pricing at the largest Pet Food & Supply distributor in the US, Phillips Pet Food & Supplies based out of Easton, PA. My wife and I also owned a small business in Bethlehem, PA between 2016 and 2019.
I have been a shareholder of Tesla since 2012 and currently own Tesla stock. Nothing I say constitutes as investment or financial advice.
My thoughts are my own and are not representative of everyone who currently works, or has worked at Tesla.
If you like this content I would greatly appreciate your likes and shares!
www.farzadmesbahi.com
www.linkedin.com/in/farzadmesbahi
Twitter @farzyness
#Tesla #elonmusk #twitter #tsla #TeslaNews #TeslaFSD #TeslaFSDBeta #teslastockanalysis #TeslaUpdate #investing #ElonMuskNews #teslanews #EV #FSD #investingtips
No One Understands Tesla’s Robotaxi’s
Tesla's AI Domination Has Begun.
HYPERCHANGE Is BACK For Tesla
Listen Up, Because HYPERCHANGE Is Back For Tesla
Gali Is HYPED For Tesla. But He Also Has DOUBTS
Gali. HYPED For Tesla or Calling BS?!?
Tesla's Lead Is Unbeatable.
so much fun thanks for having me Farzad!!!!
Our pleasure!
Great to see both Gali and Dave Lee back with their Tesla content! Both were there long before anybody knew what Tesla was about.
We've got to start working on Mauer now...
Gali and Dave are OGs. I started watching then in 2019.
Now we just need Rob back!
This is realy horrible,all the crazy greedy pumpers bullshitting the Public with disinformation!Shamefull!
If people understood robo taxi a used Tesla would not be available to buy
You nailed it my friend - which will go up 1st - used Tesla prices or the stock price…?
If more people believed owner-provided robotaxis are coming fairly soon. If they did then it's like a lot of people would be snatching up used Teslas and renting them out for enough to cover the payment.
100% correct!
Totally agree. Trying to find more excuses to get more teslas for that future….
The problem is FSD has barely progressed in the last years. It’s not even close to being a useable product.
Gali scheming is awesome.
22:07 - I have done that experiment, many times over, with my M Y FSD Supervised as an Uber driver. The results are that most people don’t even realize I am not driving. When you ask them they are typically positive on the subject, interested, excited, amazed. Some are hesitant (typically older). I have had one bad reaction, but that person was an irrational wacko anyhow and he was giving everyone one star ratings due to personal problems that had absolutely nothing to do with anyone but him. I did Uber for seven months or so, about 100 trips per month, and it was a ton of fun. The pay is too low to support a family, so I had to stop and go back to my other job, but it’s a fun less than part time gig.
Bet Tesla data centers will be the most energy efficient. Tesla will likely be working quietly toward grid independence in the background. All moving machines will eventually be autonomous and smart. What a time to be alive!
Check out Gali’s channel here: youtube.com/@hyperchangetv?si=7N3AJkP1Sh37iIQc
56:25 Re: FSD Jonesing. Ask anyone who uses FSD >80% how they feel after being suspended or whatever forces them to “relearn driving” for a week. Just as EVs are taking over, the current (forever improving) FSD will become as ubiquitous & just as irresistible. Look out for the exponential, when WS finally wakes up to this FACT. Forget regulatory approvals. What happens when regulators get a taste of “FSD Cough Medicine”! It’ll flip concerns for robotaxi acceptance on their heads! Think of the constant partying/entertainment traffic at all “off” hours! Think of the ads revenues generated by such “destinations” & to/from drinks/food services!! Imagine the new economies needed to support the new “ride to fun zones” businesses!!! F gang ridden public transport. The economic overflow will more than make those “desperation lifestyles” obsolete. F the FUD, just GO FOR IT!!!!
Farzad, I see your man cave with all your great gadgets and I wish to have just the dust of your desk .... however, we the 90% who believe in the company owns 100 shares and a couple Model Y's for ride sharing. It's not really fun to see the value of our investment dropping and just some of the chosen ones are so bullish over how many millions can be made when the replacement of drivers are indeed a reality...
Turtle mode the first 2 rides. Jersey mode after that
It definitely has to be called Jersey mode now. 😄
@11:25 "The research around this thing (the human brain) is not fully captured" Farzad. You are being kind - I'm a medical doctor with a strong interest in the human brain and an avid follower of all the scientific progress in understanding the human brain and I can tell you we are not even close to understanding how the hell it does what it does. And nobody anywhere on the planet can explain the mechanisms behind consciousness.
You’re so right on. All other autonomous driving focused on rules-based, secondarily on emulating human thinking. Tesla bypasses both by finding ML secret sauce to writing code for training video and bypass human thinking for driving by solving driving without trying to emulate human thinking. Now they can leverage much shorter code and lower compute to create AI bots.
You consider the last 80 years of brain research useless??
It is at the foundation of most all the work that Stanford, Carnegie Mellon, and MIT have done to create all this technology we are currently using successfully.
Sure there is lots to go, but it sure seems like you have decades of information to catch up on.
They will never really be able to explain the actual way consciousness is experiential and aware, as the phenomena of consciousness is embedded in the Universe. Best they can hope for is identifying a set of states of material and electrical or even subatomic particle activity for circumstances of consciousness but there is not a real answer anywhere.
Why do I walk fast? Why do I speed? Because I'd rather give my attention to something else but I have to focus on not running into things so let's get to our destination asap to free up my brain space for the thing I'd rather do. Robotaxis instantly solve this.
Gali was my LONG TIME Tesla Dude who introduced me to Rob and of course we all had to check out Dave Lee (the Sage of Tesla) before it was cool to have a Tesla channel When Gali left TH-cam full time, THANK GOODNESS FOR FARZAD! I love seeing this and hope we will get more great CONVERSATION AND IDEAS which organically erupt from great minds just thinking out loud and trying to break their own model. Thanks guys, This is classic Tesla content!
As an Australian, i pissed myself hard when Gali was talking about Aussies patting crocodiles, hahaha mind of a maniac analogy, love it 😂
Farzad, you think New Jersey is bad - take a taxi drive across Rome, Italy sometime. Traffic lights, lane markers, speed limits - are merely suggestions. I’ve never gotten out of an Italian taxi ride without having to reach down to break the seal between the seat and my ass.
100% As a 17yr old exchange student in 1970 , I can verify this as true, driving on the sidewalk, extreme competition on the road - madness.
I was with my Italian family - 5 adults in a Morris Minor (1100ccs - combined weight of passengers probably 800 lbs - the mother weighed 300 by herself) we went barreling into a dark tunnel from complete sunshine to find all traffic at a stop. The son did a good job of trying to stop the rolling mass but we still slightly ran into the cars ahead of us, what mayhem, EVERYONE got out of their cars to engage in the verbal battle and then it turned out we were on the trolley rails so everyone on the upcoming trolley got off to engage- I just stepped aside and shook my head.
I don't think Elon should be promoting breaking through the law through speeding, just because the mob is.
I've driven in Rome. I've driven in India. It's just a matter of learning what the local customs are and driving like a local. I'd much rather pass a slow moving truck on a two lane highway in India than in the US. The custom there is for the truck being passed to help the car passing it. And a truck in the other lane will slow to allow a safe pass. Just the opposite in much of the US.
Farzad! My man! Thank you for speaking up for the Jersey driver! We have been talking about a Jersey mode since I got FSD. FSD gets bullied in the Tri-state. Also, it has no clue about toll roads but I digress. Cheers to the family from the Garden state.
1:00:45 Pursuant to this thought of cities transforming once we have robotaxis-- I believe that one underappreciated aspect to the robotaxi revolution will be how much space will be freed up by the elimination of parking spaces and how much safer walking, bicycling and other micromobility devices will be when there are no more distracted drivers. I think that the parking spaces on the side of the street will become dedicated bike paths, separated from car traffic by physical barriers and that the future of transportation will be the combination of micromobility and robotaxis.
46:00 Re: FSD Jonesing. Ask anyone who uses FSD >80% how they feel after being suspended or whatever forces them to “relearn driving” for a week. Just as EVs are taking over, the current (forever improving) FSD will become as ubiquitous & just as irresistible. Look out for the exponential, when WS finally wakes up to this FACT. Forget regulatory approvals. What happens when regulators get a taste of “FSD Cough Medicine”! It’ll flip concerns for robotaxi acceptance on their heads! Think of the constant partying/entertainment traffic at all “off” hours! Think of the ads revenues generated by such “destinations” & to/from drinks/food services!! Imagine the new economies needed to support the new “ride to fun zones” businesses!!! F gang ridden public transport. The economic overflow will more than make those “desperation lifestyles” obsolete. F the FUD, just GO FOR IT!!!!
Chill needs to be chiller. Assertive needs to be more assertive. Sounds like it just needs a wider range.
Yes, a wider range and perhaps a slider right on the screen that the passengers can adjust anytime. For new riders, default to the chillest setting, thereafter it adjusts to whatever that person's last setting was set to
Galli. About your question of when will we see the first cyber taxi service on the street -- i believe that the 7 story multi parking lot is not for employees but for building a large stock of taxi cars out of the factory and get enough stock in big numbers to suddenly fill a single city with enough cars to provide a good experience of instant availability.
They could have a tesla taxi assistant on the screen live to talk to the new user for every new user that as never experience the robotaxi. It would help the user have a good first experience. Then if all good, you don't need to talk to the user next time around. Or replace the taxi assistant by ai after that.
Why in the vehicle?
You have that on your "Device" (Aka digital assistant/ "phone")
I'm a simple man. I see Gali and Farzad on the thumbnail, I click it.
I'm so glad y'all are having this discussion!!! 0:58
I think the way to solve the personalization of FSD when you first use it; will be easy once they integrate Grok, you can literally just tell Grok through maybe a dialogue about your preferences and adjust them on the fly by telling Grok hey can you slow down on your turns.
Gali's monorail tests were epic! It is good to see him on YT again!
On Gali’s “lack of accountability and lack of understanding”. Black box mystery intelligence, I do really think that having an integrated LLM as the interface to the car that we can use as direct communication. I would love to be able to tell the car to go around a slow person, or take an alternate route more known by locals, etc. being able to explain to the car and the car articulating choices back to you in conversation would be amazing
Kinda like the judge in Delaware. Mistaken, with authority to wreck your life, because they have authority....! A loose cannon.
Gali’s demo and intro video in the robo taxi is awesome. Non Tesla-ists need reassurance for sure. Friend were afraid just seeing the lack of dials in my Tesla. Young generation will be totally accepting - they born into the tech as magic age
He’s absolutely right about the speed being too quick and making passengers nervous. My mom loves the Tesla and wants me to drive full self driving every time we get in the car… But it does scare her sometimes! If you’re not holding the wheel or in control of the car, it just seems like it’s going too fast when it makes major transitions.
27:20- just black out all of the windows and put in a bigger TV screen: immersive home theater. People can watch a movie and then viola, they’re at their destination. People don’t even need to know or care to know what is happening outside of the vehicle. Same thing with a plane flight. No one but the pilot and copilot see or understand what is happening with a commercial plane flight, but they still get on it and they still land just fine.
People will be throwing up in the back seat !
I think you three have absolutely got it and are understanding the problems AND the solutions. I drive professionally and use second level thinking less than half the time and only because I am a professional driver. Scanning for potential edge cases, like deer or road hazards, is how I do it, but most drivers don’t. I believe FSD simply scanning without distraction will get us there… by June 25. 🤞
just a thought. what about releasing fsd fully and IF a corner case comes out and the internal computer cannot figure it out, for some reason, then have a ready employee dial in, very quickly, to solve it as a human? so car pulls over stops, asks for help from master base..:) and remote control fix.....should be very rare and doable if the corner cases are so rare.
Big problem I see now is traffic jams needing to get over 2 lanes to your exit.
Sometimes you have to force cars to yield way.
I have been using my MYLR with FSD for ride share and educating people on the safety and goodness of FSD. It’s very educating to know that some people are shocked and some just take it in stride as if it is inevitable. I am loving it so far
Someday hopefully there will be some sort of version to version autocorrect for things that have been solved in one version can be propagated into other versions. if it is a csv file, track changes and tweet those specific values in other versions? Sounds so easy. luckily the world has the amazing Tesla engineers to actually achieve this.
29:25 1. There will be enough customers.
2. We do need public approval.
3. 20% people will loose their jobs. Wages pressure will be huge.
HW3 was designed in 2017. That is mind blowing that there was so much insight in the compute required in the end.
Farzad. I think I speak for everyone when I say we love your videos and we are gonna need you to post that Sourdough recipe! 😊
Gali is hilarious lmao
I want to see Gali and Steven Mark Ryan. Devil’s advocate vs ultra bull
Gents… this was my fav episode to date. It’s fun to see you guys wrestling with the same things in wrestling with. So fun.
It will be most interesting to see how FSD handles the middle of Paris and the middle of Rome. Good discussion you guys. Thank you for airing it.
A Waymo drove by on my side street yesterday afternoon in Hollywood, CA. No driver, couldn't see if there was a passenger. But it LOOKED like Robby the Robot! You know, 50's sci fi! A Tesla looks cool. Cybertruck just looks otherworldly on a city street or on the freeway. Can't wait for Cybercab!
Clearly the smartest person in this room who thinks of the right questions to be asking at this time, is the guy who is holding his microphone.
I love that he’s holding that performance mic!🎤💯
I respect them all .. but actually think his questions are a little 2d ... Of Of course people are going to be frightened of it. Just imagine how people felt the first time they got on an airplane. Just my two cents... But I do think he adds tremendous value and the conversation would not be nearly as good without his perspective.
Hardware 3 was first installed in model 3s in mid May 2019
Embodied AI brains will be amazing but the distributed latent compute power in all the cars is beyond supercomputer.
I'm very interested in the day this starts being utilized
Galli.... it would be much more efficient to deliver cinnamon rolls by drone. ;-D ....change cinnamon roll to something too heavy for drones.
I feel the conversation surrounding FSD and how the perception of safety is important is a really interesting one. My opinion is that as the system becomes more robust they will likely implement an algorithm based on your riding preferences using a rating system to suit you personally, which will help solve the subjective nature of user experience.
The wild west at 21:06 with the arms flapping around was hilarious. Gali is great! Good episode!
The car doesn't learn, that's done in the center. But maybe the car could support that better.
The car should be able to, just like a human driver, recognize that it has made a mistake.
It would then upload that situation to the center, where it's used for improving it.
Human interventions wouldn't then be needed for FSD's learning. These will also become fewer anyway, when FSD isn't supervised anymore.
A few situations where FSD should recognize that it made a mistake:
;If another car drives out of turn because FSD is to slow to act.
If another car with the right of way reacts because FSD is a hinder.
It being honked!
It's in the wrong lane for the planned path.
It has to do drastic maneuvering to avoid something.
It doesn't let another car in, and it does something dangerous.
There's a lot of situations that FSD could identify as reasons for improving the driving.
The next problem would be, how does FSD go about finding out how to avoid those situations in the first place, and how to get out of them if it did get into them. That's driving intelligence!
I guess it's a while before FSD self-improve in this manner, thiugh, but if it did, it would become a super-driver!
40:01 .. amazing return.. thank you needed
46:21 hmm..
47:35
49:20 inferance is human mechanic.. that's human remote drivers.. not ai
😂😂 55:02 cinnamon rolls
Can listen to Gali Speak for ages
24:33 Sub-model training of FSD based on geographic driving style might be useful. NYC vs Denver vs Seattle VS Mumbai VS Australia. The world doesn’t drive the same way.
Remember how everyone used to ask about SAR radiation concerns about different type of cell phones? Does anyone ask about that anymore? Is it more safe now or have ppl just accepted it?
You don't generally toss the weights when you do a model update, you'll 95% of the time doing Fine Tuning with new data if you've invested in RLHF through the Tesla driver team
24:28 why does the car not have an aggressive and chill setting?
Awesome discussion. I learned a lot.
IMPORTANT Farzad! I have a friend who is a cyber security expert who told me she will never ride in a car controlled by software because it just a matter of time when you are in a car that just got hacked and instructed to do something scary. They can patch this but a LOT of damage could happen between the hack and the fix. This issue really needs to be addressed and WILL become an issue and could stall the cause significantly. Some type of failsafe needs to be developed to reassure riders/owners that this situation is safe. The answer is NOT how AI makes code that is hard for humans to "decode" and that updates happen "all the time". My friend said this code can be and is already hacked today ....using AI. Great show Farzad!
Tesla holds regular "hackathons" for hackers to try to hack their software, they give away a car or significant cash as prizes for successful hacks. Smart way to catch a hack before it is fully developed or implemented for evil. Everyone has a "friend" who says crazy stuff.
I've never heard of a neural net hack... yet.
How many othe Teslas have been hacked? Attempts...Yes.
Fact: Tesla pays hackers to find security issue. A few 6-figure payments and a couple of free cars later, haven't heard much!
Tesla regularly runs zero-day-exploit hacking challenge events. World's best hackers compete for generous prizes trying to hack the latest Tesla software.
@cathyk9197 hi Cathy. The fact that some can hack it every time is the issue. It will only take one or two gruesome accidents to kill all efforts. ...it shouldn’t but will. ...thanks to crap journalism.
@@RichardSalmore hi Richard. There needs to be more information readily available from tesla about this to reassure the public. So much FUD... you know the vultures are circulating to monetize Tesla’s failures... perceived or not.
It’s still remarkable to me when people come up and say we only need current devoted task, fixed position robots. Yes the human form is not as efficient for one task, it is also true it is very, very good for generalized tasks and this is why humanoid robots have such potential and also allows humans to better accept them. No one wants to work with or interact with alternatively a giant metal crab for instance.
Elon understands the Terminator side of this great Idea, the time of abundance. Let the bot take care of everthing. It starts with driving our car for us....................then whats next?
Thanks for an interesting video although I didn’t understand all the acronyms. I did enjoy the video and found it somewhat informative. A thought pass through my mind. Wonder if Elon would ever sell FSD to other companies considering he’s putting it in his robots as well. Sounds like not just cars in the future. This could be used theoretically in the farm fields at harvest time so that farmers would not have to work 24 seven during harvest season, perhaps one day or it could be used to run a sprayer and monitor field conditions, such as laying out Roll before planting. These are the things that go through my head as I watch these videos in the morning before I get to work although this doesn’t benefit the majority it benefits the very hard-working, who don’t have time to do other things in life such as vacations because they’re running the type of job that requires constant monitoring and work. Thanks for a great video again. and a thought provoking one at that.
Farzad, your counter points are just amazing. I just can’t see people not adopting this. Once the snowball starts it will be unstoppable. A whole new exponential curve, but an extremely steep one!!!
If AI achieves agency, then perhaps we can program skin in the game in which it will die (“pull the plug”) if it does something dangerous
If they make the cab super safe in a crash and probability of death in an accident it almost zero. People will take it all the time just for the safety aspects! If some how fsd fail the car are still crash safe for mortality, it’s a winner!
Great ideas dudes!
One more idea to throw out for you: How cool and safe it will be if you can summon a Robotaxi that can take your kids to school ? That will eliminate the fears for parents who don’t trust strangers driving them.
Great info. However, way too many commercials.
the thing i find incredible to think is in twenty years from now there will be people who have no idea how to drive a car and driving tests will no longer me done as nobody will need to know how.
Chill should be more mellowed. Assertive should be more assertive. 👍🏻
We need to clarify some parts here. Sorry for long text.
The Apple Neural Engine (ANE) has been present in iPhones since 2017. Just as a reference.
Teslas chipset is NOT in competition with NVIDIA. We have to get better understanding of when and how both of these products are needed and one cant replace the other. What you could argue is that Tesla could target marketshares from Intel or AMD (as it looks today). NVIDIA is years ahead of any other manufactue that process big data.
I also saw NVIDIAs pressconferance. He is spot on with his vision. He mention that in the future AI will write the code for e.g. Photoshop on the fly based on some fixed guidelines. Never again update any software. Always maximum security, The funny part is that I saw Microsoft AI PC pressconferance first. I was already thinking "thats it". Then NVIDIAs came as a suggestion. Watch both and think about how many years ahead they are. Microsoft show us paint AI. More or less make a ugly draft and it will show you suggestions what it could be. NVIDIA talk about that code isnt needed, that their GPUs will have new archetecture every year because they use AI to develop next gen. They stat 500% increased performance YoY. Thats numbers we never have seen. Normally you keep the archetecture for 6 years. Halft way you make a refresh. During those 6 years you get 30% better performance.
Everything the car uses when you activate autopilot is based and processed by the car. The Tesla computer is good, but it’s not even close to what NVIDIA can do. Even if you put an NVIDIA server in your Tesla, the battery would be gone very fast. These two are completely different things, but both are AI.
The statement about restarting the training data is not correct. That wouldn't make sense. What you experience is that, in order to fine-tune the training, you have to offset the value. The car will recognize that new data is better than the old (offset values). If you don't offset, the car will play it safe. As it should.
You misunderstand what the computer does in the cars. Of course, Elon has made sure that older cars will provide data to Tesla to improve. There is no difference between HW3 and HW4 when it comes to "decision making." A faster computer is not more "correct" than a slower one. It's the same data. What happens when you get updates for the car is just that added scenarios are being suggested. The reason why older cars can still benefit is because this kind of data processing is not heavy load. The code is constantly optimizing, and there are A LOT of pre-coded scenarios. More or less, the car just picks a pre-made script that fits the scenario. That's why it needs training. If it were capable of making everything on-the-fly, there would be no need for training.
@37:00 The scenario you put up is correct. Today we tell the module what we want, and it gives us an answer. This is processed on a remote server. To let the AI think "freely," we still need limitations. What NVIDIA has developed is an AI that you can ask for a "goal." Imagine you upload the drawings for a chipset and tell the AI to design a new one with double the performance. Then it gives you suggestions. This is what NVIDIA has done. This is the reason why NVIDIA presents two architectures with 12 months' space. That’s just insane. This calculation today takes time and a huge amount of compute. It's not even on the agenda to do that kind of calculation on a car, on the fly.
Letting the AI make completely independent decisions won’t happen. It's impossible for the car to manage that with current compute power. If you think about how many times a sharp right corner has been taken in a Tesla, the same amount of scenarios the "free" AI has to process to be on the same level. What you can do is add data for the AI to analyze. This is what they have been doing. On top of that, the cost of a chipset that can process this data would be very high. The cars would double in price.
If you remember, manufacturers didn’t get Elon's statement that Lidar is not the future. You get so much more detail from a camera. Lidar is a fast way to get somewhat self-driving up and running, but when a camera can recognize that there are wet leaves on the road making it slippery, what argument should the Lidar manufacturers use to justify less security for humans?
I would argue that Tesla could roll out FSD today, and it would work OK. Why should Elon wait? Because he has much more data than the competitors. Since we’re talking about a computer being in charge of your safety, only the best (safest) will be used as the bottom standard. In my experience, Tesla’s autopilot is much better than any other solution I have tested. The noticeable difference is not a good sign for other manufacturers. I would also say that if something happens when FSD is approved, it will be suspended right away, and it will take years to be reviewed again.
I highly doubt the "taxi" idea. Would you actually want random people to use your car that you bought? You don't give out keys to your neighbors when you buy a new car. It could work if you get half price on the car and it has to be available for X amount of hours per week. We, as humans, like to own the stuff that we buy.
The robots will certainly be a very good product as long as they are capable of doing all the boring stuff we do: washing clothes, cleaning the house, sorting drawers, suggesting purchase lists, and so on. This has to be done faster than you could manage yourself. You don't want a big robot making noise for several hours just to do two simple tasks. It’s like the first robot vacuum cleaner. The first week you think it’s cool, so you don't bother where the station is. After a few weeks, you only clean when you're not home, and you want to hide that ugly station. Your fancy robot will be cool for a while, then you’ll feel it’s always in the way. If it’s also slow, you will not use it.
I’ve been driving FSD 12.3.x now for a few weeks and i put in the address to where I want to drive to and it drives me to the front door with it any intervention, but being I know my area and I often would have driven a different more direct route because I know the route and FSD wants to take a different route. That is were I do the intervention because of the route and I explain that if why I intervened. What I am waiting for the FSD 12.4 that we get to choose a parking space and hand signals in traffic by others like people bikes etc is coming and I can’t wait!!!
6:08 "HW5" is called "AI.5", right?
_Elon Musk also officially announced Tesla’s FSD hardware 5.0, which he says Tesla is now calling “AI 5”. AI 5 is expected to hit the production lines for customer vehicles in approximately 18 months - around December 2025, with a massive slate of improvements. It is expected to be approximately 10x better than hardware 4.0, and up to 50x better in terms of inference power alone._
When Robotaxi operates in a jurisdiction, Tesla assumes all liability, barring exceptional circumstances such as the passenger interfering with the FSD system. In that case, in Australian jurisdictions, Tesla would probably be assigned points like any other driver, and would lose points on violations, such as speeding. Jersey driving in Canberra (55 km/h in a 40 zone, etc.) would see the system losing all points on the first day of operation.
No hyperchange showing in the Tesla stock. Still suppressed (for years)
How expensive is public transportation, bus, train and subway pr miles. How will Robota i compare?
Yes, totally too fast and aggressive for my wife. But I come from a racing background and I love it. It could even be more aggressive for me. Yes, they should have an adaptive mode almost like your comfort 'profile' for seat position and the likes.
It's a question of how safe it is for the government to make it legal. Passenger airplanes and cars were legal when by today's safety standards were horrendous. Countries will want to be the first to legalize it and accept failures to be ahead.
Very much agree on starting on chill mode. Have a voice command to increase aggression levels, but I don’t want a wild ride, I want chill.
It s a coastal thing, Seattle here as well and we are much more laid back than the aggression on the east coast
You will pay attention the g forces. That section seriously highlights the differences in chill levels of the coasts.
I don’t know if it’s the lack of glasses, the facial hair or just a more chill attitude but Gali looks and sounds all grown up
I need to see actual results for FSD where I live in the Upper Midwest in snowy conditions with icy roads.
Good discussion
I would have liked to watch the whole interesting discussion, but after 15 interruptions for commercials, it just ain't worth it.
These were great thought exercises. I love the challenging of the ideas and considering the potential failures. Why wouldn't you??
This is something I thought about previously. Given the speed of calculating outcomes, will this computer sacrifice one car to save many? As the computer is programmed to prevent crashes, will it also know how to make an inevitable crash save the maximum number of collisions? It’s inevitable that decisions will be made that are beyond our comprehension.
Hey, anyone know if Tesla gives any notice at all even a few days if they are going to lower/ raise prices on a certain model??? Looking to get a model X but wondering if they will drop prices..thanks
There will be a very easy way to understand a decision the AI has made if you don’t understand it … an explainer button that generates a written and/or spoken explanation of a decision. This would be very easy to implement.
You teach it to explain itself when there's an anomaly, and of course you have to teach it what an anomaly is that would be a deviation from the norm
THE BEST IS YET TO COME🔥🔥🔥🔥🔥🔥🔥
Would it be conceivable that Tesla could virtually combine the computing power of these many vehicles and robots (with hardware 5.0) and thus create a kind of "Vehicle (Compute) to Grid" system instead of a "Vehicle (Battery) to Grid" system. The battery capacity of the vehicles could be used to fill up with cheap electricity over lunchtime and instead of simply returning the electricity, the computing power of the overly powerful chips could be returned to the Internet.
I would like to discuss whether it could be strategically worthwhile for Tesla to strategically utilise these possibilities. Perhaps it doesn't make sense because a large, on-site data centre is much easier to set up and operate. But perhaps Elon Musk is once again way ahead of us in his strategic considerations and already sees the next profitable opportunity on the horizon. What do you think?
Need more Tesla cars on the road to run autonomous
Just a thought. FSD needs massive and fast compute since it literally does not know “what is round the next corner” and is responsible for lives … whereas Optimus lives in a mostly predictable and well scoped task and physical environment. So Optimus might well manage perfectly well with more modest compute needing much less power? Of course Optimus may move from task environment A to task environment B, necessitating a learned “program” change . It’s all very exciting but I think that Optimus might be easier to solve than FSD once the physical aspects are completed, especially given the flying start already provided by FSD experience?
Humanoid robots are NOT coming in your lifetime.
They are simply too dangerous. Ever see videos of them? Do you ever see a human standing in close range? The servos in robotic technology are VERY powerful to articulate everything correctly. In other words if one of these bots grabs you and squeezes by mistake, bones are broken. A Tesla employee has already been hurt by one. Around pets and children? Not in this lifetime holmes.
Happy to see Tim is selling his Tee Shirts successfully ....
Tim Dodd sells the highest quality tee shirts across the entire Muskonomy related economy🥰
We will have supervised robotaxi this year (in a few geographies). Having FSD driving me 99% of the time, with minor mistakes and no critical interventions (in CA), it is ready for supervised robotaxi - similar to how Waymo and Cruise deploy in new geographies. Then, in some geographies robotaxi will go driverless in 2025 based on data. The supervising drivers will be told to NEVER intervene unless an accident is imminent. Once Tesla will see ZERO interventions in specific area it will open it for driverless. This is going to happen faster than most expect.
Gali needs to learn how to play poker so that he can get on the All In podcast and become a bestie. The All In team needs an infusion of talent.
Honestly, when do you think FSD will meaningfully hit the bottom line? When do you think a Bot will be able to diagnose and fix a complex plumbing problem, or do a complete brake job on a car. This "displacing of human labor" will take decades if not longer. The human body and human movement is so utterly complex and advanced that until there are true genetically grown cyborgs, many nuanced human movements can't be replicated with a mechanically limited humanoid robot. The best we can hope for is that moving forward, everything new is designed from the ground up to be fixed using rudimentary movements. I fear the advancements in bots (similar to where we are currently at with FSD) will put us in an endless "hurry up and wait" loop with nothing but updated promises that never quite land a repeatedly require a new shift in architecture. When will the local maximums hit with this recent attempt at FSD. Its always just around the corner! What happens when it turns out the last 5% is 100x harder than the first 95%, then...oh wait, the last 1% will be 100x than the first 99%, then ...oh wait...each 9 to the right of the decimal is 100x harder than the last. People will soon realize what true exponential difficulty means.
Agreed… Theres no technology training called User Training Video as per a User Manual.
Driving on a NJ road now and listen Faz talk about NJ drivers lol 😂
“ tesla, come pick up my dog and bring her to the kennel for the day to play“. I need this. Especially today as I sit hear hungover 😅
Being tech people we all know those frustrating users who cant figure the simplest features in an app or crm or whatever and always wants someone to show them what to do. With ai, this should be more than doable. A grox assistant will sort it out. This also applies to tesla letting the world know how great they are.. if they dont want to do advertisements..do documentaries or Tomorrow's World, BBC Click, 60 seconds type shows.
I don’t think Tesla is late to the party setting up a data center at Austin as Gali says. Things are happening at the appropriate speed aka as needed.
Cinnamon Rolls with Raisins Bro. 😂 1:01:20
How many title changes, two? Three?
In regards to the P&L, in my opinion you either believe in what a company is trying to accomplish or you don't. Period. There's no other way to perceive it. You'll never have a crystal ball.
Is it just me or does Hans looking like a really really smart Amish guy?
Looks smart? He looks like a high school graduate from somewhere in Pennsylvania. So I get the Amish part. The smart part? No. He reminds me of those kids that don’t quite fit in anywhere, not smart enough to be a nerd, not athletic enough to be a jock, not cool enough to be a player, not disciplined enough to get through business college. He probably worked at autozone before he discovered his misfit comic book hero Elmo the clown. Now he dazzles the townsfolk with tails of H100 training clusters and has convinced himself he is a Tesla engineer. He uses the term “we” when is talking about Tesla. He is a true ride or die pumper.
Tesla is providing the "explanations" that are needed in a court of law, RIGHT ON THE SCREEN.
Those drawn pixels of cars, trucks, people, bicycles, curbs, cones, stop signs, stop lights, warning signs, ladders flying off of pickups tracked as unknown objects (but it knows where it is) are just like a real time MidJourney, but instead of /imagine prompts with text input to MidJourney - the car's internal neural network is asking to render those images. And drawing those images at 10 to 20 frames per second.
ALL those visual objects recognized hold up in a court of law. The things that aren't on that screen, but on the 8 video cameras are potentially what may be the issue in court, but otherwise you can see the blue line where the car is deciding to go.
I can year the complaints that we don't necessarily know WHY the car wants to go that way, but if Tesla can manage to create an output Neural Network for the screen, they can create more output NNs that give the basic driving laws that it is adhering to. It only needs to be taught to output those during training.
That is how we currently see the pixels on the screen. They didn't magically appear there on their own. And they don't have any direct part in controlling the steering, brakes, or acceleration of the car. They are output just like the steering and braking are, but the recorded video pixels are produced there to build trust with the humans - both in the car and in the courtroom.