I convinced two family member to buy Tesla, one was set on a RAV4.. I'm 100% sure. FSD will drive me around being essential to my life, in addition, think about all the disabled and too old to drive people, it's truly a game changer
X5 to x10 more miles in between interventions. So if as an example it’s 1 intervention per 1000 miles then it would be one intervention per 5000-10 000 miles. It has to be an other 9 on the march of nines.
I am a believer of FSD. As I get older my attention span is terrible. At age 26, I was a race car driver on famous tracks like Sears Pt. Raceway and Laguna Sega. Idea that I would go into a turn like turn 11 at Sears Pt Race making a almost U turn on the track was crazy enough. Before the approach of the turn I am coming in a 145 mph and I would find my favorite make on the road to begin stepping on the brakes. My Datsun 510 full race car would violently jump up and down the track and slide sideways both left and right. From the pictures my brother took, smoke was happening on all 4 wheels. So I am making a inside turn with another race car driver on my left. It's a bit crazy so for a second I win moving ahead of a driver that was ahead of me. Only for 2 seconds, the other driver surrender to me cause I am on the inside part of the U turn. I am no longer 26 and my age is 74, I don't have what it takes. I been thinking of buying a Tesla Roadster and take it onto the track, not for racing but a solo event. I not sure if I have what it takes to drive like my former skills as a race car driver. This will be intersting when I buy this roadster. I promise to have hired help to video the event. Maybe I can teach people how to drive a track. I am a good driver having set the lap record on both Sears Pt Raceway and Laguna Sega in my class. It will be fund to do it one more time I quit racing in 1979 so that over 44 years
Another way to think of it: 1/5th to 1/10th the number of interventions over any arbitrary number of miles. Or also 80% to 90% fewer interventions. It also means you have to be alert 20% to 10% of the time that you used to in order to prevent a mistake.
@@Teddy_M85 I think the better it drives, the more the nagging to stay alert is needed. People think that just because it has driven safely the last few hours, it's safe to look away.
@@Will-be-free Yes, this is something I'm concerned about! How are they going to deal with this? It's driven perfectly for days - many miles and then all of a sudden, out of nowhere, while you've been lulled to complacency, it tries to kill you.
Been saying for 2 years that FSD autonomous will first be approved, released and adopted in China. There are far too many Tesla haters in the current US government and legacy companies. Seems to be a non brainer. Maybe it will be almost simultaneous now, as the US also doesn't want to fall behind.
80% of my disengagements are mood related. A mix of: I want to speed and enjoy booting the car around; We’re late going somewhere and have no room for patience; My wife’s anxiety from a car accident before the tesla
I also want to comment on the 5x to 10x less interventions probably being accurate. For instance, if you have a route that has a road with 4 huge potholes and the car can't avoid them, you have to take over. That could be 4 interventions for one drive. But if the car learns to avoid them, then that drive goes from 4 to 0 interventions. These edge cases are quickly being addressed and I see the light at the end of the tunnel. Right now, they're laser focused on these details. At this point, 5x to 10x is likely very soon.
Telling that this announcement had no impact on the share price. That's because for the large majority of users FSD has little value until it can operate unsupervised with Tesla taking responsibilitty for any failures. If it still requires supervision, being5x, 10x or even 100X better doesn't change anything.
If it turns out to be a 5x improvement when it’s tested by the driving public then I think at that point the stock price will rise. If large incremental improvements are made with each release then it will be reflected in the stock price. The problem so far has been that FSD has still been quite underdeveloped and not sufficiently capable to give investors and Wall Street the confidence that it will be solved. Time will tell if Elon is not just over hyping 12.4
Regarding safety on city streets vs human, there hasn’t been a single death on supervised FSD. I so guess it’s are hard to quantify zero vs a human right now.
5 to 10x I think if you take FSD to work and there's a chance of 1 or 2 interventions today, the update will mean your possible interventions will now take a week. 1 FSD intervention to work. Becomes 1x5 day work week is (5x). If round trip, 1x10 trips is 10x improvement for the same week.
What makes these guys experts? Remember to do your own thinking. FSD has many things to workout and we will only know if it's 5-10x better once released and battle tested in real world usage.
Just received the free trial as my MY was on the 12.8.7 branch. Have used FSD 12.3.6 for the past 4 days with no less than 3 critical interventions every time. The trips were all less than 15 miles. Hesitation, then going when it's too late on unprotected left turns, had to take over & steer into the shoulder on a right turn to avoid being hit by oncoming traffic, veering to the off ramp going 75 mph & having to force it back into the lane, braking way too late for traffic in front & slamming the brakes to avoid a collision, wo dering jnto the other lanes when turning where there are multiple turning lanes.....
I get the feeling it is trained to work well in LA. Need more training for the rest of the country. Elon said 1 mile in one million miles of driving is useful for training.
good debates on FSD - regarding tick rate, it doesn't matter because this is a moving target. You could wait until the next 2 improvements or you could wait for when you're going on a trip where you want to use FSD. The data on tick rate maybe is relevant if you look at annual tick rate, 2023 vs 2024 vs 2025. It doesn't matter one month vs another month when we're in the middle of these big improvements.
Today’s guest seemed to fall below your normal high bar, Herbert. Combative and speaking from emotion vs experience. Love the channel keep up the good work.
The cooperation with China reminds me of a good buddy in high school Wukuon Zhan. He was very kind, smart, athletic, reserved but social. The best kind of friend he was.
China wants the best self driving in country. Whit the hope of copying it. Maybe even getting banned computer chips into the country. Such a good friend. This is mutual. Tesla needs China and China needs Tesla, for a while.
Looking for: not running over curbs, not going straight in a left turn only lane, not turning on red when it's forbidden. I experienced all those myself with 12.3.x
It's practically impossible to train for every edge case since the number of weird scenarios drivers can encounter is basically endless. Until AI is able to actually reason, instead of using pure recognition/prediction, it won't be reliably safe unless it's geofenced to lower the chances of encountering edge cases.
@@TheSpartan3669 An edge case could be something like a life sized blow up cow falling off the lorry in front. I don't see how geofencing would help in this or 99% of edge cases. Humans don't need to know the roads to drive in foreign neighborhoods or cities.
FSD is the reason I am not even seriously considering any other brand of EVs. Probably only come close to temping, is the Caddy Lyriq, what a stunning vehicle but dealerships are such shit heads to deal with.
Elon withholding the most important statistic the number of miles per intervention. If we have that number we would know how many more 10x reductions to get to full unsupervised FSD.
Interventions are not created equal. Some are a matter of safety, some just a matter of style, some a matter of the car taking a different route than expected, some are plain mistakes (by the driver). A raw intervention rate won’t tell us much.
Interventions are not rare now, unless you are someone that doesn’t mind pissing off or confusing other drivers. I have to intervene on many left and right turns by pressing the accelerator. Two curves on my drive to work I disengage so FSD doesn’t turn on the turn signal. Etc. Maybe these seem nitpicky but I have to be comfortable with how my car is driving and I have to intervene to be comfortable. I’m going to say there are roughly 5 interventions on my 21 mile work commute. If Elon is right there will likely be no interventions on my commute so I’m hoping for that with 12.4.
When I first saw this, my mind went directly to RoboTaxi. As a photographer, I have a theory that is likely the case. The position of the camera is likely because the potential new camera's wont fit in a M3 for these tests. The camera in that window is likely a larger (more light sensitive) lens. There is also no reason other than what I'm saying that the M3 rear camera in the original location couldn't be used, so I think the trunk camera must also be larger and won't fit in the original M3 location. Same for the odd protrusion of the side camera. It's bigger and won't fit in the regular spot. If a RoboTaxi is to only drive itself, it needs eagle-eyes! Larger lenses see further in very dark situations, especially on the sides and back where there is no bright illumination. As a photographer, I know there is no substitute for a large glass "capture" area to deal with dark situations. I'm confident that RoboTaxi cameras will all be higher quality since there will be no backup human eyes. I would also be surprised if there wasn't at least one front bumper camera. That being said, it could be hardware 5 and may just need to be sized down for all new Tesla's moving forward.
What if instead of an intervention, the passenger can now talk to FSD (ala ChatGPT4o) and request that it take certain actions that results in the car taking modifying behaviour without an “intervention” by taking over the drive controls?
Next year we'll probably be having conversations with our Tesla car robot and the AI will be able to pick out how to improve its driving skills. Jeannine
let's think about robots for a minute... They should allow lease or rent robots for the home. We pay a monthly fee... they get training data. We get Optimus to mow the lawn to the shock of the neighborhood. They can make a hardware update and deliver it to you and they take the old one. It can download over the air software updates. Liability, the customer signs a waiver and acknowledges it is Beta testing. Some people will complain, many will gladly pay to get a test robot at home. Celebs upload video shorts.
oh, buying leaps... How about plan to buy 100 shares per week. When the price dips very low, buy 5 leaps with assorted expiration dates. In addition every now and then with a normal dip, also buy 1 leap. Leaps by definition is monthly or quarterly (not available at each month). You'd need 4 leaps per month to average 1 per week. Maybe if the price has gone up enormously you could be well funded for converting those leaps into shares... But if you can't you can't, you could cash some out to fund a weekly purchase. If the chart has gone into a phase change you could have lined up valuable leaps going into the future. When the chart climb slows down maybe you can open puts which you will plan to cash out to fund purchases of more leaps.
The improvement will be exponential but I doubt this 5-10x means its ready for lvl4+ let alone robotaxi and until that happens and regulators give their stamp of approval TSLA will not move significantly IMO... I do see the robotaxi release in limited areas and monitored by Tesla employees remotely like Waymo, until full regulation approval.
@@danharold3087 I suppose he'll be right eventually, but only Muskophiles report Musk's prediction of a 5-10x improvement as though it's a sure thing and wonder why journalists don't report Musk's latest pronouncement similarly. We'll know soon enough how many miles per intervention the next release of FSD produces. Why would anyone believe any number that's not empirical, much less a number from the most notoriously optimistic voice on FSD ever?
like safety belts they dont save 100% of lives, they are just way better. even if fsd doesnt save 100% of lives it will be way better. we need a ralph nader to drill home legistlation.
5x to 10x is simply multiples. Say you had 5 interventions in the last release, 5x improvement means 1 intervention in the new release. Reduce the interventions by 5 times.
5 to 10 times means the improvement is between 50% to 90%. The variance between the lower limit and the upper limit is way too much for the statement to provide any meaningful direction.
FSD update is very exciting. But I'm cheap and only driving a couple of thousand miles a year. So I wouldn't spend the money on it. As a Tesla investor I hope the take rate is very high and other OEMs start to sign licensing deals soon.
The entire liability problem goes away when you buy the self driving package and insurance from the same vendor. They will be responsible for everything. Is self driving reducing accidents the insurance part should decline.
iv been driving FSD for 3 years now and it was a waste of time up into about 4 mouth ago and it has been getting better however, it t can do a lot of thing a taxi can do that my tesla 3 can do i hope this 5, 10 xxxx make me feel a lot different about it and I can trust it to pull out of my parking spot and saf;y pick me up in front of my condo and bring me to the grocery store 1.5 miles away and park my car in the car port and shut down for the night. im a tesla bull and a stock holder and want him to do well its simply not ready as of yet im hoping very soon !! however hope is not a strategy let get it done asap.
Imho, my gut feeling tells me the first licensing agreement would be with byd. It doesn’t make any sense to do it with oems in the U.S. as they are backing out from EVs. Besides there aren’t that many of their EVs to move the needle. Maybe eventually they will realize they are being stupid.
There is nothing specific to FSD that ties it to an EV. An advanced ICE or hybrid that can turn the wheel (lane keeping) and traction control (actuate the brakes) should be a good fit. Not saying they will do so.
I was given a free 30 day trial of FSD in my six month old model Y last week. I have used it four times. It is unpleasant to say the least. Driving down a straight side road, the speed is always just slightly jerking up and down. Just enough to be irritating. My passenger said the right mirror came with in an inch of a garbage bin handle as we were going 20 MPH down the side street. At a stop sign the car takes all day to get up to where it can see. When a gap in traffic appears it just slowly starts out into the lane going such that you want to push the accelerator to get going before that gap in traffic closes. On the highway today it waited way too long to brake when the cars in front of me suddenly came to a halt. I slammed on the brake to keep a safe distance back. I'll check back in five years.
When an insurance company is willing to take liability is when you can argue it’s ready to be deployed. You can debate what date that will happen, but today, no insurance company (including Tesla) is willing to take liability. Let’s be realistic and say that happens in 2026.
Tesla insurance will be the first to take liability, and will the the only one I would buy insurance from. Because if there were an accident there is no debate. Tesla insurance covers it. That is so long as it is true Tesla insurance and not a company that paid to use the name.
I liked your comment until you put a year for when it should be ready. It's easy to say at current rate of improvement it will take a long time, but what if the improvement suddenly is 10 times faster? With the rate Tesla is building out their training computers, we are not looking at linear improvement rate.
Beta driving for 3 years, there has been improvement but still mostly frustrating to use on a daily basis. If I need to replace my brakes early it will be because of FSD. Beta drivers have been duped into paying $8000-$15000 to work for tesla as beta testers adding wear and tear on our cars.
@@surf-n-turf Slow turns through busy intersection, hard unnecessary braking during merges, lane change to middle lane just before needed to take exits, erratic speed adjustments, still phantom braking, proximity to parked cars and curbs is alarming at times...I could go on!
@surf-n-turf If you recall, they said FSD would be ready by the end of 2021, enticing many people to buy Tesla equipped with FSD, duped! My brother paid $5000 for FSD in 2019...$5k for a product still far from autonomous. Now his car has 150k miles and he can't transfer his software to a new car on his terms? What a con.
@@4lessstressI agree with you it has been a long frustrating wait. But it's not a con job. They are literally spending billions in compute power trying to figure this out... But is a hard problem... I dont pitty those that believe blindly without looking at the facts. Even tho I'm all in and a fan boy, I didn't buy FSD at 10k when I bought my car. Plenty of evidence showed it was still a few years away. Now it's useful enough to justify a 99/mo sub for most ppl. Definitely not for busy urban areas. IMO. I don't buy this 5-10x bs. I'm sure it will be an improvement but regulators still need to approve it for Lvl 4+ and. robotaxi use.
@Teddy_M85 I'm a believer, don't get me wrong, but frustrated as hell. The fact that they offered a one time FSD transfer to boost sales at the end of a quarter last year shows that they know its the right thing to do. But instead of doing the right thing by letting you transfer your FSD at anytime, they try to get you to trade in a car you've had for 2 years and spend another $50-$60k or more. You need to repurchase something that still isn't fully functional? That's a con to me. If there are other Beta drivers out there like me, they got Beta fatigued and use FSD less and less. It's not near ready, I live in Queens, NY :-/
1) Interventions (critical interventions, like avoiding collision, stopping at stop signs, getting to destination, etc.) may be rare places but here, and others, they are very frequent, like one per mile. 2) I don’t believe Elon’s star for one sec. Maybe there is a place it is doing 5-10x better, but most places? Naw. And I fully expect it will be almost no different here, as has been the case for the last 2 years 😢
Expect 5-10x better everywhere so you should experience one intervention every 5 or 10 miles. Soon you will have days of no interventions when they get to one every 50 miles. I think they need to get to 1 in 100,000 miles or better for Robotaxi.
Ummm not pulling over for an emergency vehicle IS a single reason why Robo Taxi can't be deployed today...also reversing and understanding special road signs. You pull into a town with a non standard sign saying bridge ahead closed... your fsd drives off an unfinished bridge. You die. Yeah,.
We have FSD free trial in Canada. We must have a really early version because I had some really bad interventions. Car stopped mid turn, tried to turn into a closed road, tried to turn left from the middle lane of a one way street. Odd things I didn’t expect this late in the game.
You guys sound high...lol 5 to 10x is 500 to 1000% better. so pick a number of miles between your own interventions, say 20. Now it could be 100 to 200 miles between interventions....on average.
Why are people even worried about take rate. It’s a subscription service and we know how easy it is to subscribe for something. When the product is ready millions will be on it. You have morons taking data if true from only 3500 people. Think about it, as easy as clicking a button, Tesla doesn’t have to bring in the cars and install the software and hardware it’s already in!
I am worried that 10X is meaningless. seems to me that is just one 9 in the march of 9's. I love tesla. Musk has done the unimaginable with reusable rockets and he has pushed the world towards EV's, but his words on autonomy are meaningless to me. after 6 years of it being just around the corner he is the last person I want to hear about autonomy from and especially because its always some vague tweet or comment. Waymo continues to have big problems with just being in a few places. Uber gave up. Now tesla is attacking it with ever greater amounts of compute. AI is draining the power grid. I am sorry, i just dont think we have the right approach. if the current software is draining the power grid I am just convinced that the way we are going about it might not be the right path. tesla needs to show some better data and get a spokesperson who has the time to present some actual data. musk saying I am confident it will be feature complete by the end of the year. or the next update is really good. I remember when version 9 was supposed to solve autonomy. then each version after that was all hyped up. then it was 12 and now 12.4 and 12.5 are supposed to be even better. The problem is there is a neverending number of corner cases. I would really love to see tesla be able to bring down the cost of its cars. I want to know how much tesla is spending per car sold on autonomy. if its 10 billion a year that is 5,000 per car assuming 1,8000,000 cars sold. is autonomy a fool's errand right now? that question needs to be asked more.
Before Elon came along the closest I got to FSD was trying to stay awake while my wife is driving.
I convinced two family member to buy Tesla, one was set on a RAV4.. I'm 100% sure. FSD will drive me around being essential to my life, in addition, think about all the disabled and too old to drive people, it's truly a game changer
Just got back from 2800 mile trip in my MS, 60% on FSD 12.3.6 and it is definitely a game changer. Can't wait to get 12.4
I like Nick… he gives his honest objective opinion. Goes against the grain
Goes against reality.
X5 to x10 more miles in between interventions. So if as an example it’s 1 intervention per 1000 miles then it would be one intervention per 5000-10 000 miles. It has to be an other 9 on the march of nines.
I am a believer of FSD. As I get older my attention span is terrible. At age 26, I was a race car driver on famous tracks like Sears Pt. Raceway and Laguna Sega. Idea that I would go into a turn like turn 11 at Sears Pt Race making a almost U turn on the track was crazy enough. Before the approach of the turn I am coming in a 145 mph and I would find my favorite make on the road to begin stepping on the brakes. My Datsun 510 full race car would violently jump up and down the track and slide sideways both left and right. From the pictures my brother took, smoke was happening on all 4 wheels. So I am making a inside turn with another race car driver on my left. It's a bit crazy so for a second I win moving ahead of a driver that was ahead of me. Only for 2 seconds, the other driver surrender to me cause I am on the inside part of the U turn.
I am no longer 26 and my age is 74, I don't have what it takes. I been thinking of buying a Tesla Roadster and take it onto the track, not for racing but a solo event. I not sure if I have what it takes to drive like my former skills as a race car driver. This will be intersting when I buy this roadster. I promise to have hired help to video the event. Maybe I can teach people how to drive a track.
I am a good driver having set the lap record on both Sears Pt Raceway and Laguna Sega in my class. It will be fund to do it one more time
I quit racing in 1979 so that over 44 years
In Nashville and Charlotte I still get safety mistakes regularly. I’m a huge Tesla fan and a stockholder
Another way to think of it: 1/5th to 1/10th the number of interventions over any arbitrary number of miles. Or also 80% to 90% fewer interventions. It also means you have to be alert 20% to 10% of the time that you used to in order to prevent a mistake.
Hopefully, ppl will stay alert. It's easy to get complacent already.
@@Teddy_M85 I think the better it drives, the more the nagging to stay alert is needed. People think that just because it has driven safely the last few hours, it's safe to look away.
@@Will-be-free Yes, this is something I'm concerned about! How are they going to deal with this? It's driven perfectly for days - many miles and then all of a sudden, out of nowhere, while you've been lulled to complacency, it tries to kill you.
Great point Nick on finishing a puzzle… fewer pieces at end to resolve!!!
Been saying for 2 years that FSD autonomous will first be approved, released and adopted in China. There are far too many Tesla haters in the current US government and legacy companies. Seems to be a non brainer. Maybe it will be almost simultaneous now, as the US also doesn't want to fall behind.
Good call. Might be what we need to get over these political idiots here.
China is far more supportive of it's electric automakers. This is why Elon went to China to build the first gigaFactory.
Too much left wing extremists again Elon . Here in China 50% EV on the road - air quality significantly improved compared to to 5yr ago
80% of my disengagements are mood related. A mix of: I want to speed and enjoy booting the car around; We’re late going somewhere and have no room for patience; My wife’s anxiety from a car accident before the tesla
I also want to comment on the 5x to 10x less interventions probably being accurate. For instance, if you have a route that has a road with 4 huge potholes and the car can't avoid them, you have to take over. That could be 4 interventions for one drive. But if the car learns to avoid them, then that drive goes from 4 to 0 interventions. These edge cases are quickly being addressed and I see the light at the end of the tunnel. Right now, they're laser focused on these details. At this point, 5x to 10x is likely very soon.
15:00 perhaps its the camera position of an other car, from another manufacturer, and they test if fsd would run with their camera setup.
Nick is fast becoming my favorite commentator on Tesla.
5x means drive 5 times more miles per intervention. That's 500%
10x means drive 10x more miles per intervention - 1000%
I’m still getting interventions on every drive so looking forward to 12.4!!!
8/8 is Elon’s TRIPLETS birthday not twins! Sorrow to nitpick
Yes right! Thanks
For a big improvement navigation needs improving, speed limits need to be amped correctly, reading the signs only works some of the time.
Telling that this announcement had no impact on the share price.
That's because for the large majority of users FSD has little value until it can operate unsupervised with Tesla taking responsibilitty for any failures.
If it still requires supervision, being5x, 10x or even 100X better doesn't change anything.
If it turns out to be a 5x improvement when it’s tested by the driving public then I think at that point the stock price will rise. If large incremental improvements are made with each release then it will be reflected in the stock price. The problem so far has been that FSD has still been quite underdeveloped and not sufficiently capable to give investors and Wall Street the confidence that it will be solved. Time will tell if Elon is not just over hyping 12.4
Regarding safety on city streets vs human, there hasn’t been a single death on supervised FSD. I so guess it’s are hard to quantify zero vs a human right now.
5 to 10x
I think if you take FSD to work and there's a chance of 1 or 2 interventions today, the update will mean your possible interventions will now take a week.
1 FSD intervention to work. Becomes 1x5 day work week is (5x). If round trip, 1x10 trips is 10x improvement for the same week.
I enjoyed the debate
I love that someone is testing so thoroughly and charting it @46:47
As interventions reduce, solving the edge cases become easier.
In a way, but we're also coming up against harder edge cases to solve. Cases that require more human like understanding.
Thank you , very good report
Robots don't need mirrors. They are testing for optimal camera positions for cybercab
Impressive progress
What makes these guys experts? Remember to do your own thinking. FSD has many things to workout and we will only know if it's 5-10x better once released and battle tested in real world usage.
Just received the free trial as my MY was on the 12.8.7 branch. Have used FSD 12.3.6 for the past 4 days with no less than 3 critical interventions every time. The trips were all less than 15 miles. Hesitation, then going when it's too late on unprotected left turns, had to take over & steer into the shoulder on a right turn to avoid being hit by oncoming traffic, veering to the off ramp going 75 mph & having to force it back into the lane, braking way too late for traffic in front & slamming the brakes to avoid a collision, wo dering jnto the other lanes when turning where there are multiple turning lanes.....
I get the feeling it is trained to work well in LA. Need more training for the rest of the country. Elon said 1 mile in one million miles of driving is useful for training.
good debates on FSD - regarding tick rate, it doesn't matter because this is a moving target. You could wait until the next 2 improvements or you could wait for when you're going on a trip where you want to use FSD. The data on tick rate maybe is relevant if you look at annual tick rate, 2023 vs 2024 vs 2025. It doesn't matter one month vs another month when we're in the middle of these big improvements.
What’s 100% vs 5x Nick?
Today’s guest seemed to fall below your normal high bar, Herbert. Combative and speaking from emotion vs experience. Love the channel keep up the good work.
Not even. The guest just wasn’t willing to play the nit-picking game.
The cooperation with China reminds me of a good buddy in high school Wukuon Zhan. He was very kind, smart, athletic, reserved but social. The best kind of friend he was.
China wants the best self driving in country. Whit the hope of copying it. Maybe even getting banned computer chips into the country.
Such a good friend. This is mutual. Tesla needs China and China needs Tesla, for a while.
Looking for: not running over curbs, not going straight in a left turn only lane, not turning on red when it's forbidden. I experienced all those myself with 12.3.x
I won't use the free month anymore my car is freaking out one too many times it was doing some really stupid stuff
Slowly at first and then all at once - the nature of exponentials. I think 8.8 will be the launch of robotaxi service somewhere.
If edge case A. by its very nature is rare, how do you collect enough of it to train on?
Repeat it artificially?
It's practically impossible to train for every edge case since the number of weird scenarios drivers can encounter is basically endless. Until AI is able to actually reason, instead of using pure recognition/prediction, it won't be reliably safe unless it's geofenced to lower the chances of encountering edge cases.
@@TheSpartan3669 An edge case could be something like a life sized blow up cow falling off the lorry in front. I don't see how geofencing would help in this or 99% of edge cases. Humans don't need to know the roads to drive in foreign neighborhoods or cities.
FSD is the reason I am not even seriously considering any other brand of EVs. Probably only come close to temping, is the Caddy Lyriq, what a stunning vehicle but dealerships are such shit heads to deal with.
Unless US legacy sheds or buys out their dealers they are toast.
If the metric is true... all the layoff and drastic move to fsd makes sense. Makes sense to reorganise the company
Elon withholding the most important statistic the number of miles per intervention. If we have that number we would know how many more 10x reductions to get to full unsupervised FSD.
Interventions are not created equal. Some are a matter of safety, some just a matter of style, some a matter of the car taking a different route than expected, some are plain mistakes (by the driver). A raw intervention rate won’t tell us much.
@@peterfireflylund Better than nothing
The amount of time you guys spent at the beginning just going over examples of what 5 to 10 x actually means is pretty hilarious
I feel pretty safe driving even when FSD isn't on in my GT.
Interventions are not rare now, unless you are someone that doesn’t mind pissing off or confusing other drivers. I have to intervene on many left and right turns by pressing the accelerator. Two curves on my drive to work I disengage so FSD doesn’t turn on the turn signal. Etc. Maybe these seem nitpicky but I have to be comfortable with how my car is driving and I have to intervene to be comfortable. I’m going to say there are roughly 5 interventions on my 21 mile work commute. If Elon is right there will likely be no interventions on my commute so I’m hoping for that with 12.4.
I won't use the free month anymore my car is freaking out one too many times it was doing some really stupid stuff
Might get to Robo ready by FSD 12.7
U are not nitpicking these are valid concerns
@@user-xj5xp6qz5g fake acct
First from Azerbaijan ❤❤❤
I don’t think they will launch 8-8, but will reveal the taxi, so they can start on road testing shortly after, and the launch in a city q2 next year.
5-10x that’s like gotta be at least 100% better.
Yea bro!!!
When I first saw this, my mind went directly to RoboTaxi. As a photographer, I have a theory that is likely the case. The position of the camera is likely because the potential new camera's wont fit in a M3 for these tests. The camera in that window is likely a larger (more light sensitive) lens. There is also no reason other than what I'm saying that the M3 rear camera in the original location couldn't be used, so I think the trunk camera must also be larger and won't fit in the original M3 location. Same for the odd protrusion of the side camera. It's bigger and won't fit in the regular spot. If a RoboTaxi is to only drive itself, it needs eagle-eyes! Larger lenses see further in very dark situations, especially on the sides and back where there is no bright illumination. As a photographer, I know there is no substitute for a large glass "capture" area to deal with dark situations. I'm confident that RoboTaxi cameras will all be higher quality since there will be no backup human eyes. I would also be surprised if there wasn't at least one front bumper camera. That being said, it could be hardware 5 and may just need to be sized down for all new Tesla's moving forward.
What if instead of an intervention, the passenger can now talk to FSD (ala ChatGPT4o) and request that it take certain actions that results in the car taking modifying behaviour without an “intervention” by taking over the drive controls?
Next year we'll probably be having conversations with our Tesla car robot and the AI will be able to pick out how to improve its driving skills. Jeannine
Pump, 10, 20x - pump pump pump!
you need the approxal of one person to FSD a thing in China, 50 states 50 auths mininum for rollout in US
let's think about robots for a minute... They should allow lease or rent robots for the home. We pay a monthly fee... they get training data. We get Optimus to mow the lawn to the shock of the neighborhood. They can make a hardware update and deliver it to you and they take the old one. It can download over the air software updates. Liability, the customer signs a waiver and acknowledges it is Beta testing. Some people will complain, many will gladly pay to get a test robot at home. Celebs upload video shorts.
oh, buying leaps... How about plan to buy 100 shares per week. When the price dips very low, buy 5 leaps with assorted expiration dates. In addition every now and then with a normal dip, also buy 1 leap. Leaps by definition is monthly or quarterly (not available at each month). You'd need 4 leaps per month to average 1 per week. Maybe if the price has gone up enormously you could be well funded for converting those leaps into shares... But if you can't you can't, you could cash some out to fund a weekly purchase. If the chart has gone into a phase change you could have lined up valuable leaps going into the future. When the chart climb slows down maybe you can open puts which you will plan to cash out to fund purchases of more leaps.
I won't use the free month anymore my car is freaking out one too many times it was doing some really stupid stuff
The improvement will be exponential but I doubt this 5-10x means its ready for lvl4+ let alone robotaxi and until that happens and regulators give their stamp of approval TSLA will not move significantly IMO... I do see the robotaxi release in limited areas and monitored by Tesla employees remotely like Waymo, until full regulation approval.
What do you call a "puzzle" in your country?
In Scotland, we call it a Jigsaw. 😎👌
Elon says 5-10x better, and you can take that to the bank ... 'cause when has Elon ever been too optimistic about FSD?
Have you watched the FSD drive videos over the years. With all the additional training computers I give him a 50% chance of being right this time.
@@danharold3087 I suppose he'll be right eventually, but only Muskophiles report Musk's prediction of a 5-10x improvement as though it's a sure thing and wonder why journalists don't report Musk's latest pronouncement similarly. We'll know soon enough how many miles per intervention the next release of FSD produces. Why would anyone believe any number that's not empirical, much less a number from the most notoriously optimistic voice on FSD ever?
It better be damn good for all those FSD doubters to finally adopt the idea that it's a feasible part of the business.
like safety belts they dont save 100% of lives, they are just way better. even if fsd doesnt save 100% of lives it will be way better. we need a ralph nader to drill home legistlation.
5x to 10x is simply multiples. Say you had 5 interventions in the last release, 5x improvement means 1 intervention in the new release. Reduce the interventions by 5 times.
5 to 10 times means the improvement is between 50% to 90%. The variance between the lower limit and the upper limit is way too much for the statement to provide any meaningful direction.
Then you can read it as. At least 50% and die happy. psst: 5 to 10 times is inclusive.
FSD update is very exciting. But I'm cheap and only driving a couple of thousand miles a year. So I wouldn't spend the money on it. As a Tesla investor I hope the take rate is very high and other OEMs start to sign licensing deals soon.
what price per month would you pay?
The entire liability problem goes away when you buy the self driving package and insurance from the same vendor. They will be responsible for everything.
Is self driving reducing accidents the insurance part should decline.
S&P 500 and Nasdaq close at All Time Highs while Tesla stock closed at 58% below its All Time High 😢😮 There is something wrong about Tesla
Amigo cadê a tradução não tenho como compreender
It needs to be able to handle emergency vehicles in order to be robotaxi
This weird camera placement is tested for other cars with licenced fsd in my opinion.
"Humans do that all the time" @47:20...the point is for FSD to be multiples better than humans.
A 5X to 10X is absolutely required to make the progress needed to achieve robotaxi functionality not in the distant future..
20x
iv been driving FSD for 3 years now and it was a waste of time up into about 4 mouth ago and it has been getting better however, it t can do a lot of thing a taxi can do that my tesla 3 can do i hope this 5, 10 xxxx make me feel a lot different about it and I can trust it to pull out of my parking spot and saf;y pick me up in front of my condo and bring me to the grocery store 1.5 miles away and park my car in the car port and shut down for the night.
im a tesla bull and a stock holder and want him to do well its simply not ready as of yet im hoping very soon !! however hope is not a strategy let get it done asap.
Been driving FSD since 2018. Has never been a waste of time!!
12.4 to the public by mid June?
I can only imagine where FSD will be by 8/8...12.6?
Imho, my gut feeling tells me the first licensing agreement would be with byd. It doesn’t make any sense to do it with oems in the U.S. as they are backing out from EVs. Besides there aren’t that many of their EVs to move the needle. Maybe eventually they will realize they are being stupid.
There is nothing specific to FSD that ties it to an EV. An advanced ICE or hybrid that can turn the wheel (lane keeping) and traction control (actuate the brakes) should be a good fit. Not saying they will do so.
@@danharold3087 thank you for this response, I always wondered about this.
12.5 MIGHT nail it???
Maybe 12.6
Why would Elon say 8/8...should have waited until 12/8
should spread the good news abt Baidu n Tesla to everyone. no one knows this yet. this is a huge news!
Please have a show about how much money us Bulls on Tesla have LOST
I was given a free 30 day trial of FSD in my six month old model Y last week. I have used it four times. It is unpleasant to say the least. Driving down a straight side road, the speed is always just slightly jerking up and down. Just enough to be irritating. My passenger said the right mirror came with in an inch of a garbage bin handle as we were going 20 MPH down the side street. At a stop sign the car takes all day to get up to where it can see. When a gap in traffic appears it just slowly starts out into the lane going such that you want to push the accelerator to get going before that gap in traffic closes. On the highway today it waited way too long to brake when the cars in front of me suddenly came to a halt. I slammed on the brake to keep a safe distance back. I'll check back in five years.
Check back in July. You will be amazed
This guy is mental. It is not safe enough for tobotaxi. How about trying to take a 90 degree exit at 75 mph?
When an insurance company is willing to take liability is when you can argue it’s ready to be deployed. You can debate what date that will happen, but today, no insurance company (including Tesla) is willing to take liability. Let’s be realistic and say that happens in 2026.
Tesla insurance will be the first to take liability, and will the the only one I would buy insurance from. Because if there were an accident there is no debate. Tesla insurance covers it. That is so long as it is true Tesla insurance and not a company that paid to use the name.
I liked your comment until you put a year for when it should be ready. It's easy to say at current rate of improvement it will take a long time, but what if the improvement suddenly is 10 times faster? With the rate Tesla is building out their training computers, we are not looking at linear improvement rate.
I hope you’re right and I’m wrong.
'It reverses all the time. They just haven't allowed it to reverse'. What exactly he is saying there? Is he locked in a cell?
I love elon.
😎
It is not going 20 in School zones
Wait, cnBS and the big short😂
Stock down
Buy
8/8 Chinese robocab will be launched. Jeannine
Who cares how many people subscribed - BOOM
Beta driving for 3 years, there has been improvement but still mostly frustrating to use on a daily basis. If I need to replace my brakes early it will be because of FSD. Beta drivers have been duped into paying $8000-$15000 to work for tesla as beta testers adding wear and tear on our cars.
Specifically, what problems and interventions have you reported to Tesla Engineers?
@@surf-n-turf Slow turns through busy intersection, hard unnecessary braking during merges, lane change to middle lane just before needed to take exits, erratic speed adjustments, still phantom braking, proximity to parked cars and curbs is alarming at times...I could go on!
@surf-n-turf If you recall, they said FSD would be ready by the end of 2021, enticing many people to buy Tesla equipped with FSD, duped! My brother paid $5000 for FSD in 2019...$5k for a product still far from autonomous. Now his car has 150k miles and he can't transfer his software to a new car on his terms? What a con.
@@4lessstressI agree with you it has been a long frustrating wait. But it's not a con job. They are literally spending billions in compute power trying to figure this out... But is a hard problem... I dont pitty those that believe blindly without looking at the facts. Even tho I'm all in and a fan boy, I didn't buy FSD at 10k when I bought my car. Plenty of evidence showed it was still a few years away. Now it's useful enough to justify a 99/mo sub for most ppl. Definitely not for busy urban areas. IMO. I don't buy this 5-10x bs. I'm sure it will be an improvement but regulators still need to approve it for Lvl 4+ and. robotaxi use.
@Teddy_M85 I'm a believer, don't get me wrong, but frustrated as hell. The fact that they offered a one time FSD transfer to boost sales at the end of a quarter last year shows that they know its the right thing to do. But instead of doing the right thing by letting you transfer your FSD at anytime, they try to get you to trade in a car you've had for 2 years and spend another $50-$60k or more. You need to repurchase something that still isn't fully functional? That's a con to me.
If there are other Beta drivers out there like me, they got Beta fatigued and use FSD less and less. It's not near ready, I live in Queens, NY :-/
Wish we had teslas in the country of Elons birth
So 12.4.0 is ~7.5x better than 12.3.0 which is 5x worse than 12.3.6... Gotcha, E...
It needs to be able to send your car to a charging station and charge on its own
Stop and charge on its way to the dog groomer.
1) Interventions (critical interventions, like avoiding collision, stopping at stop signs, getting to destination, etc.) may be rare places but here, and others, they are very frequent, like one per mile.
2) I don’t believe Elon’s star for one sec. Maybe there is a place it is doing 5-10x better, but most places? Naw. And I fully expect it will be almost no different here, as has been the case for the last 2 years 😢
Expect 5-10x better everywhere so you should experience one intervention every 5 or 10 miles. Soon you will have days of no interventions when they get to one every 50 miles. I think they need to get to 1 in 100,000 miles or better for Robotaxi.
Ummm not pulling over for an emergency vehicle IS a single reason why Robo Taxi can't be deployed today...also reversing and understanding special road signs.
You pull into a town with a non standard sign saying bridge ahead closed... your fsd drives off an unfinished bridge. You die. Yeah,.
Tesla needs to release a 25k car for more adoption. COuld be a money pit too I guess.
We have FSD free trial in Canada. We must have a really early version because I had some really bad interventions. Car stopped mid turn, tried to turn into a closed road, tried to turn left from the middle lane of a one way street. Odd things I didn’t expect this late in the game.
Great job Nick. Herbert always like to cut off guests and has not good points other than other peoples opinions
You guys sound high...lol
5 to 10x is 500 to 1000% better. so pick a number of miles between your own interventions, say 20.
Now it could be 100 to 200 miles between interventions....on average.
These people can’t do math, they must be banned from RUclips, and never allowed to be an influencer again.
American education! These two don't grasp percentage.
And I'm told it's even worse now!
You need more Chinese in your country! 🙂🙂
Why are people even worried about take rate. It’s a subscription service and we know how easy it is to subscribe for something. When the product is ready millions will be on it. You have morons taking data if true from only 3500 people. Think about it, as easy as clicking a button, Tesla doesn’t have to bring in the cars and install the software and hardware it’s already in!
Take what Elon says with a grain of salt. He tends to embellish.
I am worried that 10X is meaningless. seems to me that is just one 9 in the march of 9's. I love tesla. Musk has done the unimaginable with reusable rockets and he has pushed the world towards EV's, but his words on autonomy are meaningless to me. after 6 years of it being just around the corner he is the last person I want to hear about autonomy from and especially because its always some vague tweet or comment. Waymo continues to have big problems with just being in a few places. Uber gave up. Now tesla is attacking it with ever greater amounts of compute. AI is draining the power grid. I am sorry, i just dont think we have the right approach. if the current software is draining the power grid I am just convinced that the way we are going about it might not be the right path. tesla needs to show some better data and get a spokesperson who has the time to present some actual data. musk saying I am confident it will be feature complete by the end of the year. or the next update is really good. I remember when version 9 was supposed to solve autonomy. then each version after that was all hyped up. then it was 12 and now 12.4 and 12.5 are supposed to be even better. The problem is there is a neverending number of corner cases. I would really love to see tesla be able to bring down the cost of its cars. I want to know how much tesla is spending per car sold on autonomy. if its 10 billion a year that is 5,000 per car assuming 1,8000,000 cars sold. is autonomy a fool's errand right now? that question needs to be asked more.
One full 9 is still significant