You should talk more about and to Hotz. His team and his product works well! I think he will win big. (I have Comma 3x right now and it took my car from normal adaptive cruise control to full on super cruise with a bit of tinkering.
Go Jeff… a multi decade entrepreneur with a volcanic mind!! So lucky to have you… agree with Herbert’s hand motion about Jeff’s comment that there being a lot of smart people at Tesla 🙄😂
On 13.2 I have had what I consider critical interventions in 300 miles. In LA the pedestrian lights turn green several seconds before the traffic lights do. Somehow FSD mistook a pedestrian green light and took off forcing me to slam on the brakes and leaving me in the middle of the intersection while the light was still full-on red. If that is not considered critical then the data is misleading because there are likely many incidents that are unacceptable but not counted as critical. And I am a huge fan.
Your's definitely was not, however, I wonder if some interventions are preference. I have a MY with 13.2.2 and it started to pull out into traffic in a way which wasn't necessarily dangerous, but didn't allow for the margin of error that I usually would like.
I wonder if the 300miles per intevention is more of an average that a guarantee for every car. Hopefully it will get substantially better now that Tesla has far bigger data training capacity. I think this is only the beginning of the curve in accelerating the quality and performance of automated driving. Since as the number of interventions goes down - the training is going to be even more highly focussed to solving these cases. I’m impressed how their cars have been performing in states beyond California already. I’m looking forward to seeing how it performs in China, where there’s so much more non-car traffic.
I think the place that will really challenge FSD in China will be in the big intersections during rush hour. Qinian St and Chongwenmen W St looks like a good candidate (puzzle). The large long distance highways are exactly the same as in the US and Europe. They follow the same international standards, so these will be easy. There is in many rural places the problem of no shoulder, or dangerous shoulders. This where if you go off the roadway you're in a ditch. I saw a minibus tipped over just trying to make a U turn on a divided local road. One of the two sides was much higher than the other. A turn, you go up or down a steep bit, but a U Turn in a high center of gravity could be hazardous. (So no problem Teslas!). I've not driven in China, but I've been all over in taxis and minibuses, and large buses. An FSD rollout in China should be terrific. The Chinese are going to love it.
HOTZ was the first person Elon asked to head his self driving program. The problem was he wanted some part in the ownership of the program. He always said that privatly that Tesla was on the right track. If you have an older model S that doesn't have the board, Hotz company can supply the hardware to do some aspects of supervised self driving.
until you look at the data - that says they are YEARS away. They are at 700km without intervention it needs to be 675,000km without intervention for full FSD . Theres a reason they won't release the data. Also the x10 improvement is for their highway stack - which they have barely touched since 12.0
What is Hotz’ path to FSD success? Tesla is telling us it takes the right software approach, massive real world data, and massive compute. Where is Hotz on these? Sounds like he has the first thing, if that’s end to end neural nets. Two to go? How is he going to get those?
Please remember that while the incoming administration may make the process coming to market easier, that it is not liable for anything. So that liability is on Tesla or somebody else. Liability may not occur often but be assured that there is lot of lawyers ready to pounce.
If legacy automakers license fsd, they must also buy the inference board, in which case they have locked themselves into being 100% dependent on Tesla. All they’ll have left is styling and some of the screen UX. I find it hard to believe any company would want to do that. On the other hand, they’re fucked if they don’t. Tough decision. 🤪
Come on, Herbert. Tell us everything George Hotz said. You left out the part where he said fsd was 3-4 orders of magnitude away from human level driving. He's far from the fanboy you portrayed.
If that's true, along with platforming the guy saying "FSD is at 50k miles per critical disengagement" without proof; then I'm losing a little trust in this channel.
Buying a Tesla? Use my referral link and get up to $2,000 off.
ts.la/herbert23392
My daughter: ts.la/amanda443323
You should talk more about and to Hotz. His team and his product works well! I think he will win big. (I have Comma 3x right now and it took my car from normal adaptive cruise control to full on super cruise with a bit of tinkering.
Thank you Herbert. Really appreciate you and your team members!
Thank you JCL
Hotz actually deserves a lot more cool points for predicting the shift to end-to-end way before Tesla. Smart dude
Agreed 👍🏼
Go Jeff… a multi decade entrepreneur with a volcanic mind!!
So lucky to have you… agree with Herbert’s hand motion about Jeff’s comment that there being a lot of smart people at Tesla 🙄😂
On 13.2 I have had what I consider critical interventions in 300 miles. In LA the pedestrian lights turn green several seconds before the traffic lights do. Somehow FSD mistook a pedestrian green light and took off forcing me to slam on the brakes and leaving me in the middle of the intersection while the light was still full-on red. If that is not considered critical then the data is misleading because there are likely many incidents that are unacceptable but not counted as critical. And I am a huge fan.
Your's definitely was not, however, I wonder if some interventions are preference. I have a MY with 13.2.2 and it started to pull out into traffic in a way which wasn't necessarily dangerous, but didn't allow for the margin of error that I usually would like.
I wonder if the 300miles per intevention is more of an average that a guarantee for every car. Hopefully it will get substantially better now that Tesla has far bigger data training capacity. I think this is only the beginning of the curve in accelerating the quality and performance of automated driving. Since as the number of interventions goes down - the training is going to be even more highly focussed to solving these cases.
I’m impressed how their cars have been performing in states beyond California already.
I’m looking forward to seeing how it performs in China, where there’s so much more non-car traffic.
Critical means something that would have happened,
I think the place that will really challenge FSD in China will be in the big intersections during rush hour. Qinian St and Chongwenmen W St looks like a good candidate (puzzle). The large long distance highways are exactly the same as in the US and Europe. They follow the same international standards, so these will be easy. There is in many rural places the problem of no shoulder, or dangerous shoulders. This where if you go off the roadway you're in a ditch. I saw a minibus tipped over just trying to make a U turn on a divided local road. One of the two sides was much higher than the other. A turn, you go up or down a steep bit, but a U Turn in a high center of gravity could be hazardous. (So no problem Teslas!). I've not driven in China, but I've been all over in taxis and minibuses, and large buses. An FSD rollout in China should be terrific. The Chinese are going to love it.
Thanks!
Thank you Grace!
It’s no longer a rumor
It’s a sell the news event
Your analysis of XAI105e was spot on! I’m ready for those 10x gains, especially with Elon Musk backing it!
Jeff's desk chair looks like it came out of a cyber cab
HOTZ was the first person Elon asked to head his self driving program. The problem was he wanted some part in the ownership of the program.
He always said that privatly that Tesla was on the right track. If you have an older model S that doesn't have the board, Hotz company can supply the hardware to do some aspects of supervised self driving.
Coma AI is years away and won’t be able to catch up with Tesla. Coma AI is limited to level 2 autonomous only.
Music at the start is often too loud.
until you look at the data - that says they are YEARS away. They are at 700km without intervention it needs to be 675,000km without intervention for full FSD . Theres a reason they won't release the data. Also the x10 improvement is for their highway stack - which they have barely touched since 12.0
But is that evenly distributed across the country? Some areas are probably much better.
Herbert Is 50k miles per critical intervention anywhere near realistic?
Yeah, thinking fsd is anywhere near 50K miles between critical disengagements right now is just absolute copium/dreaming
What is Hotz’ path to FSD success? Tesla is telling us it takes the right software approach, massive real world data, and massive compute. Where is Hotz on these? Sounds like he has the first thing, if that’s end to end neural nets. Two to go? How is he going to get those?
Just bought XAI105e after watching your video! The buzz from Elon Musk’s podcast is contagious!
What happened to all the licensing deals Tesla was going to sign with other automakers?
Drove from the Orlando airport to Port St Lucie without touching the steering wheel once
Sounds great. Mind telling us how far it is?
120 miles each way
Jeff is a MDSCE@CQO and now he is just CEO, thank goodness.
The TAM is *_all_*_ transportation_ covering the approximately 510 million square km (5.1 x 108 km2) or 196,900,000 square miles of planet *Earth.*
Nobody turns a 10 minute video into 30 minutes like you.
I enjoyed the whole 30
Lol. I don't know. I see a lot of competition out there.
Please remember that while the incoming administration may make the process coming to market easier, that it is not liable for anything. So that liability is on Tesla or somebody else. Liability may not occur often but be assured that there is lot of lawyers ready to pounce.
Once FSD is safer than a human driver liability won't be an issue since it would be cheaper to insure than a human driven vehicle.
@ you gotta love people that believe in the infallible humans. You be you and I won’t be you.
If legacy automakers license fsd, they must also buy the inference board, in which case they have locked themselves into being 100% dependent on Tesla. All they’ll have left is styling and some of the screen UX. I find it hard to believe any company would want to do that. On the other hand, they’re fucked if they don’t. Tough decision. 🤪
FSD will essentially be everything important in a car in the future - without it , a car is a motor on wheels!
Come on, Herbert. Tell us everything George Hotz said. You left out the part where he said fsd was 3-4 orders of magnitude away from human level driving.
He's far from the fanboy you portrayed.
🤣
If that's true, along with platforming the guy saying "FSD is at 50k miles per critical disengagement" without proof; then I'm losing a little trust in this channel.
could tesla buy waymo or cruise just to get their data?
Lmao what data
@27:02 Just because you put _lipstick on a pig_ doesn't mean anybody's going to _kiss it._
我只能告訴這些汽車廠或是Robot製造廠，不要跟Nvidia合作，你們不會成功的。比方說，Ford汽車車上有150個軟體系統，Nvidia的系統加進來就是第151個軟體，一樣整合不了的，而且Ford得不到VW等其他汽車的data，Nvidia手上則會得到Ford的data，還有VW等其他車商的Data，最後是Nvidia得到足夠的Data訓練完成他的自駕系統，而其他每個車商還是必須向Nvidia買自駕系統。與其花錢花時間幫Nvidia做好系統之後還要花錢跟Nvidia買，甚至不知道能不能買到還要看Nvidia臉色，倒不如一開始就去找Tesla授權FSD，又方便又自駕能力完美。
To use FSD any OEM car will need an onboard computer, cameras, sensors, and the ability to update ota.
Who has that?
Who will invest?
2026 $800 price for Tesla all’s reasonable Wall Street is still asleep!
Did you see the Herbert episode from the other day where they said it's going to $30,000 in a couple of years? 😲
first
$XAI515T will be king 🦄
$XAI515T COMMUNITY WHAT UP 😎