Ex-Uber CEO: Tesla's FSD 10x Better! | Jeff Lutz

[@BrighterwithHerbert]

Buying a Tesla? Use my referral link and get up to $2,000 off. ts.la/herbert23392 My daughter: ts.la/amanda443323

[@InspireLifeGoals]

You don't need clickbait titles. People come here because you and your guest offer excellent content. Appreciate you.

[@metatron3942]

He didn't slam Tesla cyber cap and FSD I think that's a clickbait title and that should be beyond your channel

[@LazerEyez]

I agree. Don’t go down that click air route pls

[@toddborstad504]

Yeah Herbert, clickbait is a huge turn off and insults my intelligence. Your channel is better than that!

[@415BurberryKid]

Yup facts! I watched the interview and thought the same

[@janh-r8h]

❤

[@abcqer555]

Agreed. I want solid Tesla content and analysis. This seriously turns me off and makes me take you and the channel far less seriously. Plus it is annoying and inaccurate. If the title is a lie... What else is.

[@timnewell4015]

I watched the podcast. Your headline is pure clickbait. But I suppose it worked, because here I am interacting. That being said, I usually enjoy your videos.

[@mkenoyer]

Same for me. I knew it was Clickbait and came here anyway. It’s not necessary and turns off your true customers

[@erktrek]

The issue with ICE cars as AVs is as others have said - greater maintenance, less predictable performance, less safety - greater fire risk, greater issue of something going wrong in transit.

[@talajon5469]

Why using “Slam” if he is so positive about Tesla’s FSD and Cybercab?

[@abcqer555]

It's clickbait nonsense.

[@sat7755]

Travis is probably willingly forgetting Mega Pack storage which will at least double available electricity.

[@steverice2318]

I agree with what Jeff is saying about "Who has the manufacturing ability and data" to be able to do real world autonomous? The video of Avatar car in China with Huawei autonomous driving assist was very impressive. I'd like to know how Huawei had done what they are doing without the training data? It's not using geo fencing. It is using end to end neural networks. How are they doing it. Is what they are doing going to become the way other companies can commoditize level 5 autonomous driving? Does that destroy Tesla's moat sooner than we think? Something to think about. I am a big Tesla investor and believer. But, that Out of Spec video made me think. Thanks, guys. Great content as always!

[@wildcatmahone-md6me]

I would need at least a few sources to believe anything the CCP puts out.

[@talajon5469]

If GM or Ford had Tesla’s FSD and Cybercab their stock would hit 5X at this level. Tesla/ Elon haters is the prominent reason Tesla stock is not 5X at this level. IMO

[@paulhill182]

Mr Uber's problem is that he lives in CA, not in the civilized part of the country. Most states have sufficient power reserves and the ability to add new capacity, as needed, in a reasonable time frame. How many new power plants has CA built in the last 10 years? How much long range planning for increased power requirements has CA done. The CA solution appears to be, why should we worry about increased power needs, we can fixed it easily since their people can just move to another state that does have the power.

[@nickmcconnell1291]

First they laugh, then they get angry, then they get worried and say something stupid, then they die.

Just dont see how ICE autonomy would be as cost effective, still the 40% more maintenance. They already have decent driver assist software. I could just see improving that to the pointnof increased safety/collision avoidance, but not level 5 autonomy.

[@DavidSaintloth]

Travis may not have worded what he was trying to get across properly but I can tell you as an IT professional and machine learning researcher that he's 100% right about cheap AI leading to cheap autonomy. It's literally what's been happening as the hardware technology is developed, Tesla and other companies that have been working with machine learning models have been able to take bigger models and run them on hardware which initially was very expensive but over time becomes more commodity. But the more important issue is what happens when you find ways to optimize the training of the parameters in a given model. Parameter space is multidimensional and the ability to encode complex understandings of the world into a network is determined by how one can identify the directions in parameter space that are associated with highly relevant predictions once a trained model is deployed into the world... What that means is here you are, A bag of meat and water that at some point learned how to drive... Did it take you 100,000 miles of driving before you could drive adequately on pretty much any road?? No. Are you powered by a nuclear power system or gigawatts of energy? Were you powered by such energy while you underwent the learning process for driving?? No. This tells you immediately, That we aren't even close to the levels of compression that we can get of these models in terms of the amount of intelligence that they can pack into a tight set of parameters and be able to make predictions with very low energy and after significantly low levels of training... Again, most people can confidently learn to drive on any roadway after only training for several weeks in practice. So a few hundred miles of training enables them to do inferencing for a lifetime... That's significantly more efficient than any of these neural models that we are currently deploying.... So for those who think that the innovations did deepseek revealed are the end-all to be all of this compression guess again?. We saw this before with algorithms across the different types of data types and the history of computing. We've watched new, more efficient algorithms be developed for doing things like compressing video data, compressing audio data, compressing regular text.... In that parameter space finding the feature similarities involves learning to efficiently utilize as many dimensions as possible. This way, the multidimensional vector that takes a model from a given input to a precise output prediction becomes extremely efficient both in temporal cost and in energy utilization. Anyway, hopefully Tesla is early enough along the curve of optimizing the vision model that it is utilizing nfsd that it can reap the benefit of the rewards of efficiency that it has achieved for the next decade or so before the competition can do the same thing... My guess is more realistically. It'll be about 5 years of dominance.... After all, it's not like Tesla is impervious to also potentially being hacked in some way and have the weights and biases of the various modules of its end-to-end architecture borrowed by outside actors. It's important now is speed of execution to get to market first and then dominate as much as possible. Because smaller and smarter models running on much lower power hardware are coming... It's not a matter of possibility, it's just a matter of brief time.

[@valentineni208]

Yeah I also wanted to point out to Huawei adas 3.2 system. The advantage against waymo in terms of geofencing is only applicable to US companies, Huawei adas is not geofenced and it was also mentioned that the lidar sensors not that expensive around 200$ per sensor and avatar 11 has 3 of them. Herbert could you please comment on that. Feels like Tesla will take US market in terms of robotaxi but in China it will be competitive

[@khuo0219]

How do you know that Huawei's system is not geofenced?

[@jonbowes5999]

As it stands Tesla is handicapped in China because China won't allow Tesla to export the training data back to the US and the USA won't let Tesla export the training compute capacity (and possibly the AI intellectual property) to China. Trump has suggested he is ok with Chinese companies coming to the US and setting up manufacturing there the question is will that allow the Chinese companies access to US located AI compute and IP if they can find a vendor in the US.

[@westcoast8562]

LOL I WONDER WHAT THE TOTAL ghg'S PRODUCED BY xai AND TESLA SUPER COMPUTERS IS BY THE END OF THE YEAR VS. THE GHG SAVED BY RUNNING TESLA EV'S WOULD BE

[@westcoast8562]

HOW DID I FAIL TO SEE, ITS NOT ABOUT ABOUT EV ITS ENTIRELY THE SEARCH FOR SINGULARITY

[@GrantRobinson-r1r]

Question. What does a freight consignment and delivery look like with autonomy. The first and last metre in particular. How will the freight be handed over to the receiver or even from the sender to the freight company? 10:36

[@tompava3923]

I think Travis‘s napkin has huge holes in it.🙄

[@stevenhill3136]

Clickbait title!👎

[@tedg1609]

Rooftop PV + batteries are the likely solution for a failing LA grid. It’s kinda sunny in LA.

[@milesinaz6027]

Why would I tune in to listen to this?

[@christinehalk5550]

Why not put a Tesla FSD in next year's ICE cars? That idea could save Chevy Ford Toyota etc while they retrofit. Many big computer companies use other companies' hardware inside. Legacy could do the same. I posted this same question yesterday. It would also give the grid time. Win win win.

[@GalacticMarine2012]

How are you going to make the FSD computer turn the steering wheel? You would have to change and modify so many mechanical parts of an ICE car, you might as well just build an EV that's made for that type of thing.

[@wildcatmahone-md6me]

They would have to design a complete vehicle including subsystems from the ground up. $$$$$$

[@ricwilmot1654]

Why not charge people 20$ if they leave trash. Also get optimi bots to clean the cyber cab inside and out.

[@janh-r8h]

Clickbait title again.👎

[@bru512]

ICE AVs? Technically it's possible, but the low, low cost of BEV operation will allow the capture the greatest profits

[@rangerthedog8981]

CLICKBAIT

[@GoroDan]

There is an example of Norway their vehicle fleet is about 25% evs, how has it affected Norways energy consumption. Does anybody know?.

[@andrewvercillo7584]

Of course he trashed them! He is jealous. Gas cars will never be software based.

[@stevem8602]

Herbert lost any objectivity a long time ago

[@westcoast8562]

HOW DID I FAIL TO SEE, ITS NOT ABOUT ABOUT EV ITS ENTIRELY THE SEARCH FOR SINGULARITY

[@michaeltelaak1452]

One area will not go 100% electric all at once.

[@tobias7252]

They are talking like they haven t seen the huawei avatr 11 video.

[@mikebailey2970]

why not deploy optimus to drive any ICE vehicle ?

[@highlanderapparel]

Let’s not pick on the Highlander for holding onto his Dodge minivans for 15 years at a pop. I’ve already made the public statement. My next car will be a used model three respectfully the Highlander crowd sponsored the Gruber corporation.😊

[@drawzeywazig9096]

How about Huawei 3.2 combined with manufacturing? Isn't that the combo we're discussing?

[@talajon5469]

ROO taxi market is huge. If Tesla pulls it off this year Tesla will dominate this market in US and Europe . Tesla will be ready for launching their Robotaxi fleets this year for sure. IMO.

[@goukux5908]

Huawei has an autonomous system that takes the same basic approach as Tesla (general solution rather than map based) but they use a lot of sensors; LIDAR, Radar, Sonar, and 11 cameras so the sales price is between 40K and 60K and I don't know what subsidies that includes. I saw a short clip of the car that looked like it performed well on a relatively difficult city street setting, but it was just one clip, so I can't say what the actual performance is. I think they can catch up at some point and even if its more expensive to produce I think it will be a competitor in china someday.

[@juliahello6673]

The clip I saw from out of spec didn’t seem like a challenging road. Challenging for humans, yes. Aggressive driving, yes, but that’s the style in China. The situations that they marveled at were ones my hw3 with older software can accomplish (dealing with cars elbowing in from the side, etc.). It was gratifying to see a system that good, since there is nothing that good here outside of FSD, but road and terrain recognition is the challenge for EVs and I didn’t see that in the video. Their system might be better, worse, or the same as FSD but we don’t know from that video.

[@thomasr6848]

Would also like to get the perspective of the Tesla community on this system. It seems to be very performant. And why should Huawei not license outside of China…

[@PMI551]

I would agree. The competition isn’t as far behind in terms FSD as a lot of people think. In China, Tesla can’t train their data outside of China, and the us won’t allow Tesla to train their models inside of China. I can imagine that to be a significant handicap, but necessity is the mother of invention, just look at deepseek.

[@khuo0219]

I think we have to be realistic that a country like China will never allow a foreign company to dominate autonomous driving technology, considering the winner take all nature of the market. Tesla will not be allowed to fully train their model with Chinese data until local players are equally competitive. Since Baidu vs. Google. Tesla is better off focusing on other markets in countries that do not wish to compete in autonomous driving.

[@BillB33525]

@@thomasr6848 Exactly, imagine US auto mfg could buy a box here for reasonable dollars that drives the car that good.

[@toddpossum]

Thanks!

[@michaeltelaak1452]

It feels like UBER and WAYMO people desperately trying to justify their companies existing.

[@wildcatmahone-md6me]

First comes denial, then comes fear.

[@ianteetzel]

The metric that matters on high mileage robotaxi's is COGS/mile. CyberCab at scale production looks like it will be 1/2 of other ev's including other Tesla's and 1/4 that of a gas cars 🫳🎤

[@DiO-fy5ex]

Could somebody ask Leon to get Tesla insurance in FL

[@okay4403]

Click bait. Way less likely to open videos or subscribe. It’s fluffy channel growth, not enduring. Look at Tesla Daily. Banging thumbnails but zero clickbait.

[@patrickdilauro6868]

Synthetic data will allow most manufacturers to catch Tesla

[@wildcatmahone-md6me]

Wrong.

[@curtisyoung7107]

Hmm specially trained augmented humanoid robot drivers could opperate existing vehicles better than most human drivers, but it would not be a profitable use of this kind resource when compared against Tesla with unsupervised FSD 😊

[@davab]

Lets make robohorse

[@lifer9883]

Seriously? clickbait?

[@MrJSiQ]

Being back tesla daily with rob mauer!

[@santiagovilla6219]

cheap AI? not there yet if i am not mistaken. Investment in AI yes

[@LJ-jq8og]

JEFF 💪❤

[@MrJSiQ]

Clickbait thumbnails are trash; dont be that channel

[@Jack-jk1zl]

Ridiculous episode title

[@stefanconstantinescu1576]

Have you seen Huawei AVATR FSD = ADAS ? And is is FOR FREE

[@RichardDavenport-h8s]

He scared

[@diogenez471]

scared of what ? he doesnt work anymore for uber.

[@wildcatmahone-md6me]

​@@diogenez471probably still owns stock

[@Lindathomas-h6z]

I'm 51yrs old. $40,000 weekly and I'm retired, this video have inspired me greatly in many ways that I remember my past of how I struggled with many things in life to be where I am today!!!! 👍

[@MilomirVujčić]

Huge! been trying to trade on my own for a while now but it isn't going well. Few weeks ago I lost about $10,000 in a particular trade. Can you at least advise me on what to do?🎉

[@U2B_Viewer]

This bot couldn't even be smart enough to pick a profile picture that represents a 51-year-old 😂😂😂

[@appl314]

How does Huawei's auto driving car/software compare?

[@arielbecke]

🦉

[@ChuckCobain1]

Waymo is way behind Tesla. It's very small scale, 3 times more expensive than tesla and and the AI and technology is behind and inferior to Tesla.