Comma.ai and Tesla FSD: How close are we to solving self-driving? | George Hotz and Lex Fridman

Lex Fridman Podcast full episode: • George Hotz: Tiny Corp...
Please support this podcast by checking out our sponsors:
- Numerai: numer.ai/lex
- Babbel: babbel.com/lexpod and use code Lexpod to get 55% off
- NetSuite: netsuite.com/lex to get free product tour
- InsideTracker: insidetracker.com/lex to get 20% off
- AG1: drinkag1.com/lex to get 1 year of Vitamin D and 5 free travel packs
GUEST BIO:
George Hotz is a programmer, hacker, and the founder of comma-ai and tiny corp.
PODCAST INFO:
Podcast website: lexfridman.com/podcast
Apple Podcasts: apple.co/2lwqZIr
Spotify: spoti.fi/2nEwCF8
RSS: lexfridman.com/feed/podcast/
Full episodes playlist: • Lex Fridman Podcast
Clips playlist: • Lex Fridman Podcast Clips
SOCIAL:
- Twitter: / lexfridman
- LinkedIn: / lexfridman
- Facebook: / lexfridman
- Instagram: / lexfridman
- Medium: / lexfridman
- Reddit: / lexfridman
- Support on Patreon: / lexfridman

[@LexClips]

Full podcast episode: kzbin.info/www/bejne/mn_VhaWuaZd3fbM Lex Fridman podcast channel: kzbin.info Guest bio: George Hotz is a programmer, hacker, and the founder of comma-ai and tiny corp.

[@juneshasta]

To solve self-driving, let the AI ride with my pal Howard, and whatever he does, don't do that.

[@y1ink]

🤣

[@john_smithchiropractor3931]

Let Asians into Harvard just don’t use them as a driving model.

[@nat3llite]

Lmao

[@RabbitConfirmed]

Or James May!

[@krishp1104]

@@john_smithchiropractor3931 where funny?

[@knpstrr]

Hotz is a unique dude but you can tell he really respects Tesla even though it's a competitor.

[@markfernandes2467]

I'm not a fan of many ppl, but I gotta admit, I do like George. He just gets it.

[@yingle6027]

Imagine the day when FSD can drive better than the top human rally car drivers and can take you for the ride of your life around race car tracks

[@71kimg]

Can probably do that today. Problem is when eg a truck blocks a small road - and people solve that by getting out of car and talk with truck-driver or what not. Also visibility can be so bad - that no-one should actually take that left/right turn

[@mackan7086]

No thanks

[@onlythebest3311]

Driving in loops around a closed track with no other cars is trivial and already solved. The challenge for is driving around the crappy nyc roads and highways with no lane markings with bunch idiot nyc drivers all trying to crash into each other.

[@RabbitConfirmed]

just as people said we will have flying cars in 2000s, this will take forever to be reality. So don't wait for it lol.

[@hdaviator9181]

@@71kimg No, it can't. FSD still gets confused in intersections and bridges with no cars around. No way it would complete a rally course.

[@riot121212]

would this not be a step forward in the alignment problem? human feedback being integral to the predictions performance?

[@CreamyBone]

This dude is medicated

[@themilkman7367]

define "medicated"

[@FractalPrism.]

Full Self Driving is not real until the company that sells you the car, pays for 100% of insurance.

[@TheEvilmooseofdoom]

That's just silly.

[@jingyitay6179]

Talking about CGPT. Does it mean Hotz n comma needa go big n get more funds in order to make the next DGPT? Is now the time to “sell out”?

[@cristianionascu]

This gives me Musks's 2017 statement "It's a solved problem" vibes with relation to fully autonomous driving

[@olivervasquez7690]

Exactly, “it’s been solved years ago”

[@cristianionascu]

@@olivervasquez7690 sarcasm?

[@knpstrr]

Just waiting on the computer power/data volume to catch up, which they are now starting to do

[@platin2148]

From the models i have seen from pmuch any company they seemed to be all trash. And throw more hardware at it doesn’t work. I mean you are trying to brute force card driving. Without the knowledge of how it feels works or should work. The only thing that they have right now is data which might be bad.

[@knpstrr]

@@platin2148 Which models have you seen Tesla's code? CommAIs code?

[@IvanRandomDude]

The problem with self-driving cars is I would assume that there is no "good enough" solution like for many other things. It has to be perfect, "good enough" is not acceptable.

[@adityasahasranshu7503]

if self driving cars achieve the level of driving capability of humans as per today. Will u be confident enough to use self driving?

[@BuiltInBrooklyn]

@@adityasahasranshu7503Well humans cause a lot of accidents and make stupid decisions all the time, so for me personally that’s a no!

[@actualyoutubeaccount]

It depends on the autonomy level. L5, completely unsupervised driving? Yes, it can't just be "good enough" For supervised levels of driving, good enough passes because you have the error correction of a human driver. I think ppl can be largely satisfied with "good enough". Tesla's self-driving is nearly there - it just does silly, cautious, and sometimes uncautious things still. Once it drives confidently, but safely under all circumstances I think ppl will be largely satisfied to say "meh, good enough" as long as they can error correct and only have to once in a while.

[@ranxerox10]

are you perfect in driving?

[@adityasahasranshu7503]

@@BuiltInBrooklyn so, do you drive a car now? Or just stay at home because humans cause accidents?

[@John-Edward]

I’ve been using FSD for about 3 months now, and while it is impressive technology, but I would not use it for everything at this point!! There are many moments where it makes stupid lane change things that while may not kill anyone, are very annoying!!! I will generally use FSD more on highways vs City streets… I think Autonomous Driving is 5-10 years away from being truly useful in a general sense… usa sayin’.

[@94alhf]

How about now? This comment was written 9 months ago. With the latest updates can you share your thoughts on it now

[@olemew]

"We solved it years ago" 1 minute later... "We're ten years away"

[@princeofexcess]

Funny guy to have a conversation with. Kinda feel like it wasnt very informative though. I think Comma has a place in FSD domain. It will be behind FSD beta for a long time but they offer a cheaper alternative.

[@Gruby7C1h]

It's mid 2023, just look at the amount of full self driving cars around you, especially in environments where there's snow or other hurdles... I've heard about self driving being "almost there" back in 2017.

[@bezimeni2000]

Eh. Not really

[@kianp]

he lost me when he said we solved it years ago...

[@onestepclosertoheaven7347]

As a Tesla employee, you cannot send out full self driving instantly. In order to accelerate sustainable energy , we have to let our competitors feel like they are catching up. Otherwise we put alot of peoples jobs at risk if we instantly release technology that is too far ahead. Eventually self driving will get better and better and enter an age of no steering wheels with self driving. Just have patience 😉

[@postedactive]

I am enjoy thé comments on these Géo Hotz videos 😢😂😂

[@CptSquirrely]

It would be nice to understand what George is talking about. I can’t follow all the acronyms.

[@pbasswil]

I'll point out: Self Driving isn't one problem to be "solved." It's a massive _bundle_ of overlapping challenges. FSD developers can feel in control of the vast majority of common driving scenarios, and then stumble on a situation where their software doesn't work right at all, and feel like they're back to square one. You can never 100% 'solve' self-driving, cuz unique or unforeseen problems can arise at _any time._ And even if Tesla eventually gets governments to certify FSD, eventually some tragedies will happen; and each time one does, the technology will be savagely disparaged in _all_ media across the planet. :^/

[@dannyisrael]

He’s so unbelievably arrogant and sounds so dumb. I can’t listen to him.

[@24hrdiner]

FSD will be 10x safer than a human driver .even though fatalities may happen the stats will show for every 1 death the equivalent is 10 deaths due to humans driving.

[@pbasswil]

@@24hrdiner I think you're probably right. I was pointing out two things: a) That this is an analog challenge: There is no On/Off solution to FSD, rather it consists of a l-o-o-o-n-g series of improvements, narrowing down to the handling of increasingly rare situations; and b) Regardless of how statistically safe FSD becomes, each individual fatality from it will incite a frenzy of media scrutiny, and harsh criticism from detractors. And imho that scrutiny will get even more brutal, once government agencies have dared to approve FSD (that is, when it's out of Beta and into the realm of mainstream products). Because folks _love_ to disagree with governments.

[@2204JCM]

Yeah it will never be perfect just like humans arent. But it actually learns as it goes mistakes or not and keeps getting better. So sooner or later it will more or less perfect even though it technically isn’t. Bottom line its going to be more reliable than humans and that’s what the insurance business cares about…

[@pbasswil]

@@2204JCM That's a good point: maybe good numbers from the insurance industry is ultimately the most important kind of approval. I'm glad you brought them up.

[@nkau7788]

so this is the future of the car enthusiast?

[@kd969]

My Hyundai with Comma 2 drives so much better than Tesla standard autopilot. Amazing tech

[@BatteryOnBoard]

Find yourself on a ten mile stretch of newly paved, winding mountain road, and you won't ask "Why do you feel like driving?"

[@Duskinterteling]

1:15 “it was an auto encoder and rjena andndja ajajd right?

[@bokoler9107]

Lex you would nail it right from the start, if you would present a wise entrepreneur who understood the principles of Napoleon Hills interviews. I think it is time to move into this educational episode now.

[@AtillatheFun]

Napoleon Hill was a fraud who was exposed multiple times. He never met the people he claimed to have met, and he peddled these fake claims all the way to the bank.

[@snip3d]

Openpilot is better than AP in some areas and vica versa..........just put them together already please!!

[@GarthShoemaker]

Is driving the worst/riskiest domain to chase for the first big AI "win"? I suppose handing over control of the nuclear stockpile might beat it out, but driving is not too far behind.

[@TheVigilante2000]

So FSD is a solved problem, that nobody has deployed in a general way? And even the limited uses have major problems. Not what I would called 'solved'.

[@michaelnurse9089]

Semantics. He clarified and says it has bugs. The evidence that it is solved is that there are Teslas driving people to work daily with zero disengagements most of the time. Sometimes a bug emerges and there is a disengagement. Were the first cars or computers bug free?

[@dirkbester9050]

@@michaelnurse9089 Or you can just call him a liar. I drive a Tesla with FSD and it is not solved. Unlike with Autopilot I am optimistic about FSD but pessimistic about hardware 3. Unlike Elon I am not "this year" optimistic. There are an incredibly large number of beta 11 versions in our future. This is not semantics. There is no actual proof that either current hardware or software is sufficient. I doubt we will find that out till after DOJO is pulling its own weight. I will never let his comma based software drive me, at all. He is the kind of moron that implodes in his own corner cutting deep sea "submersible".

[@mackan7086]

​@@michaelnurse9089key here is "most of the time"

[@TheVigilante2000]

@@michaelnurse9089 Getting something to spit out an answer (a human policy for driving) is not 'problem solved'. If it worked all the time like a human, then maybe it is solved, but I never 'disengage' when driving EVER! I don't stop in the middle of the road looking at a water puddle unable to tell the car what to do. I don't see a bike on the back of a car and get confused that I might hit a biker crossing traffic. I can explain why I swerved away from a child and hit 3 adults. I can talk to a cop about the closed road and take direction to drive around a crash scene. Tesla and all others are not even close to 'problem solved' [shiteating grin]

@@michaelnurse9089 That like saying someone is almost pregnant. A fully automated system cannot have "bugs". Because even bugs in every 1.000.000th hours of driving might mean a lethal accident.

[@matteoperron5436]

I love george haha

[@festeradams3972]

George has at least two things over Musk...he doesn't get into Politics, he doesn't get into the "Social Media" thing. Musk has only proved that "smart" does not extend into all areas...

[@jrcanuck1]

George is always so nervous, talking fast in a very choppy manner, that it feels like he is on speed & he comes across as a smartass, who is always hiding something. He rarely elaborates in his answers. He gives short choppy answers. In my opinion, he's very hard to listen to for more than 10 minutes. As far as I have seen, Self driving cars are not even close to being capable of handling snowy conditions.

[@DeckerCreek]

Yea self driving is bs in anything other than limited good weather areas.

[@ranxerox10]

@@DeckerCreek how can you know?

[@jingyitay6179]

I beg to differ. I could watch his twitch videos all day long. He’s pretty entertaining. That’s good science/tech education to me One of the best interviewees on lex imo

[@jrcanuck1]

@@ranxerox10 from every report and articles that I have read. Show me otherwise & I will be pleased.

[@jrcanuck1]

@@jingyitay6179 , to each there own... It seems like a number of people agree with me.

[@reh.4919]

Do either of these upload any of the video. I'm not fond of a camera staring at me all the time.

[@j.d.4697]

I like how Tesla keeps talking about having that tech it already sold with all those cars, but still doesn't give it to the people who already paid for it.

[@TheEvilmooseofdoom]

It's not release ready. It's still in testing.

[@timothyharrison]

This might be the first time I’ve heard Lex sound like this. Perhaps because it is his field. Is it a competitive tone maybe? Doesn’t matter. Unless you pseudo worship Lex like, oh I don’t know… me!

[@bravenkind7843]

Would be sick if they joined forces with Tesla for making devices that can work with ALL cars.

[@emmanuelmata]

His company will never fail, but he shouldn't be the face of it. comma ai should be worth billions right now

[@sanahaskuranage8071]

Comma ai is only good for highways. How’s it going to make unprotected lefts. Not worth billions maybe a few million. Unless they work with an OEM and introduce more camera angles. They’ll have to train again. They are more than 2 years behind tesla.

[@billthecat7536]

Good luck, George. When Tesla brings up full production of DOJO (soon), by the end of 2024, it will have 100 exaflops at it's disposal. But keep up the good fight!

[@petersmangalisongoma2013]

Yeah, George seems to underestimate the hardware requirements of full autonomy

[@_____case]

Google currently has the crown with 9 exaflops capable clusters.

[@oscarmike4596]

He literally said - whoever can scale😂

[@baynetim]

Entered

[@johnkost2514]

Hubris..

[@cifey]

There needs to be a universal swarm api where all the cars are broadcasting to eachother frame by frame about their weighted intentions. Once that happens we won't feel like driving because everything will just be zooming around the dinosaur on the road.

[@Kitten_Stomper]

Yeah, but that won't be in place until you force people to install that broadcast system. That can't be done until the cost comes down.

[@slyktech1860]

My flops are sitting here watching this video and whether or not the usage of these flops are being useful is to the eye of the beholder.

[@TeslaElonSpaceXFan]

👍

[@daxramdac7194]

Im looking for this "arrogance" that everyone in the comments is talking about. All I see is a guy who enjoys what he does having a conversation.

[@Yazolight]

That guy’s accent is so hard to comprehend

[@eyesuckle]

There's something incredibly smug and annoying about this guy's manner--the guest, I mean. Not Lex Fridman. Come to think of it, Lex's humility and openness to learning by contrast is part of what makes his guest's manner so off-putting.

[@moonloopLeo]

Exactly, he comes off as arrogant and smug in every video he's in. The worst of silicon valley.

[@dirkbester9050]

His Comma ai, public on youtube, mission statement is "to screw over other self driving companies and prevent them from making money". He is mentally deficient. You should never go into business with him. Remember how Elon vetted him and then told him to f off? Just follow Elon's good example. He managed to detect the rotten fish smell. I feel like you are smelling something too?

[@TylerHallHiveTech]

Dude drives me nuts. Watched some of his “live coding sessions” during the Twitter mess when he then “joined to improve search” and it’s very obvious that his execution is parallel to what you described about his personality. I don’t see much value with it. Sporadic, interjecting philosophical insights when just technical problems come up, inability to see the whole picture and constant framing of small problems as so small they’re solved. Except the small problems are unsolved and stack up to a mess. And then he “doesn’t think he can help” Twitter search and leaves quietly while benefiting of being loud during the hype of elon buying.

[@2drealms196]

@@moonloopLeo I wonder if his air of arrogance could be a cover/coping mechanism for his insecurities he might be having behind the scenes with CommaAI.

[@daxramdac7194]

Well when you're as smart as he is, first guy to jailbreak the iPhone and the PS3, and then decides to tackle selfdriving cars, that's probably gonna come with some, let's say unique personality traits.

[@billbradleymusic]

Simply saying "... just stupid bugs" doesn't do it for folks with a brain.

[@GuyFromGeorgia]

DriveGPT. That's good! Surprised Tesla hasn't grabbed that moniker, especially since Elon's trying to make everyone aware of Tesla's 'AI moment'

[@keyserzoom9154]

It is unlikely that FSD will ever be completely perfect when operating in an imperfect world, just as there is no perfect driver. I suspect adoption will be driven by insurance and accident statistics, with some interesting liability clauses..

[@ramraj08]

“How close are we to curing cancer?” Owner of drug company: we already solved it! Big surprise.

[@tomw485]

Except cancer is a catch all term for thousands of ailments who’s commonality is uncontrollable cell growth often requiring unique therapies in which not all patients have the same response to. We both have and we have not cured cancer. Have we cured stage 1 cervical cancer. Sure. Have we cured stage 4 pancreatic cancer. Nearly always terminal. Don’t think in terms of curing cancer like curing an infectious disease like Smallpox or even HIV.

[@EpicHardware]

There are some minor bugs, will solve it i a year or so. Everyone knows this stuff -Hotz

[@ProfCantonius]

@@tomw485 Same with Driving as there's city driving, highway driving, driving in the rain and snow vs sun. Driving in Mumbai. Highway driving with sun and no edge cases probably solved. Everything else probably not. Driving in Mumbai, not even close.

[@dfjkl1]

I’d argue that “Tesla isn’t doing anything wrong.” I’ve stopped attempting to use any of those features of Tesla because they just keep making worse decisions over time. They can’t properly deal with basic speed sign inputs or provide override mechanisms for their mistakes. All this investment of money and resources would be better spent on mass public transit networks.

[@oscarmike4596]

They are Tesla competition- that is absurd

[@briantep458]

FSD is basically a 16 year old teenager practicing for their driving test

[@thestonksmarket1897]

Totally false. Tesla's FSD Beta is leaps and bounds safer than a 16 year old driver... it's at the point now where it's safer than the average DRIVER regardless of age. But that's not saying a ton since there are a shit ton of bad drivers out there dragging down the average.

[@briantep458]

@@thestonksmarket1897 I have not used FSD, I have only seen videos. I will tell you that I can only use autopilot for 20 mins at the most before intervening. The software isn’t broken, the model is. Its very difficult to teach software to anticipate bad driving. A teenager with very limited driving experience can instinctually detect when a person will cut them off just based on profiling. There are other nuances too. For example the car will always drive in the center of the lane when in reality its safer to drive towards the left edge to give room to parked cars. I will feel safer when 99% of cars are equip with self driving technology, or if people just stop driving like jerks all the time but we are not there yet.

[@thestonksmarket1897]

@@briantep458 I’d recommend trying FSD Beta (there are affiliate links to get 3 months free in various places). It’s WAY better than basic and advanced autopilot as it can handle many many more driving scenarios than the other two. Teslas have 360 degree vision at ALL times. Humans do not. When we look one direction we are NOT looking in any other direction. Therefore a Tesla with FSB Beta in its current form is leagues better than the average 16 year old driver. And I’d wager it’s better than the average driver. Once more data is compiled, processed via Dojo supercomputer and Fed back into Tesla’s FSD feedback loop you’ll see the software become better than MOST drivers, not just the average.

[@erikm9768]

Why does geohotz believe Tesla is always 1 year ahead of us. Its not like they are anywhere near Tesla were 1 year ago, thats just laughable. Does he actually believe this? Tesla has like a fleet om millions of cars. Comma has a few hundred at best.

[@TypicalBlox]

Comma actually has over 10,000+ units on the road, it's a great product

[@0013619511]

Wait it might take 1 year or it might take 10 years to be 100 % Self driving ???????

[@tomjensen618]

These guys are all bad drivers. They think they are ahead but, they don't know what a good driver can do. Lies will not sell defunkt systems. Don't rush to buy their hype.

[@ElektronicEscalation]

Its not about how an excellent driver would do, it’s about how the AVERAGE driver would do. And to be honest the level isn’t that high in that regard. It’ll only take a small amount of time from being more safe than an average driver vs excellent driver though, once we’re at that stage

[@EzraSchroeder]

3:18 a person will always feel like driving at some point. humans are autonomous. we crave autonomy. that's how babies come into the world.

[@sanahaskuranage8071]

He seems like he knows more than anyone else. But I don’t think he does.

[@rolla2035]

It’s not solved. Typical tech bro Silicon Valley arrogance. It’s not solved until you have real AI. It’s still a bunch of if thens , neural net or not. It’ll still fail in weird situations and it can’t decide anything as well as a human.

[@DigSamurai]

FSD doesn't have to be perfect it just has to be much better than a human because humans are terrible at making decisions under pressure in complex situations.

[@MrAngryCucaracha]

How can you say they are terrible when there is currently no better system?

[@DigSamurai]

@@MrAngryCucaracha because the evidence is clear. 40,000 people die a year in car crashes in the US. Do you know how many people die in plane crashes? Some years it's below 100. That's because the level of training required to be a pilot is extraordinarily high compared to training to drive. The potential of FSD is all cars will talk to each other via 5G virtually eliminating crashes. How many years will that take? I don't know but there's no question that's going to happen.

[@michaelnurse9089]

It is already 10x safer than humans. It is just about 10x more embarrassing.

[@10162131]

It’s not about being x times better than a human, it has to be perfect before someone will actually seat on the back seat and let it drive around. Will you send your kid to school if it’s even 1000 times better than a human? Don’t answer it just to win an argument. Truly think about it and honestly say you would while you go to work or do anything else.

[@kennyg1358]

​@@10162131when I'm convinced it's 10% better than a very good human driver I'll let it drive me, my kids, my dog, my rare Faberge egg collection...

[@soraaoixxthebluesky]

I always think of it as feed fwd nn Snake AI game. It learns that he was wrong if he's dead (when human override the car). He will to try not to die the next time from the experiences of him hitting his own body or the walls (disengagement). The reward is like eating the apple which then he knows he makes a right decision. Like no human engagement for every 5 miles, or when he can take all the type 1,2,3,4,5,6 turn without human disenggament at "x" amount of speed. Turn type can be classified using rally racing code at which number refers to the severity of a corner ahead. The spectrum goes from 1, referring to a slower to 6, which refers to a fastest. Rewards can also be given if the system can self-brake when needed without human enggament. Every single action taken with no human engagement can be consider as rewards.

[@grahambrown42]

Comma compute power 40 Petaflops, Tesla heading for 100 Exaflops (100k Petaflops), so 2,500x Comma 😉

[@sacsauce4200]

It seems like a human not disengaging is a lower bar than what a human would do...I can be marginally dissatisfied and still not disengage, right?

[@sadlizzard1]

I smell pot in the air..

[@JotbotAI-Octavio]

Lit

[@Ty_Neadik]

George Grow Up! Say "My research went nowhere yet" and move to the next edge on the graph, do not be like 3l0n(manchild). RTOS asics cannot yet learn Newtonian physics of cars like a person behind the wheel, I Know when to push the brake unlike an AI's I/O ports.

[@dfend451]

Hmmm gambling with your life in the hands of software. I'm sorry David, I've made the wrong decision.

[@sanz_96]

What is this dude talking about.

[@Jaycren86]

This is an odd duck. Does he seem snarky to anyone?

[@sandeeps2]

Textbook imposter

[@briankwest]

The arrogance of the 'we've solved it' is weird to me, its way more than they are letting on here.

[@henrikandersson3286]

Who decided self-driving is a goal?

[@robby12320]

Would be a goal for me. I hate driving.

[@DuckDuckGeeseMusic]

Human drivers did.

[@paulhamrick3943]

@@robby12320I just paid $20 plus tip for an Uber last night, would have been great to only have to pay $10

[@Kitten_Stomper]

@@paulhamrick3943 This, and the cost of truck drivers bringing food to our grocery stores.

[@Sergio-yj6pc]

Wtf?

[@jonesingforfood6380]

All these comments about George being smug don’t understand what he knows. Smart people like George come across this way cause it is so obvious to him. FSD is a solved problem. It just needs a large data set and the compute power to train the AI.

[@arky1923]

Isn't the limitation of any powerful AI that any tom , dick or harry fantasizes about ..

[@NewYork7914]

I bet that George has ADHD.

[@jingyitay6179]

😂

[@is1nameleft]

He is scary. In the road to hell kind of way. But then again all humans are

[@tylerknight99]

Lex Fridman get more sleep before hosting challenge [IMPOSSIBLE]

[@erikm9768]

Every self driving question is answered with what comma does, geohotz have problems realizing that hes like 5 years behind Tesla, also they are like 5 person compared to 127,000

[@morticus7650]

Can't watch his interview, he sounds disingenuous with the way he answers.

[@AgathaWhispers]

To be fair - he can’t say too much - when he so happily said they are 1-2 years behind Tesla - I instantly thought - maybe not

[@Charieldeleon]

He’s the type of guy who kisses himself in the mirror lol 😂 he’s so disingenuous!

[@Jaycren86]

Agree.

[@dickheadrecs]

he’s the kind of guy you want in the race, but not to win

[@jonesg9798]

He knew End 2 End was the answer 3 years before Tesla

[@EpicHardware]

i feel he's lying to safe the company when it says minor bugs

[@24hrdiner]

5:15 tesla will always be 1 to 2 yrs ahead of them

[@fredrickkun]

Never Watch Jetsons Technology Is just as bad Total recall was a virtual reality in a reality no one could get it right It’s just like walking while chewing Bubalicious chewing gum baseball pack inevitable of falling we can’t even fly planes ✈️ without crashing in open sky 🌌 Satellites 🛰 crash all the time We walk and bump into each other 💭🤔 How smart are we really We make sex dolls And the sex dolls run 🏃 🏃‍♀️ away 🎙 🎤 💥 💥

[@chrisbarry9345]

It feels like the decision to move away from lidar was a big mistake for Tesla

[@TubaTones]

Incorrect.

[@TubaTones]

It might feel like that, sure, but LiDAR is very expensive and will only worked on premapped areas of cities, it won’t work anywhere else, and you have to constantly remap those areas. It’s a limited technology for general automotive use, for factories, sure, but radar does just fine there as well. Pure vision based AI autonomous driving, like where Tesla is heading, will in the end be far superior due to its application literally anywhere on the planet. Tesla’s vision based FSD has gotten so good, they’ve removed the radar sensors because they are creating too much noise (interference) and getting in the way.

[@pppparanoidddd]

Humans have no lidar and humans can drive, cars then also do not need lidar

[@TubaTones]

@@pppparanoidddd bingo. Vision based AI all the way.

[@hi-gf5yl]

⁠​⁠@@TubaTonesthey actually added mmwave radar in HW 4.

[@sporefergieboy10]

I wonder why the arabs don’t like these people