Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Andrej Karpathy leaves Tesla

twitter.com

Post reply on HN
Print view View original

381–390 of 441 posts

Re: Andrej Karpathy leaves Tesla

#381

Post by coenhyde »

giantrobot wrote: post #351
coenhyde wrote: post #317

Earlier quoted context omitted.

> That doesn't make any sense. "Legacy auto" knows how to make cars. To their final assembly factories there's not a huge difference between a BEV, PHEV, or an ICE drive train. So long as they feed in components they get cars out. They definitely know how to get components made to feed into their factories. That's the problem. When they do it that way, they don't make a profit on BEV's. "Legacy auto" got good at buil…

> That margin will continue to grow and then turn into a weapon when Tesla needs to be competitive. This does not follow at all. Tesla sells their barely up market cars at luxury car prices and makes a significant amount of money off selling emission credits and raking in other government subsidies. As "legacy auto" manufacturers ramp up their EV production they won't need to buy emission credits from Tesla. They'll…

> This does not follow at all. Tesla sells their barely up market cars at luxury car prices and makes a significant amount of money off selling emission credits and raking in other government subsidies.

Emission credits were 3.6%[1] of revenue last quarter (Q1 2022). I don't know what "subsidies" you are referring too but you may be thinking of the EV tax credit, which was exhausted by Tesla in late 2019[2]. So in that regard other EV automakers have an advantage over Tesla.

> I also don't see Tesla being able to compete on luxury as they have significant problems with fit and finish. Now that traditional luxury brands have solid EV offerings it seems unlikely Tesla will be able to keep their margins and increase their sales volume.

Tesla has certainly had fit and finish issues but they are solving this. Their new Model Y produced in TX uses their "giga castings" which removes the opportunity to screw up alignment by removing ~150 welds for the body. Look it up if you want to know more.

I think error in your thinking is that Tesla will remain static while others catchup. That's not going to happen. Tesla is innovating (increasing margin) and scaling at a faster rate than "legacy auto".

[1] Q1 2022 Financials https://tesla-cdn.thron.com/static/IOSHZZ_TSLA_Q1_2022_Updat... [2] https://fueleconomy.gov/feg/taxevb.shtml

Re: Andrej Karpathy leaves Tesla

#382

Post by dlkf »

8jef wrote: post #143
ok_dad wrote: post #108

Earlier quoted context omitted.

It’s almost as if we should invest in public transportation and try to reduce cars rather than make a train car for each individual!

But for people, in the same way as for parcels, power, cable, fiber-optic, there's the last mile problem. Europe has almost solved it with high speed trains and efficient city transit. Then, it solves nothing for all those folks living in rural areas and having to drive f-150s.

I’m going to guess you don’t live in Europe? Plenty of cars here.

Re: Andrej Karpathy leaves Tesla

#383

Post by dlkf »

uoaei wrote: post #247
impulser_ wrote: post #8

Earlier quoted context omitted.

Yeah, but Google's vision + lidar tech also doesn't seem any better at solving it either. They have been working on this problem the longest and they aren't even confident enough to produce a product with it. Google is probably the leader in AI and AI research. They are also the leader in data and mapping. They have billions of cash to play with. Yet it seem like they haven't gotten any closer at solving this problem…

> Any reason why everyone seems to be stuck on this problem? Because they're all trying visual- or line-of-sight methods only, I call this the "robo-human" fallacy in ML: trying to automate the processes that humans undergo so that you eventually have a drop-in replacement for a human. But that is a myopic and unimaginative approach because you could be re-assessing the system itself and eliminating inefficiencies th…

I’d like to hear an actual response from the people downvoting this comment. It’s an interesting perspective, and while I can imagine legit criticisims (SPOF, privacy concerns, coordination with vehicles outside the system, etc) actually articulating them would yield a more productive discussion than downvoting without comment.

Re: Andrej Karpathy leaves Tesla

#384

Post by DebtDeflation »

elif wrote: post #289
codechicago277 wrote: post #3

Tesla has bet the company on robotaxis, but their vision only tech stack doesn’t seem capable of solving it, which is a problem because Tesla has repeatedly promised FSD is right around the corner, or less than a year away. It’s hard to believe Karpathy would step down if he felt they were close to solving the problem anytime soon. This announcement comes after a 4 month sabbatical where Karpathy said he wanted to ta…

FSD beta tester here. I think they are minimum 3 years away from anything exciting beta.. but the localization, mapping, and visualization are not the reason. I don't think LIDAR would contribute substantially to improvement. The fundamental flaws are in the decision-making being based upon 10-30 second feature memory, ignoring features outright, and only depending on visible road features instead of persisted map da…

>I think they are minimum 3 years away

I vividly remember a conversation I had with an acquaintance who had recently taken an engineering position with an AV company that occurred circa late 2018. He claimed that they were at most 1 year away. In fact, his exact words were something along the lines of "They're already here. It's just a few edge cases to work out and some regulatory hurdles to overcome."

The reality is that the first 80% of the problem had been solved quickly and significant progress had been made at the time on the next 10%. The end was in sight. Unfortunately, that next 10% ended up taking as long as the first 80% to solve, and the final 10% will likely take decades if it's even possible.

Re: Andrej Karpathy leaves Tesla

#385

Post by rsynnott »

antisthenes wrote: post #316
rsynnott wrote: post #45

Earlier quoted context omitted.

> Any reason why everyone seems to be stuck on this problem? Because it's really, really difficult. A lot of AI-ish stuff pretty rapidly gets to the point where it _looks_ quite impressive, but struggles to make the jump to actual feasibility. Like, there were convincing demos of voice recognition in the mid-90s. You could buy software to transcribe voice on your home computer, and people did. And, now, well, it's be…

> You could buy software to transcribe voice on your home computer, and people did. And, now, well, it's better than in the mid-90s certainly, but you wouldn't trust it to write a transcript, not of anything important. Is this the fault of software or people generally frequently using bad sound equipment in poor and noisy conditions, such as talking on the phone while driving, on the street, poor connection, wind/rai…

Even in ideal recording conditions, machines are quite bad at transcribing human speech. Look at subtitles on live TV shows, typically produced these days by machine transcription; they're typically barely usable.

Re: Andrej Karpathy leaves Tesla

#386

Post by mrmlz »

01100011 wrote: post #362
mrmlz wrote: post #358

Earlier quoted context omitted.

As someone who has worked for almost a decade with safety critical software, system design and functional safety analyses (both aerospace and automotive) etc. I don't think I ever met anyone who believed they would be anywhere near FSD/Level-5 what-have-you in the timeframe they presented. The only ones I met who thought so was managers and software developers and ML-people, i.e. people who has never in their life se…

> As someone who has worked for almost a decade with safety critical software, system design and functional safety analyses And you were right, but most people aren't like you, and people like you don't drive the discussion on what trends are coming. The world needs more folks with your background. Generally no one wants to listen to folks like you because you come off as a Debbie Downer despite being correct most of…

Hah, to be fair I am somewhat of a Negative Nancy from time to time ;)

I think Mercedes has a reasonably sensible game-plan as far as I know. They currently have Level-3 on some highways in Germany which is still freaking hard to achieve, but still a narrow enough scope that you might be able to pull it off.

Time will tell, my bet is its still 10+ years off.

Re: Andrej Karpathy leaves Tesla

#387

Post by elif »

FireBeyond wrote: post #319
elif wrote: post #289

Earlier quoted context omitted.

FSD beta tester here. I think they are minimum 3 years away from anything exciting beta.. but the localization, mapping, and visualization are not the reason. I don't think LIDAR would contribute substantially to improvement. The fundamental flaws are in the decision-making being based upon 10-30 second feature memory, ignoring features outright, and only depending on visible road features instead of persisted map da…

> I think they are minimum 3 years away from anything exciting beta I think that's optimistic. Ten plus years. There's too many exceptions. In my home town, an intersection. Imagine, if you will: Traveling westbound there's four straight ahead lanes. There's a traffic light for each straight lane. The traffic lights will alternate "left two straight, green; right two straight red" and then "left two straight, red; ri…

If the pavement is clearly marked i would give it a good chance. We have a 6-way intersection with similar lighting that it succeeds at... Other than completely and recklessly disregarding the clearly visible "no turn on red" sign.

We also have a road with a reversible lane in the middle. Thankfully it recognizes the red X sign for this. Unfortunately, if it's driving for 11 seconds without seeing anothee, it will suddenly decide to merge into the incoming lane to make an upcoming left...

Re: Andrej Karpathy leaves Tesla

#388

Post by elif »

ncmncm wrote: post #313
elif wrote: post #289

Earlier quoted context omitted.

FSD beta tester here. I think they are minimum 3 years away from anything exciting beta.. but the localization, mapping, and visualization are not the reason. I don't think LIDAR would contribute substantially to improvement. The fundamental flaws are in the decision-making being based upon 10-30 second feature memory, ignoring features outright, and only depending on visible road features instead of persisted map da…

The elephant in the room is fraud. Tesla has been collecting thousands of dollars, each, from car buyers and utterly failing to deliver what it represented, and keeping the money year after year. Would be let GM, Toyota, or Audi do this? Where is the criminal prosecution? Where are the refunds?

I wouldn't call it fraud.

I bought it knowing it was in beta and with no timeline for production release. They haven't represented it as anything more than that.

Being along for the journey is a big part of why I decided to pay.

Re: Andrej Karpathy leaves Tesla

#389

Post by elif »

kaba0 wrote: post #348
elif wrote: post #289

Earlier quoted context omitted.

FSD beta tester here. I think they are minimum 3 years away from anything exciting beta.. but the localization, mapping, and visualization are not the reason. I don't think LIDAR would contribute substantially to improvement. The fundamental flaws are in the decision-making being based upon 10-30 second feature memory, ignoring features outright, and only depending on visible road features instead of persisted map da…

That’s not 3 years away, it’s lacking fundamental knowledge of the world, which might require GAI to begin with. It is just shitty drive assist features sold as much more.

Agree I think 5-6 is more likely

3 years was my minimum, assuming they scrap their current model and begin working on one with real geo-spatial awareness now.

They have the data for it already.

To over-essentialize the problem with their approach, you simply cannot safely drive in Atlanta as a person who has never seen the streets. Humans are no exception. You have to know how each road works, and remember that for next time.

Re: Andrej Karpathy leaves Tesla

#390

Post by asdajksah2123 »

elif wrote: post #388
ncmncm wrote: post #313

Earlier quoted context omitted.

The elephant in the room is fraud. Tesla has been collecting thousands of dollars, each, from car buyers and utterly failing to deliver what it represented, and keeping the money year after year. Would be let GM, Toyota, or Audi do this? Where is the criminal prosecution? Where are the refunds?

I wouldn't call it fraud. I bought it knowing it was in beta and with no timeline for production release. They haven't represented it as anything more than that. Being along for the journey is a big part of why I decided to pay.

Elon Musk has literally said in a public conference keynote that robotaxis would be running in 2019, and earning Tesla owners 30k/yr since 3 years ago, and that it would be financially foolish to buy anything but a Tesla.
Post reply on HN