Live data from Hacker News

Andrej Karpathy leaves Tesla

twitter.com

331–340 of 441 posts

Re: Andrej Karpathy leaves Tesla

#331

Earlier quoted context omitted.

FSD may be a huge fraud but the whole company is not. The cars are real!

The cars are real but they are... Mediocre. The first year of ownership is an amazing honeymoon period, assuming you got one with decent build quality. But the parts are cheap, and they break quickly. I've had one for 5+ years and it's gradually become something I prefer driving less and less.

speak for yourself

i friggin love our model 3. two years and almost 60,000 miles in!

i want to get the model s refresh and be a two tesla household, but my pay needs to go up a little :)

Re: Andrej Karpathy leaves Tesla

#332

Earlier quoted context omitted.

Yes, but Steve Jobs delivered. He was rude and whatnot, but far from a fraud.

And Jobs' "reality distortion field" was primarily used to motivate / burn-out his team by demanding things that were technically very difficult but feasible (one of the original anecdotes was about reducing the boot-time for the original Macintosh), not something like solving AGI which FSD would ~apparently require.

I fully believe Elon and Jim Keller truly thought the deep learning HW delivered by Tesla was capable of self-driving in the early 201x time frame because, frankly, a lot of us did. More than likely, like the rest of us, they slowly came to see that the problem was much more difficult, and deep learning much less capable, than we had originally envisioned. Most of us didn't make a bunch of stupid promises to shareholders based on the original, incorrect assumptions though. Unlike Elon, we're able to admit we were wrong.

Re: Andrej Karpathy leaves Tesla

#333

Earlier quoted context omitted.

> I don't see legacy auto being capable of refactoring their cars, manufacturing processes and business models fast enough to survive. 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…

> 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…

> But they don't have the 30% gross margin that Tesla does.

Sure, but they'll make it up on volume. Tesla also will have much lower than 30% gross margins in 10 years -but can it catch up with "legacy auto" on volume?

Re: Andrej Karpathy leaves Tesla

#334
post #38
post #11

This comes after yesterday's news of Tesla letting 229 people go who were working on Autopilot: https://techcrunch.com/2022/07/12/tesla-laying-off-229-autop... .

That's pretty much a totally different category of employee, though: "Tesla is laying off 229 data annotation employees who are part of the company’s larger Autopilot team [...] Most of the workers were in moderately low-skilled, low-wage jobs, such as Autopilot data labeling" Data annotation can be cheaply outsourced or scaled up and down without much affecting progress on self-driving. Karpathy leaving says more ab…

> Data annotation can be cheaply outsourced or scaled up

Only if you don't care about quality. In my experience using annotation companies leads to a conflict of interest - annotate more to earn more, or annotate better.

Re: Andrej Karpathy leaves Tesla

#335

Earlier quoted context omitted.

The counterexample to your point is that brand new 16-year-old human drivers seem to do a pretty good job of driving with very little training data. Yes, they do kill people and they're much worse drivers than 40-year-olds who have much more training data, but the fact that 16YOs kill as few people as they do with so little training data means some other mechanism than raw data quantity is important.

It's not a counterexample. Your 16-year old human has millions of years of evolution into its brain. That's a lot of pre-training data.

Are you suggesting that orange cones and semi-trailers cutting you off were common on the African savannah?

Re: Andrej Karpathy leaves Tesla

#336
post #91

Earlier quoted context omitted.

> any of those could be forged / faked without some kind of way of securely validating the message Any reason routine crypto methods would not solve this? Seems like one of the easier parts to me.

Who controls the signing keys and the whole signing process? What about key revocation if someone steals the key? Will a municipality in Texas really be willing to not be allowed to create a new stoplight without approval from the federal agency in charge of the keys? What’s to stop someone stealing a “real” stoplight from bumfuck nowhere and putting it in the middle of the 101 at rush hour? What about replay attacks…

You raise some genuine possible concerns but generically I'd probably tend to just say, "laws will stop them" just like they already stop someone deliberately endangering lives by placing a real fake stop light in the middle of the 101 at rush hour.

The real concern would be whether someone can engineer a terrorist level mass scale attack but as long as it requires physical tampering that adds up to a tremendous amount of work. So if the signalling is largely burned into fixed infastructure it eliminates a lot of that or at least sets the bar high enough that its probably more work than various other types of attack that are likely to be just as impactful.

Re: Andrej Karpathy leaves Tesla

#337
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…

It’s naive to call yourself beta tester. People who have access to beta are basically just a part of PR exercise by Tesla, that tries to hide lack of meaningful progress towards multiple missed deadlines. You aren’t beta testing complex automation system that’s operating on public roads. To do any meaningful testing you should have defined operation domain, specific behaviors to test, direct line to the engineering t…

You're thinking of QA testing. Beta testing is the final stage of testing before release and aims to test the system in as close as realistic environment as possible - that's why betas are often seeded to the general public, you want to find out if your system doesn't work on some configuration or circumstances that you didn't anticipate in internal testing. Better to have it fail for a limited number of users who signed up for that possibility than to have to do a Cyberpunk style recall.

Re: Andrej Karpathy leaves Tesla

#338

Earlier quoted context omitted.

This sounds like solid evidence for the fact that he won't ever get his hands on a time machine.

Or it'll be like the time machine in Primer, where you can't go back before the time machine was turned on.

That kinda only works if we’re the only civilization in the universe, right? What if some other civilization discovered it a long time ago?

Re: Andrej Karpathy leaves Tesla

#339

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…

Isn't Tesla supposed to be producing Optimus, their human-like android, next year? Elon has been over-promising(i.e. flat out lying) about self-driving every year since.. 2014(there's a youtube video compilation of it)? It seems like his strategy is to just come up with increasingly grandiose promises every year when he fails to deliver on his past promises. He's trapped in his swirling vortex of bullshit. Very worry…

Elon said they might have a prototype in 2022. He didn’t give any other timelines.

Re: Andrej Karpathy leaves Tesla

#340
post #48

Earlier quoted context omitted.

> Tesla has bet the company on robotaxis, How so? They're not selling robotaxis or building factories to build them > Tesla has repeatedly promised FSD is right around the corner Which means it's years away and/or "FSD" means "automatic cruise control and lane keep assist" or whatever standard feature from auto manufacturers they've renamed

> How so? Because they chose to back themselves into that corner. Musk says that Tesla is worth nothing without full self-driving. Certainly it's the only thing left to justify the stock price: https://electrek.co/2022/06/15/elon-musk-solving-self-drivin... > Which means it's years away and/or "FSD" means "automatic cruise control and lane keep assist" Well, more precisely it means Musk has been lying about it for ni…

In the long run it’s all robotaxis and Elon sees that. Consumer purchases of cars will basically end.
Post reply on HN