Can anybody give insights in to how key this guy is to Tesla. Is it a big deal? What’s his USP?
Andrej Karpathy leaves Tesla
71–80 of 441 posts
Re: Andrej Karpathy leaves Tesla
#72Cars will never drive themselves.
Re: Andrej Karpathy leaves Tesla
#73The autopilot part of Tesla has never made much sense. Is Tesla the electric car company, or is it the luxury car company? Either way, why does the power train (EV or ICE) come into play at all? Not only that, but Tesla has played the Innovator's Dilemma game from the position of the upstart financially, but targeted the segment of the market that incumbents will defend to the death (luxury cars). Tesla could have go…
And yet Tesla is dominant player in luxury segments they have products in
Re: Andrej Karpathy leaves Tesla
#74Earlier quoted context omitted.
Irrespective of what he said, it won't be 0, Tesla would still be one of the best electric cars available.
For how long though? It seems like traditional automakers have mostly figured out EVs now. I love my Chevy EV, and it's 4 years old now. Similarly, the Kia EV I recently drove was excellent(although the Bolt one pedal driving is better IMHO).
I'm now seeing a lot of VW ID.4s and Skoda Enyaqs (basically the same car, different styling) in Sweden.
Re: Andrej Karpathy leaves Tesla
#75Earlier 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? ML maximalism focused on the narrow problem of 'solving driving' while not recognizing that any task as complex as driving requires probably something closer to general intelligence, and theoretically the field has been impoverished in favor of "throw more graphics cards at everything".
Even then, even if you can solve every case involving actual roads with perfect markings and intact signs and functioning signals, I’ve found myself just this week:
- driving across an unmarked grassy mound to park a car at a store in their designated area.
- paying a fee with coins to enter and exit a toll road.
- stopping to move around roadworks based solely on hand signals from one of the workers.
These aren’t even scratching the surface in terms of edge cases that could be encountered regularly.
Re: Andrej Karpathy leaves Tesla
#76Earlier quoted context omitted.
I remain convinced that "real" self driving (as in: go ahead and sleep in the backseat) will never happen without changes to road infrastructure and possibly some sort of segregation between robot-driven cars and people-driven cars. Things like traffic signals that actively communicate their status to nearby robot cars (more than just a red lamp that can be occluded by weather, other vehicles, or mud on the camera le…
What if we just, you know, walk a bit more instead? Or even cycle? I hear great things.
The pattern of zoning and development that the U.S. has used for the past 70 years assumes that nobody wants to walk or cycle anywhere and that they feel so strongly about it that they want cities to be built such that people will die if they attempt to do so. I grew up in a neighborhood where the elementary school was about half a mile away... on the other side of a five lane highway with no crosswalk and no sidewalk on the other side. Zero chance to assert your independence as a kid. Zero chance to live responsibly as an adult.
A part of the reason people wish they had FSD so badly is because they want to be rescued from this fundamental failure of NA-style urban planning that necessities driving, all the time, across both short and long distances.
Re: Andrej Karpathy leaves Tesla
#77Earlier 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…
Consumer and grid-scale battery storage
Solar
And I also think they could do some clever stuff with home HVAC, possibly using waste heat from crypto miners as the H.
Lastly, afaik they went camera-only in their cheap cars (3/Y) and still use fancy stuff in the S/X.
They gave fsd customers new computers once. What’s to stop them from going back to vision+lidar or whatever once the parts are available and retrofitting as needed?
The whole ‘vision-only is better’ gag seemed like an obvious ploy to keep being able to ship cars from the beginning of supply chain problems.
And yeah, Elon does not appear to be a good person.
Re: Andrej Karpathy leaves Tesla
#78Earlier 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 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…
I'm old enough to remember all the breathless pronouncements in 2015 about how self-driving cars would be everywhere within 5 years. I was skeptical then and I'm still skeptical that it will ever happen outside of some limited range of well known paths that have been pre-determined to be safe - ie we'll have (and we do have) some self driving cars but they'll essentially be on a closed course and will not leave the course. There are already some small bus lines like that - the bus just goes around a closed circuit picking up and dropping off people always taking the same route.
Re: Andrej Karpathy leaves Tesla
#79Earlier quoted context omitted.
Because self-driving has a bunch of tricky edge cases and most of them will kill people. Problems with hundreds of important edge cases cannot be solved by simply throwing more training data at the problem; that's how you solve AI problems in a "dumb" manner, and it works for lots of problems (like recognizing dogs in images) -- but not for self-driving. To solve the self-driving problem we need "smart" A.I., which m…
And we haven't even adressed that drving is not a purely technical endeavor, it's largely a social one.
A social one where you use experience to make inferences about what is going to happen next. Who hasn't observed an aggressive driver coming up from behind, weaving around other cars and said to themselves something like "that guy is going to cut me off, better ease off the gas so there's a little extra room in front for when he does". That level of situational awareness isn't coded for is it?
Edit: I have a whole mental model for other drivers and different approaches for them. Someone driving like a grandma? Pass when available. Nervous/erratic/lost driver? Keep extra distance then pass as soon as possible. Aggressive driver? Relax, give some space and let them get ahead. And so on. I get that stereotyping is bad but ignoring the subtle signals other drivers give off seems like it would be myopic. An AI that doesn't anticipate what others will do on the road will always be reactive rather than proactive.
Re: Andrej Karpathy leaves Tesla
#80hmm, I think the real time aspects of self driving are hard. There is lots of promise with masked language and image modeling, but plenty of work to be done, even for non real time applications.