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

Tesla is building its own AI chips for self-driving cars

techcrunch.com

Post reply on HN
Print view View original

81–90 of 238 posts

Re: Tesla is building its own AI chips for self-driving cars

#81

Post by Retric »

simion314 wrote: post #74
Retric wrote: post #70

Earlier quoted context omitted.

It's an option so insurance companies compare the exact same car with and without that option. In terms of paying attention I really doubt most people are doing this very well. I am going to be stuck in the car either way so by default I am going to be watching the road, but as long as my relative stress level drops that's a huge win. PS: You can make your own estimates. If people pay attention sufficient to avid acc…

But if say me+ a computer program together defeat you at chess you can't conclude that the computer is better then you, just that the computer helped me maybe in some tricky situation , you can't conclude that the computer will win without me. Same with Tesla, this is why I would like to see the number the drivers had to intervene and prevent crashes, without this number the only conclusion is about driver asist and…

IMO the difference between self driving and driver assist is who is usually in charge of the car. If the car is going to hit the breaks because it thinks you are going to hit someone otherwise that's driver assist. If you as a driver need chose to hit the breaks because you don't trust the car that's self driving.

Further, the difference between OK and an accident is normally fractions of a second. So, my suspicion is humans are really bad at this role and only catch the most obvious cases. AKA crap I am an a corn field, not hmm it's not slowing down enough I need to apply slightly more breaking power or I am going to hit that guy simply because you don't have time to react with the second situation.

Having said that, humans may be preventing a lot of issues well before they turn into accidents. The 'wanted to make a wrong turn down a one way street' is something a person can deal with easily, though it's not necessarily going to cause an accident.

Re: Tesla is building its own AI chips for self-driving cars

#82

Post by azernik »

TeMPOraL wrote: post #73
ethbro wrote: post #15

Earlier quoted context omitted.

And yet squishy meatbags manage to sensor fuse a stereoscopic image (plus nascient sound based imaging) into a reliably rendered picture of the world...

Yeah, and they do that with dozens of years of accumulated knowledge. Do not have the impression that what you perceive with your eyes is the photons hitting your retina. All that input is heavily fused with decades of experience in dealing with people and the world.

Experience and some of the most sophisticated evolved machinery in the human brain.

Re: Tesla is building its own AI chips for self-driving cars

#83

Post by Fricken »

tim333 wrote: post #77
Fricken wrote: post #45

Right now, with HW2 Tesla is in kind of an awkward no man's land where their hardware is overkill for basic ADAS, but nowhere near enough for any kind of autonomy that allows the driver to turn their attention away from the driving task. By going to a custom ASIC with inferencing. I have little doubt about Musk's claim that it'll offer a 10x improvement over the current stripped down PX Drive2 variant for deep learni…

It's hard to say for sure they are no where near full autonomy in their research although their current vehicles crashing into barriers and fire trucks obviously don't cut it.

I'm only commenting on what little there is to glean about the compute, but there are so many other red flags about their autonomous development strategy. Crippling per-unit cost constraints, No Lidar, SLAM-lite, ANN heavy, insufficient redundancy and they've got a major data collection problem because they're relying on what they can get over the air. Plus Musk scared away most of their their top engineers.

Re: Tesla is building its own AI chips for self-driving cars

#84

Post by r3bl »

squarefoot wrote: post #62

Curiously there already is (was?) a probably eastern European chip manufacturer named Tesla (search images for "tesla ttl chips") but I wasn't able to find info on the manufacturer and if they're still in business.

You're thinking of this: https://en.wikipedia.org/wiki/Tesla_(Czechoslovak_company)

It's not a unique name in any shape or form. There's a smartphone "manufacturer" from Serbia that uses Tesla as its brand name: https://tesla.info/en/

Re: Tesla is building its own AI chips for self-driving cars

#86

Post by nbeleski »

3pt14159 wrote: post #25
titzer wrote: post #3

This is pretty bad from a hardware/software verification perspective. People who care about correctness can just go home, I guess. It's amazing that we're gonna let these things on the streets. When you weigh almost 2 tons and can go >100mph, it doesn't matter if you have a gun mounted on you or not, you are a weapon .

I worry less about software / hardware crashes leading to death and more about cyberattack. I've written about it before[0]: We need regulations before a mass breach of autonomous systems happens. Not after. This whole bullshit where corporations claim all sorts of things about their security then get owned trivially has happened for every OS and device out there. The supposedly secure QNX was installed into countles…

We already have a number on embedded processors and controllers in cars today, including systems responsible for critical operational parts of the car.

How autonomous systems will make anything different? Why do you believe people aren't currently hacking cars?

I don't intend to be critical at all, I believe this is a very interesting point too that should be discussed. (I hope to check the links on my interval).

Re: Tesla is building its own AI chips for self-driving cars

#87

Post by Retric »

simion314 wrote: post #79
Retric wrote: post #70

Earlier quoted context omitted.

It's an option so insurance companies compare the exact same car with and without that option. In terms of paying attention I really doubt most people are doing this very well. I am going to be stuck in the car either way so by default I am going to be watching the road, but as long as my relative stress level drops that's a huge win. PS: You can make your own estimates. If people pay attention sufficient to avid acc…

>PS: You can make your own estimates. If people pay attention sufficient to avid accidents 99% of the time the car might be in 100x more accidents without drivers assuming those drivers where perfect. Drop that to IMO more likely 75% and it's closer to 4x at the absolute best case. You can't invent numbers and statistics. Remember the Tesla that crashed in a side barrier and killed the driver, people could reproduce…

Computer failures are reproducible, but you don't get "1000 Teslas driving trough that section in the problematic interval" which shows up in the actual statistics.

Computers fail in different ways than humans do. But, that does not mean you can't reason about failure modes. If you ask someone to pay attention to something for an hour without doing anything most people are just really bad at this. Based on that I am rather shocked how well Tesla's systems work as I was expecting vastly more problems.

Could I have flipped to far in the other direction probably. But, I base things on my experience and the data I have available not just unchanging gut feelings.

Re: Tesla is building its own AI chips for self-driving cars

#88

Post by simion314 »

Retric wrote: post #81
simion314 wrote: post #74

Earlier quoted context omitted.

But if say me+ a computer program together defeat you at chess you can't conclude that the computer is better then you, just that the computer helped me maybe in some tricky situation , you can't conclude that the computer will win without me. Same with Tesla, this is why I would like to see the number the drivers had to intervene and prevent crashes, without this number the only conclusion is about driver asist and…

IMO the difference between self driving and driver assist is who is usually in charge of the car. If the car is going to hit the breaks because it thinks you are going to hit someone otherwise that's driver assist. If you as a driver need chose to hit the breaks because you don't trust the car that's self driving. Further, the difference between OK and an accident is normally fractions of a second. So, my suspicion i…

Autopilot/self driving is more complicated there are 5 levels I think.

Anyway I am waiting for some government institution or maybe the lawsuit to get more data then we can use that to make informed conclusions.

Re: Tesla is building its own AI chips for self-driving cars

#89

Post by jaggederest »

Nae3Au5x wrote: post #21
titzer wrote: post #3

This is pretty bad from a hardware/software verification perspective. People who care about correctness can just go home, I guess. It's amazing that we're gonna let these things on the streets. When you weigh almost 2 tons and can go >100mph, it doesn't matter if you have a gun mounted on you or not, you are a weapon .

In the same sense that any object that can potentially kill someone (hint: everything) is a weapon. Which is technically correct but a useless observation. If we use a more narrow definition of weapon, e.g. a tool optimized for the primary purpose of injuring or killing people then it certainly is not a weapon.

As a modification of the Kzinti Lesson: A vehicle is a weapon in direct proportion to it's effectiveness as a vehicle.

Re: Tesla is building its own AI chips for self-driving cars

#90

Post by titzer »

colordrops wrote: post #75
titzer wrote: post #72

Earlier quoted context omitted.

For example, chips have flaws. Hardware designers work incredibly hard to verify that their chip designs are functionally correct, because the cost of a correctness bug can be catastrophic. With the proliferation of custom hardware designs, verification of the whole stack is even more important. It's bad from the perspective that now instead of building on top and reusing a chip company's huge investment in a verifie…

Being skeptical is fine, but it would be fair to back such a strong statement up with facts rather than completely ungrounded speculation.

This is not how verification works, generally. You have to prove your chip is correct.
Post reply on HN