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

Tesla AI Day [video]

livestream.tesla.com

Post reply on HN
Print view View original

71–80 of 187 posts

Re: Tesla AI Day [video]

#71

Post by tw04 »

MontyCarloHall wrote: post #46
TheJoeMan wrote: post #31

I might be completely wrong here, but hasn’t their collision incidents often been with weird things in the road such as sideways 18 wheelers and yellow lines heading into barricades? They’ve put lots of effort into tracking some pedestrian waaaaay over on the side passing behind trees, and that is really really impressive. But it seems to me the only reason it doesn’t hit a car in front on the road is because the neu…

> I would like to see some system that also doesn’t collide into any sort of random object directly in front. The flip side of this is false positives: you don’t want a car slamming on the brakes because of a plastic shopping bag or other soft piece of trash lying in the road. This is a hard problem, regardless of technology employed. On the other hand, you need the car to distinguish a mattress lying in the road fro…

>The flip side of this is false positives: you don’t want a car slamming on the brakes because of a plastic shopping bag or other soft piece of trash lying in the road. This is a hard problem, regardless of technology employed.

Is this really a problem with current vehicles? I've had ford, gm, and dodge/ram vehicles over the last ~8 years with forward collision sensors and automatic braking and never once had any of them slam on the brakes due to a plastic shopping bag or other piece of soft trash lying in the road. I honestly haven't even had them give warning from one of those items being in the road (or any item that wasn't a car to be honest).

Re: Tesla AI Day [video]

#72

Post by GrantZvolsky »

Invictus0 wrote: post #15

I like Andrej and I think he's a brilliant guy. At the end of the day though, there is no way for me or anyone to audit the neural networks or their training data. How can anyone trust them? I think this is really a fundamental problem with using neural networks in self-driving cars--I don't want to put my life in the hands of a fragile system that I, and no one alive, can understand. We are being asked to close our…

How can anyone trust any other complex system? Even something as seemingly trivial as a bridge. Picture driving on an unfamiliar road when suddenly there's a bridge ahead. How can anyone in that situation trust that bridge? Without knowing how it was built, whether it is in a state of disrepair, or whether the two sides of the bridge connect without a precipice inbetween[1]. The answer is that we don't need to completely trust a system as long as our experience suggests reasonable safety. Once unassisted Tesla proves to be safer than human drivers, we still won't trust it, but that won't prevent us from using it.

[1]: Not relevant to the above comment, but I once encountered such a bridge. It was while driving in the dark, at about 1am and with no traffic, when the navigation in my, iirc, Android 2.1 phone, suggested I cross an unfinished bridge. The only warning sign ahead of the precipice were a couple of traffic cones. The road was familiar, though, and I drove very carefully, curious about the new bridge.

Re: Tesla AI Day [video]

#73

Post by ajross »

nightski wrote: post #67
MontyCarloHall wrote: post #62

Earlier quoted context omitted.

> Because human-caused accidents are not uniformly distributed among the driving population. This is such a great point. Controlling for driving circumstances (e.g. weather, location, etc.), human caused accidents are not uniformly distributed across the population, but machine caused accidents are. This, in my mind, is the fundamental reason people are distrustful of self-driving cars. Everyone thinks they’re far be…

It's not just about skill. A lot of accidents are under the influence. That's a choice. While you can't control what other drivers are doing (and neither can the Tesla), you can at least significantly increase your odds of safety by some simple choices such as this.

> A lot of accidents are under the influence. That's a choice.

Only for one of the drivers! There are at least two cars involved in almost every fatal accident.

I made this point elsewhere but it bears repeating. Even if you think your driving is so perfect you can't benefit from a self-driving car, don't you want everyone else in one?

Re: Tesla AI Day [video]

#74

Post by abc_lisper »

ModernMech wrote: post #30
2bitencryption wrote: post #28

A big part of the presentation so far was an engineer describing the huge difficulty of stitching together the multiple cameras into one vector space that can be the input to the network, instead of treating each camera individually. Seems like the biggest problem was each pixel from a camera does not tell you how far away it is, so even if you know the camera is X feet off the grounding pointing at Y degrees, you do…

Yes. This is exactly what LIDAR is for. This is the reason every single team in the 2007 DARPA Urban challenge that finished the race was equipped with a Velodyne laser. We knew it was a critical enabling technology 14 years ago, and it baffles me Tesla is eschewing it. Tesla's lack of this technology is the main reason I feel they will never achieve what they claim to without a huge breakthrough in AI.

Idk.. animals seem to do fine without a lidar

Re: Tesla AI Day [video]

#75

Post by ajross »

tw04 wrote: post #71
MontyCarloHall wrote: post #46

Earlier quoted context omitted.

> I would like to see some system that also doesn’t collide into any sort of random object directly in front. The flip side of this is false positives: you don’t want a car slamming on the brakes because of a plastic shopping bag or other soft piece of trash lying in the road. This is a hard problem, regardless of technology employed. On the other hand, you need the car to distinguish a mattress lying in the road fro…

>The flip side of this is false positives: you don’t want a car slamming on the brakes because of a plastic shopping bag or other soft piece of trash lying in the road. This is a hard problem, regardless of technology employed. Is this really a problem with current vehicles? I've had ford, gm, and dodge/ram vehicles over the last ~8 years with forward collision sensors and automatic braking and never once had any of…

Radar isn't going to signal on a shopping bag, but I bet you it would for a mylar baloon. Existing radar-based collision sensors really aren't that accurate anyway. They are quite late with detection in most cases, precisely because they have to be tuned not to brake on false positives. They want a good strong signal, at which point the best they can do is reduce the speed of an inevitable collision.

Re: Tesla AI Day [video]

#76

Post by joshuamorton »

karpierz wrote: post #44
bpodgursky wrote: post #24

Earlier quoted context omitted.

Unless your model actually has trillions of parameters (and it doesn't, even gpt-3 only has 175 billion) it is not even possible to overfit on 1.4 trillion training inputs. You can't actually pigeonhole it.

Suppose that you train a neural network to predict the next number in an arithmetic sequence (a, a+b, a+2b, a+3b, a+4b, ...). As input it gets two numbers, the last number and the current number and has to predict the next one. Suppose you had 1.4 trillion examples in the following test set (using a model with 175 billion parameters): (1,2)->3 (2,3)->4 (3,4)->5 ... Do you think it is possible to overfit and score per…

I think you've specified this problem in a very strange way. But if you're saying that you're trying to train on the specific dataset where a = 1 and b = 1, then your model will fit the data perfectly with 175 billion parameters. It will also fit the data perfectly with, like, 15 parameters.

If you're trying to fit to some more complex space where a and b are unknown and you're given 3 numbers in the sequence, then what you're trying to fit is `f(a, b) = a + 2(b - a)` (or 2b - a, however you want to represent it), which is a swell function, but if you only give data that can be equally represented by `f(a, b) = b + 1`, you're mis-training your model.

But you could once again do that with a model with a dozen parameters. In both cases, the issue isn't overfitting, but misrepresentative data.

Re: Tesla AI Day [video]

#77

Post by awal2 »

mdorazio wrote: post #51
rnaud wrote: post #36

Earlier quoted context omitted.

Too expensive + Your eyes already do the distance estimation and they are basically cameras so we can do this just as well or better.

> Your eyes already do the distance estimation and they are basically cameras so we can do this just as well or better. This is not really accurate. Our brains do the distance estimation, and they use all kinds of tricks and contextual clues to do it, not just parallax (~2.5 inch parallax for objects more than ~100ft away isn't all that helpful). And this is the whole problem with vision-only autonomous driving - ML…

Monocular depth estimation has gotten really good recently though[0]. Not saying this one paper/method is 100% sufficient, but we're closing the gap in this one capability (depth estimation from pure vision) quite rapidly.

[0] https://roxanneluo.github.io/Consistent-Video-Depth-Estimati...

Re: Tesla AI Day [video]

#78

Post by ec109685 »

2bitencryption wrote: post #28

A big part of the presentation so far was an engineer describing the huge difficulty of stitching together the multiple cameras into one vector space that can be the input to the network, instead of treating each camera individually. Seems like the biggest problem was each pixel from a camera does not tell you how far away it is, so even if you know the camera is X feet off the grounding pointing at Y degrees, you do…

He talks about the approach here. They actually used radar and LIDAR to train the depth sensing neural network: https://t.co/osmEEgkgtL?amp=1

Re: Tesla AI Day [video]

#79

Post by ipsum2 »

tw04 wrote: post #71
MontyCarloHall wrote: post #46

Earlier quoted context omitted.

> I would like to see some system that also doesn’t collide into any sort of random object directly in front. The flip side of this is false positives: you don’t want a car slamming on the brakes because of a plastic shopping bag or other soft piece of trash lying in the road. This is a hard problem, regardless of technology employed. On the other hand, you need the car to distinguish a mattress lying in the road fro…

>The flip side of this is false positives: you don’t want a car slamming on the brakes because of a plastic shopping bag or other soft piece of trash lying in the road. This is a hard problem, regardless of technology employed. Is this really a problem with current vehicles? I've had ford, gm, and dodge/ram vehicles over the last ~8 years with forward collision sensors and automatic braking and never once had any of…

This is because those cars are using radar (a plastic bag probably would show up as a small blip), but Tesla is removing radar from their cars to rely solely on cameras.

Re: Tesla AI Day [video]

#80

Post by ajross »

WrtCdEvrydy wrote: post #66
MontyCarloHall wrote: post #46

Earlier quoted context omitted.

> I would like to see some system that also doesn’t collide into any sort of random object directly in front. The flip side of this is false positives: you don’t want a car slamming on the brakes because of a plastic shopping bag or other soft piece of trash lying in the road. This is a hard problem, regardless of technology employed. On the other hand, you need the car to distinguish a mattress lying in the road fro…

> Because for whatever reason (hubris?) Tesla insists on eschewing LIDAR/RADAR for 100% camera based systems. It's a retrofit problem. The moment you admit that you need LIDAR/RADAR, every Tesla on the road loses value unless you're willing to retrofit them all. Comma.ai has sidetracked this issue by basically using the built-in sensors on the vehicles, and there's rumors of a camera-only version for those without ra…

> Comma.ai has sidetracked this issue by basically using the built-in sensors on the vehicles

Comma/OpenPilot doesn't work with Lidar anywhere I'm aware of...

Post reply on HN