Live data from Hacker News

Sub-$200 Lidar could reshuffle auto sensor economics

spectrum.ieee.org

411–420 of 582 posts

Re: Sub-$200 Lidar could reshuffle auto sensor economics

#411
post #378
post #347

Earlier quoted context omitted.

Just for the record though, Musk isn't blindly anti-LIDAR. He has said (and I think this is an objective fact) that all existing roads and driving are based on vision (which is what all humans do). So that should technically be sufficient. SpaceX uses LIDAR for their docking systems. I would argue that yes, we do use vision but we get that "lidar depth" from our stereo vision. And that used to be why I thought camera…

Humans aren't using only fixed vision for driving. This is such a tiresome thing to see repeated in every discussion about self driving. You're listening to the road and car sounds around you. You're feeling vibration on the road. You're feeling feedback on the steering wheel. You're using a combination of monocular and binocular depth perception - plus, your eyes are not a fixed focal length "cameras". You're moving…

In theory, a computer should be able to do the same. It could do sensor fusion with even more sense modalities than we have. It could have an array of cameras and potentially out-do our stereo vision, or perhaps even use some lightfield magic to (virtually) analyze the same scene with multiple optical paths.

However, there is also a lot of interaction between our perceptual system and cognition. Just for depth perception, we're doing a lot of temporal analysis. We track moving objects and infer distance from assumptions about scale and object permanence. We don't just repeatedly make depth maps from 2D imagery.

The brute-force approach is something like training visual language models (VLMs). E.g. you could train on lots of movies and be able to predict "what happens next" in the imaging world.

But, compared to LLMs, there is a bigger gap between the model and the application domain with VLMs. It may seem like LLMs are being applied to lots of domains, but most are just tiny variations on the same task of "writing what comes next", which is exactly what they were trained on. Unfortunately, driving is not "painting what comes next" in the same way as all these LLM writing hacks. There is still a big gap between that predictive layer, planning, and executing. Our giant corpus of movies does not really provide the ready-made training data to go after those bigger problems.

Re: Sub-$200 Lidar could reshuffle auto sensor economics

#412
post #268

Before y'all say that now everyone will be able to get Waymo's sensor suite for hundreds of dollars instead of tens of thousands, that's the easy part. Waymo benefits from Google's unparalleled geospatial data. Waymo also has a support architecture that doesn't depend on real time remote operation, which can't be implemented reliably in almost all cases. You can't be following your supposedly unsupervised cars with a…

I was using cruise control on the highway yesterday and thinking: this is like very cheap very crude self-driving. And you know what? In its limited UNIX-like way, it's great: the car does a much better job of gradually injecting fuel than I, with my brick-like human foot, can do. Robot 1, human 0.

And from there it's easy to think: couldn't the car also detect white lines and stay within them? It doesn't have to be perfect; it can be cruise control++. If it errs a little, I can save it. But otherwise, this is a function I'd love to use if it was available, for a sub $1000 price point.

Re: Sub-$200 Lidar could reshuffle auto sensor economics

#413
post #306

Earlier quoted context omitted.

I’m not disagreeing with what you’re saying, but does Alphabet actually intend Waymo to be a trillion dollar retail car business itself, selling cars to everyone? Or would they be happy to sell all those super cool things to OEMs? In a world where “everyone” can make a car affordable that can run Waymo’s software, they may be happy to license all that to “everyone” and simply collect fat royalty checks, à la Microsof…

> but does Alphabet actually intend Waymo to be a trillion dollar retail car business itself, selling cars to everyone? Google doesn't do retail other than Chromecast and Pixel phones, and that is already annoying to them as it is because it involves something Google is notoriously bad at - actual customer support. Starting up a car brand is orders of magnitude worse. For one, people actually need to trust your brand…

> Starting up a car brand is orders of magnitude worse.

Tesla did it, and is more valuable than most other car brands added together. They had a novel product: a good EV that was fun to drive. Is that a unique situation? Could a truly autonomous car launch do it?

Your arguments make sense in themselves, but maybe underestimate the revolutionary value that a level 4 car would provide.

Re: Sub-$200 Lidar could reshuffle auto sensor economics

#414

Earlier quoted context omitted.

> but does Alphabet actually intend Waymo to be a trillion dollar retail car business itself, selling cars to everyone? Google doesn't do retail other than Chromecast and Pixel phones, and that is already annoying to them as it is because it involves something Google is notoriously bad at - actual customer support. Starting up a car brand is orders of magnitude worse. For one, people actually need to trust your brand…

> Starting up a car brand is orders of magnitude worse. Tesla did it, and is more valuable than most other car brands added together. They had a novel product: a good EV that was fun to drive. Is that a unique situation? Could a truly autonomous car launch do it? Your arguments make sense in themselves, but maybe underestimate the revolutionary value that a level 4 car would provide.

> Tesla did it, and is more valuable than most other car brands added together.

Half of Tesla's value is hopium, the rest of it is pure trust in that the current government will continue propping Elon up (even if he personally ran afoul of the Dear Leader). A lot of the promises Elon made, particularly when it comes to FSD, had to be tracked back and I don't see them ever coming to fruition - at least not for the cars that don't have LIDAR hardware.

Re: Sub-$200 Lidar could reshuffle auto sensor economics

#415
post #268

Before y'all say that now everyone will be able to get Waymo's sensor suite for hundreds of dollars instead of tens of thousands, that's the easy part. Waymo benefits from Google's unparalleled geospatial data. Waymo also has a support architecture that doesn't depend on real time remote operation, which can't be implemented reliably in almost all cases. You can't be following your supposedly unsupervised cars with a…

I was using cruise control on the highway yesterday and thinking: this is like very cheap very crude self-driving. And you know what? In its limited UNIX-like way, it's great: the car does a much better job of gradually injecting fuel than I, with my brick-like human foot, can do. Robot 1, human 0. And from there it's easy to think: couldn't the car also detect white lines and stay within them? It doesn't have to be…

Is this comment from 2010? Maybe I'm missing your point, but it seems you would be shocked by what modern cars are capable of.

Re: Sub-$200 Lidar could reshuffle auto sensor economics

#416

Low cost, sub $200 automotive grade LIDAR sensors are already available. Cepton Technologies offers Nova [0], Nova-Ultra [1] sensors both at a sub-$100 price point [2]. These feature a 120°(H) x 90°(V) FOV at 50m, with 2.7M points per second sampling. Velodyne introduced Velabit in 2021, for $100. Boasting 100m range and a 60-degree horizontal FoV x 10-degree vertical FoV. The article claims that: > What distinguishe…

>Cepton Technologies offers Nova [0], Nova-Ultra [1] sensors both at a sub-$100 price point

Where? How? I'm only seeing the Nova on ebay for between $4000 and $5000.

Re: Sub-$200 Lidar could reshuffle auto sensor economics

#417
post #347

Earlier quoted context omitted.

Based on that list it boils down to 2 things it seems: - cost (no longer a problem) - too much code needed and it bloats the data pipelines. Does anyone have any actual evidence of this being the case? Like yes, code would be needed, but why is that innately a bad thing? Bloated data pipelines feels like another hand-wave when I think if you do it right it’s fine. As proven by Waymo. Really curious if any Tesla engin…

Just for the record though, Musk isn't blindly anti-LIDAR. He has said (and I think this is an objective fact) that all existing roads and driving are based on vision (which is what all humans do). So that should technically be sufficient. SpaceX uses LIDAR for their docking systems. I would argue that yes, we do use vision but we get that "lidar depth" from our stereo vision. And that used to be why I thought camera…

Mentioning gaussian splatting for why we don't need lidar depth is a great example of Musk-esque technobabble; surface level seemingly correct, but nonsense to any practitioner. Because one of the biggest problems of all SfM techniques is that the results are scale ambiguous, so they do not in fact recover that crucial real-world depth measurement you get from lidar.

Now you might say "use a depth model to estimate metric depth" and I think if you spend 5 minutes thinking about why a magic math box that pretends to recover real depth from a single 2D image is a very very sketchy proposition when you need it to be correct for emergency braking versus some TikTok bokeh filter you will see that also doesn't get you far.

Re: Sub-$200 Lidar could reshuffle auto sensor economics

#418
post #268

Before y'all say that now everyone will be able to get Waymo's sensor suite for hundreds of dollars instead of tens of thousands, that's the easy part. Waymo benefits from Google's unparalleled geospatial data. Waymo also has a support architecture that doesn't depend on real time remote operation, which can't be implemented reliably in almost all cases. You can't be following your supposedly unsupervised cars with a…

I was using cruise control on the highway yesterday and thinking: this is like very cheap very crude self-driving. And you know what? In its limited UNIX-like way, it's great: the car does a much better job of gradually injecting fuel than I, with my brick-like human foot, can do. Robot 1, human 0. And from there it's easy to think: couldn't the car also detect white lines and stay within them? It doesn't have to be…

I think of Tesla autopilot as sophisticated cruise control. Can perform most driving tasks better than I can, saves a lot of cognitive work, still needs close of my 100% attention.

Re: Sub-$200 Lidar could reshuffle auto sensor economics

#419

Since lidar has distance information and cameras do not, it was always a ridiculous idea by a certain company to use cameras only. Lidar using cars are going to replace at least the ones that don't make use of this obvious answer to obstacle detection challenges.

One camera can't really produce depth/distance information, but two cameras sure can. The eyes in your head don't capture distance information individually, but with two eyes you can infer distance.

Re: Sub-$200 Lidar could reshuffle auto sensor economics

#420
post #343

Earlier quoted context omitted.

And barely gets enforced

2775 fines for a total of €6.8B since July 2018. It's not A LOT (I would hope for A LOT MORE fines), but it's not nothing. https://www.enforcementtracker.com/

It’s very interesting. Thanks for sharing.

But also kinda weird. There seems to be a lot of fines for hospitals for example.

Some Portuguese hospital was fined €400,000 for ‘Insufficient technical and organisational measures to ensure information security’

Post reply on HN