Waymo will beat Tesla FSD because of LIDAR

[CyberGus]

People can close one eye and still throw a basketball or shoot a gun and hit small targets.

Famously, Pataudi, an Indian cricketer lost an eye but he was still able to play international cricket and handle 90 mph bowling, with one eye and went to become a celebrated captain.

Fury approves

Sponsored

 

[Deleted member 17810]

skrulls dont count




spoiler alert

 

[cardad]

Tesla spent $2m in lidar. Stock drops $10b in a day...

$2m is nothing. Prob equivalent to 1 prototype cybertruck wheel or one AI engineer's salary...

Interesting article saying lidar is $1000 so cheaper than I thought.

Radar prob even cheaper $10. I mean I spent $110k on a truck, $1000 for lidar is better than $2000 for the tent.

It's all about the code to make it useful. They already have radar on older models which is free but basically stopped using it because it cost money to pay coders and train the AI engine vs use free image depth algorithms that for some crazy reason people "open source" for free while paying their software engineers $10m a year: https://www.theinformation.com/arti...-google-promising-recruits-a-quick-stock-bump

Guess Lidar also fugly and reduces range as not aero at all but again it's probably all to do with what code you can get to work and AI camera vision / video is actively developed.

I guess you haven’t seen how Lotus (aka CHINA) ?does it.

 

[CTSoFL]

*Disclaimer: I work for Google, but not Waymo
*Overall opinion: I use the Cybertruck's FSD all the time, even to drive 2-minutes down the street; it is awesome because it's here NOW

After seeing many FSD threads this past week and enjoying the Cybertruck's FSD myself I wanted to share a brief crash course on autonomous driving

References:

• MIT's course on Artificial Intelligence: Implications for Business Strategy
• Lecture, "Understanding Robotics"
• Lecture transcripts attached for convenience

Robotics 101: AI in the Physical World, with Sensors and Actuators
At a high level robots are just automated devices that perform physical tasks in the real world using 1. sensors (inputs) and 2. actuators (outputs)

Tesla is Level 3 Autonomy, Waymo is Level 4

All Self Driving Cars "Map" Between Feature Space (virtual) and Physical Space (real)

• All of the sensors (cameras, LIDAR, radar, microphones, etc.) are used to create a map of our real world in a virtual world called "feature space"
• The car "thinks" and "acts" entirely in this feature space
• It needs the sensors to keep populating the feature space with objects, to try to mimic our physical world as much as possible
• This why the Cybertruck "drives over curbs" while Waymo "sees the curbs", Waymo's feature space is better populated, closer to our real world

Tesla FSD is Parallel Autonomy (human-in-the-loop), Waymo is Series Autonomy (no human, mostly)
In robotics there are two types of autonomy, series and parallel:

• Series autonomy (Waymo): either human in control or vehicle in control, not both
• Parallel autonomy (Tesla FSD): aka "Supervised" FSD

Tesla's Big Failure: No LIDAR
In one sentence, robots need "laser imaging, detection and ranging" (LIDAR) to accurately perceive depth, occlusion, etc.

Longer answer is that LIDAR facilitates building a 3D point map of the world. Initially cameras were meant only for object detection. Tesla admirably expanded camera-based computer vision into FSD, but the lack of LIDAR is still problematic

Tesla's Reason for Camera-Only Computer Vision FSD: Cost
LIDAR (and the now removed radar from early model Teslas) is expensive and it's not just the hardware. The Cybertruck FSD update boasts that its neural network / machine learning model replaces over "300K lines of explicit C++ code". That's a lot of software engineering hours and millions of dollars saved

My (unverified) opinion is that there are nowhere near as many LIDAR 3D map examples as there are computer vision image and video examples. This means training a LIDAR machine learning model is not (yet) feasible

Waymo cleverly overcame this limitation by transforming the LIDAR 3D point map into a more manageable format. I have not had the time to read all the research, but Waymo solved the LIDAR point map problem, where a mess of 3D dots inhibited using machine learning fully

All of Waymo's research came at an extreme cost and they had Google / Alphabet money. Tesla just couldn't afford to run at a loss and had no other massively profitable business units to draw from

Waymo is nowhere close to Level IV because it can only perform in a closed environment.

Drop a Waymo car outside its controlled confines and it becomes a paper weight.

Tesla FSD can operate in far broader environs.

Not even a close comparison.

 

[Crissa]

They already have radar on older models which is free but basically stopped using it because it cost money to pay coders and train the AI engine vs use free image depth algorithms that for some crazy reason people "open source" for free while paying their software engineers $10m a year: https://www.theinformation.com/arti...-google-promising-recruits-a-quick-stock-bump

Tesla and Elon believe in the open source (at least for industrial products they use rather than sell) which is a good policy. No one wants to be stuck rebuilding from being wedded to a single company's tool or engine when it collapses, goes stale, or tries to strangle you for profit.

The point being, the radar/lidar repeats data they already have. It doesn't increase the accuracy of their current decision trees. The problem with the current software is based upon those decision trees, not the data it's collecting.

In nearly all the cases of collisions in the full stack FSD could see the objects. It needs to be able to identify the problems.

-Crissa

 

[cyberos]

I guess you haven’t seen how Lotus (aka CHINA) ?does it.

At this point China is a clear decade ahead of other countries in EV batteries, vehicle robotics and humanoid robotics. As controversial as Elon is, he seems to be the only futurist able to stand against the Chinese engineering juggernaut

Good for China, and it is sadly reflecting the vacuum of policy leadership in our part of the world

 

[cybertruckvegas]

What software contributions do you see Tesla put out open source? I see none useful being in that space.

By far Meta is #1 contributor with LLM and Vision. Next Microsoft eg Florence2 is pretty great.

Apple uses other stuff but nothing groundbreaking I've seen besides refactor other code.

Tesla nothing so don't give them that much credit.

Tesla and Elon believe in the open source (at least for industrial products they use rather than sell) which is a good policy. No one wants to be stuck rebuilding from being wedded to a single company's tool or engine when it collapses, goes stale, or tries to strangle you for profit.

The point being, the radar/lidar repeats data they already have. It doesn't increase the accuracy of their current decision trees. The problem with the current software is based upon those decision trees, not the data it's collecting.

In nearly all the cases of collisions in the full stack FSD could see the objects. It needs to be able to identify the problems.

-Crissa

 

[DrPhyzx]

humans don't have lidar and drive perfectly well a camera based system should be able to do the same or even better since it has 360 field of view. .

ML doesn't work like human intelligence: Decades of attempts to approach AI by mimicking the way the human brain works have all been abysmal failures. So... why should useable machine vision and spatial perception work like human vision and spatial perception? There is no reason to think it won't fail just as miserably.

 

[Crissa]

What software contributions do you see Tesla put out open source? I see none useful being in that space.

I don't recall saying they did.

ML doesn't work like human intelligence: Decades of attempts to approach AI by mimicking the way the human brain works have all been abysmal failures. So... why should useable machine vision and spatial perception work like human vision and spatial perception? There is no reason to think it won't fail just as miserably.

That's just not based upon facts.

Machine learning literally does work, and no one has been able to mimic the way the human brain works because we can't currently build a human brain using silicon.

But just because we couldn't do something in the past is no evidence we can't do it in the future. We're typing on devices which were basically impossible to even conceive a hundred years ago (although a few people did).

-Crissa

 

[jerhenderson]

Camera vision isn't getting within inches.

?? I know the width of my vehicle in a lane, using my vision.

 

[Sjohnson20]

I just used Waymo 3 times in Los Angeles today. I really like it!

A couple funny things happened though. One was a lady got really mad at the Waymo because it didn't turn right on red fast enough and was honking at it lol! Then she got next to the car and was making weird hand signals at the car. Man people are stupid.

The even crazier one was a guy walked out in the street and the car stopped and then he tried to tell it to go lol. Then he got mad at the waymo and flipped it off. Then he flipped me off in the back seat and called me an ahole hahaha

Interesting enough it felt like the early days of driving the Cybertruck. People taking pics and videos and wanting me to roll down the window. And people getting mad and flipping it off haha

 

[HaulingAss]

Yes, lidar gives accurate, mathematical, distance data, that fsd optical system has proven to not been able to deduct.

Tesla vision can give distance data quicker and more accurately than any human can guess. We don't drive by calculations, we drive by instinct. Autonomous vehicles will be the same way.

 

[Crissa]

A couple funny things happened though. One was a lady got really mad at the Waymo because it didn't turn right on red fast enough and was honking at it lol! Then she got next to the car and was making weird hand signals at the car. Man people are stupid.

The even crazier one was a guy walked out in the street and the car stopped and then he tried to tell it to go lol. Then he got mad at the waymo and flipped it off. Then he flipped me off in the back seat and called me an ahole hahaha

I feel for those Waymo. Like, when I'm driving, people not taking their right of way just delays me! Don't stand in the way, waving me on. Now I have to re-check all around you.

Or being honked at for not tailgating or driving over pedestrians. That one is terrif.

-Crissa

 

[Deleted member 17810]

?? I know the width of my vehicle in a lane, using my vision.

roughly *

Tesla vision can give distance data quicker and more accurately than any human can guess. We don't drive by calculations, we drive by instinct. Autonomous vehicles will be the same way.

Tesla vision gives incorrect distance data right now. You've posted in the threads where park assist has messed up. People drive by instinct. The point of autonomous driving is that we drive by data, not a close second.

Maybe after tomorrow.

I personally doubt it.

 

[Deleted member 17810]

At this point China is a clear decade ahead of other countries in EV batteries, vehicle robotics and humanoid robotics. As controversial as Elon is, he seems to be the only futurist able to stand against the Chinese engineering juggernaut

Good for China, and it is sadly reflecting the vacuum of policy leadership in our part of the world

There's a reason that musk is playing friendly with china for manufacturing. He's not standing against them, it's opportunistic.

Sponsored