Cybertruck, I can’t unsee this! [ADMIN WARNING: NO POLITICS; MEMBERS BANNED]

[pricedm]

The SAE Definition of Autonomous Driving Levels does not call out specific technologies, it only details the capabilities of self-driving without human intervention. LiDAR is not literally "required" and Tesla certainly intends to go forward without it.

Autonomy companies like MobilEye are dropping LiDAR due to the inherent limitations, instead focusing on radar as the preferred non-vision sensor.

https://www.autoweek.com/news/a62154234/mobileye-lidar-development/

Very interesting:
"Autonomous tech developer Mobileye is now moving away from lidar development entirely. And the reasons are perhaps logical at this stage in autonomous sensor evolution.

Mobileye says lidar is now "less essential" to its path toward Level 3 eyes-off systems, citing increased performance of imaging radar, its EyeQ6-based computer vision perception, and continued cost reductions in off-the-shelf lidar systems."

Sponsored

 

[dalton108]

I'm not worthy ?

True, but you’re our a-hole!

?????

glad to have you back, Buddy!!

 

[Chris9702L]

Very interesting:
"Autonomous tech developer Mobileye is now moving away from lidar development entirely. And the reasons are perhaps logical at this stage in autonomous sensor evolution.

Mobileye says lidar is now "less essential" to its path toward Level 3 eyes-off systems, citing increased performance of imaging radar, its EyeQ6-based computer vision perception, and continued cost reductions in off-the-shelf lidar systems."

Cybergus must be working for Luminar like Mark Rober is ???????

 

[REM]

I think if we visit the in-context, actual quote of him, we will find that he includled qualifying words along with those statements somewhere along the lines of:

"I think ..."

"If our dev work keeps trending in the right direction"

"Maybe .... Perhaps ... I hope".

It's quite disingenuous to misquote or mis-appropriate the actual language or spirit of what someone is saying. Granted, Elon has been super optimistic and he, himself, has fully owned up to that several times.

Can you show us demonstrations of real world Level 3 autonomy by those companies so that we can compare it against FSD? Perhaps we will find that they lack a knee-slapping amount of core competencies. One of them being geo-locked to specific and small areas (serious scalability issue).

I have attempted to provide a core explanation for this before, but I'll try again. Tesla's FSD is not just an Ai, it's an AGI. It is basically designed to be a virtual, truncated, and hyper focused human brain playing a video game. It is emulating special and time domain awareness using near real-time input from the telemetry stream it received from a huge collection of on and off-board sensors (predominantly from the cameras, but it also include accelerometers, temperature, barometric, gyroscopic, GPS, auditory, and a slew of others). This system has to be properly weighted and balanced to mimic how we, as humans, perceive the world around us. Why is that the case? It's because we need to be able to understand how it is working, either in error or in success. The engineers need to be able to lock into the decision making tree as if it was an actual person (and they are doing exactly that). I have a laugh everytime I see someone comment that TSLA's PE is outrageously high. It's not. It's insanely low when you consider their entire portfolio. Not a single entity on this planet has a more advanced Ai team.

Telsa definitely learned from having a forward facing radar, that there is no significant progress to be made simply by adding more (volume) or varying (lidar, Ultra-sonic) telemetry to the stack. I think what's going on here is that you may lack a fundamental understanding of how the Ai is processing this data, and how quickly it needs to do it in order to make actual lifesaving and even routine decisions. It is indeed a mimicry of how our brains actually gather and process data. Tesla's Ai day a few years back really nerded out on this topic, so I'd suggest starting there. Remember when phantom braking was a thing? Yeah, that was the radar giving good data, but bad context. The cameras give good data AND good context. Lidar gives inconsistent, grainy, data, with wildly varying degrees of context. Was that a reflection from someone rolling down a car window next to you, or is there a pothole up ahead. ?‍

Lidar is great when it is applied appropriately. It provides an awesome amount of 3d spacial awareness that is harder (but not impossible) to replicate with cameras. I think one mistake people make when comparing the two is how the NN processes occlusion. By the time the cameras render a few frames and process it into the 3D spacial stack, the Lidar is still trying to get a full rotations worth of data. And even then, when it's presented and processed to the computer in a manner that makes sense, the data is already late. In other words, cameras are faster. It's a bit of a paradox, right? Both are based on the physics of light.

Us humans don't have lidar. It's a foreign sensor to us. The closest thing we have is the hair on your body sensing pressure wave differentials in the atmosphere. Not a single company on this planet has figured out how translate this type of data into something useful for an AI NN stack.

 

[Chris9702L]

I think if we visit the in-context, actual quote of him, we will find that he includled qualifying words along with those statements somewhere along the lines of:

"I think ..."

"If our dev work keeps trending in the right direction"

"Maybe .... Perhaps ... I hope".

It's quite disingenuous to misquote or mis-appropriate the actual language or spirit of what someone is saying. Granted, Elon has been super optimistic and he, himself, has fully owned up to that several times.


Can you show us demonstrations of real world Level 3 autonomy by those companies so that we can compare it against FSD? Perhaps we will find that they lack a knee-slapping amount of core competencies. One of them being geo-locked to specific and small areas (serious scalability issue).



I have attempted to provide a core explanation for this before, but I'll try again. Tesla's FSD is not just an Ai, it's an AGI. It is basically designed to be a virtual, truncated, and hyper focused human brain playing a video game. It is emulating and special and time domain awareness from near real-time input from the data it received from a huge collection of sensors (predominantly from the cameras, but it also include accelerometers, temperature, barometric, gyroscopic, GPS, auditory, and a slew of others). This system has to be properly weighted and balanced to mimic how we, as humans, perceive the world around us. Why is that the case? It's because we need to be able to understand how it is working, either in error or in success.

Telsa definitely learned from having a forward facing radar, that there is no significant progress to be made simply by adding more (volume) or varying (lidar, Ultra-sonic) telemetry to the stack. I think what's going on here is that you may lack a fundamental understanding of how the Ai is processing this data, and how quickly it needs to do it in order to make actual lifesaving and even routine decisions. It is indeed a mimicry of how our brains actually gather and process data. Tesla's Ai day a few years back really nerded out on this topic, so I'd suggest starting there.

Lidar is great. It provides an awesome amount of 3d spacial awareness that is harder (but not impossible) to replicate with cameras. I think one mistake people make when comparing the two is how the NN processes occlusion. By the time the cameras render a few frames and process it into the 3D spacial stack, the Lidar is still trying to get a full rotations worth of data. And even then, when it's presented and processed to the computer in a manner that makes sense, the data is already late. In other words, cameras are faster. It's a bit of a paradox, right? Both are based on the physics of light.

Us humans don't have lidar. It's a foreign sensor to us. The closest thing we have is the hair on your body sensing pressure wave differentials in the atmosphere. Not a single company on this planet has figured out how translate this type of data into something useful for an AI NN stack.

Thank you ?????

 

[Sjohnson20]

Tesla is trying self driving their way and other companies are also trying in different ways. Nobody is at the finish line yet. I don't know if Tesla will succeed but they are giving it a damn good effort.

 

[CyberGus]

I'm certain that Will Rogers never met you.

 

[BannedByTMC]

I think if we visit the in-context, actual quote of him, we will find that he includled qualifying words along with those statements somewhere along the lines of:

"I think ..."

"If our dev work keeps trending in the right direction"

"Maybe .... Perhaps ... I hope".

It's quite disingenuous to misquote or mis-appropriate the actual language or spirit of what someone is saying.

Indeed

We'll be able to do a demonstration guide of full autonomy all the way from LA to New York...from home in LA to Times Square in New York. And then have the car go and park itself,.

Read More: https://www.jalopnik.com/elon-musk-tesla-self-driving-cars-anniversary-autopilot-1850432357/

"This year...the car will be able to find you in a parking lot, pick you up, take you all the way to your destination without an intervention," Musk said in 2019. "I would say that I am certain of that. That is not a question mark."

Read More: https://www.jalopnik.com/elon-musk-tesla-self-driving-cars-anniversary-autopilot-1850432357/

Sponsored