Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

For years people acted like self-driving safety was all about perception — what kind of LIDAR, radar and cameras are you using? In the real world, we see that it is all about understanding. The issue here wasn’t whether LIDAR could see something or cameras could see something. It’s...

28,703 Aufrufe • vor 23 Stunden •via X (Twitter)

8 Kommentare

Profilbild von Russ
Russvor 22 Stunden

teslas cant see thin ropes or chains, and dont know when they're on an upper level of a multi-level parking garage

Profilbild von bye
byevor 23 Stunden

Perception is necessary, but understanding is what makes autonomy work. The real breakthrough is teaching the system to understand the scene and make the right decision.🚀

Profilbild von Scott Endsley
Scott Endsleyvor 23 Stunden

Clearly a self driving fail but how does the human driving the SUV not see it, stop or even attempt to avoid it??

Profilbild von A.
A.vor 23 Stunden

They still don’t grasp that intelligence is key, not a hundred sensors.

Profilbild von ĐOGEARMY GENERAL
ĐOGEARMY GENERALvor 21 Stunden

Just a narrative short sellers and morons use that LIDAR is needed. The fucktards clearly don’t realize had this been true, Tesla would have left LiDAR on their cars. It’s not that expensive. 🤣

Profilbild von Dragonbyte🐉🦁🍀
Dragonbyte🐉🦁🍀vor 23 Stunden

Safety gets real when the car understands the scene, not when it only catalogs what it can see. Perception without comprehension is still guessing under a nicer name.

Profilbild von TeslaNorth.com
TeslaNorth.comvor 23 Stunden

two eyes and one of them is usually on the phone. pretty low bar to beat lol

Profilbild von Gerg Oos
Gerg Oosvor 23 Stunden

well it’s clearly both. to understand a traffic situation and reason to solve it and decide what to do, we still must accurately perceive it in the first place. elon was the first to insist that vision is sufficient to achieve that understanding and that sensor-fusion is detrimental not advantageous.

Ähnliche Videos

Tesla’s vision-only approach to solve FSD got laughed at for many years. People called it reckless, said just cameras would never be enough, and even claimed Tesla was cutting corners. Fast forward to today, it’s pretty clear who was right. “It’s so obvious you can solve this with cameras. Why wouldn’t you solve with cameras? It’s 2026. The self-driving problem is not a sensor problem, it’s an AI problem. The cameras have enough information already. It’s a problem of extracting the information, which is an AI problem.” - Ashok Elluswamy, VP of AI at Tesla. Just like us humans, we drive with eyes, we walk with eyes, literally we do everything with our eyes. We don’t have lidar spinning and laser beams on our heads. We simply see the world and our brain figures it out… all from vision. For so many years, the industry treated autonomy like it was a hardware problem. Like “if cameras struggle, just add more sensors.” That’s exactly how companies like Waymo ended up with sensor towers that look like crazy science projects. So ugly btw too… That approach might have made sense in 2008, but not today! Bc back then, AI wasn’t good enough to truly understand the world. So engineers had no choice but to lean on lidar and radar for help. But today, cameras capture insane amounts of information, I’m talking billions of pixels every second. The thing that was missing was AI intelligence. And when AI can understand why a car is slowing or why a pedestrian is hesitating, or real world things, you just need a better brain instead of more sensors. And now, everyone else is realizing this… the future of autonomy was always about understanding the world through vision, just like we humans do. That’s the bet that Elon and Tesla made early on… he was already years ahead of everyone, when everyone was doubling down on LiDAR and radar, even when so many people mocked them for switching early. Tesla vision for the win!!! 🥇

Teslaconomics

19,357 Aufrufe • vor 7 Monaten