Загрузка видео...
Не удалось загрузить видео
In 2022, Elon Musk predicted all human code could be replaced by AI in Tesla's self-driving software. This approach materialized two years later with FSD v12. “The thing that's happened over time with respect to Tesla Autopilot, is that the neural nets have gradually absorbed more and more software.... show more
6,068,111 просмотров • 1 год назад •via X (Twitter)
Комментарии: 10

Nvidia CEO Jensen Huang on Tesla's newest self-driving technology: “Tesla is far ahead in self-driving cars” "It's an end-to-end generative model, and it learns from watching video... and it learns to about how to drive... and using generative AI, predict the path and how to understand and how to steer the car. And so the technology is really revolutionary."

So AI can fully code and train the self driving cars?

yes

🇺🇸🇺🇸 Team America 🇺🇸🇺🇸

Amazing Elon Musk!❤️ Abandoning coding and adopting video training of neural networks is a game changer for autonomy! Now both Tesla and Optimus have beautiful brains and the future looks brilliantly amazing! Leave the LiDAR for deep space, and hop on Robotaxi for a magic race❤️

Elon was looking fit back then. What is up with his new kegger he’s carrying around? He needs to get a trainer. Your body is your temple @elonmusk Don’t let yourself go.

"The Road to FSD v12 In 2022" When Elon Musk first hinted that Tesla’s self-driving software might someday evolve without human-coded instructions, most of the world thought it was ambitious talk—a glimpse into a distant future. But for the engineers in Tesla’s AI division, it was a quiet declaration of a new era. Tesla's approach to Full Self-Driving (FSD) shifted over the next two years, progressing from line-by-line coded instructions to vast neural networks that taught the car how to drive like a human. The idea seemed deceptively simple: feed the car’s onboard AI countless hours of video data, and let it learn to associate visual patterns with the driver’s inputs. The FSD system would analyze how human drivers handled curves, intersections, and unexpected obstacles, and gradually absorb all the rules of the road without a single line of human logic. By 2024, with FSD v12, Musk’s prediction materialized. Engineers marveled as the AI trained itself, almost entirely replacing the need for human code. There were no more painstakingly coded logic trees or hand-written responses for specific scenarios. Instead, the FSD system learned on its own, developing a responsiveness that surprised even the Tesla team. It learned to yield gracefully to jaywalkers, predict aggressive drivers, and navigate icy roads by sensing the subtle shifts in traction—all tasks it learned purely from experience. At Tesla’s headquarters, a large monitor displayed real-time data from the neural network as it drove a test vehicle. Engineers watched in awe as the car analyzed the video feed, dynamically adjusting based on subtle environmental cues. The car braked gently as a ball rolled into the street, anticipating that a child might follow. It waited cautiously at intersections, picking up on subtle patterns of pedestrian movement that no human code could capture. “It's eerie,” one engineer murmured, watching the car's steady progression through a crowded intersection. “It’s not following instructions anymore—it’s just… driving.” Musk, observing from a distance, allowed himself a rare smile. The machine had become its own instructor, teaching itself to drive in ways more nuanced and reliable than any human hand could design. FSD v12 was not just an innovation—it was a testament to the dawn of an era where AI would not merely assist but independently lead humanity into new realms, one self-taught mile at a time.

Elon’s vision coming to life—AI taking over human coding in Tesla's FSD. The future of driving is here!

I wonder how many scenarios can they run in virtual now with all that compute power...

Great, but if Lidar is cheap enough, it's worth having superpowers over visible light
