News Elon Musk reveals photos of Dojo D1 Supercomputer cluster — roughly equivalent to 8,000 Nvidia H100 GPUs for AI training

Please correct the article, the last paragraph mentions " 40,000 Nvidia AI4", the AI4 chips (Formelly known as hardware 4) are TESLA chips not NVIDIA.
 
"learning" from "video data gathered from Tesla cars", is that still a thing?
What exactly are you asking? Is Tesla still developing their AI that they're... still developing? Yes, they're still developing it, thus still training it on video data.

Companies are starting to realize video is the only real solution to a universal self driving AI, not LIDAR, as Musk stated 5 years ago. XPeng is abandoning their LIDAR system, GM is haulting their Cruise LIDAR system, multiple other Chinese companies are following suit.
 
GM is haulting their Cruise LIDAR system
I looked this up, and all I see are the opposite results. GM is expanding LIDAR usage, but is not solely relying on it. They are using LIDAR+Radar+cameras.
https://www.theverge.com/2023/3/7/23627656/gm-ultra-cruise-sensor-radar-lidar-hands-free
XPeng is abandoning their LIDAR system
This looks to be true, but is not the full story. They are using Camera + Radar, so it's not solely relying on cameras.

It makes sense to keep the radar. It's a mature, reliable, and low complexity device for collusion avoidance.
 
Slight correction to this article. In the last paragraph it said they're Nvidia AI4. The HW4 chip that is now renamed to AI4 is a Tesla computer, not Nvidia.