On September 30, 2022, North American time (expected on October 1, Beijing time), Tesla's 2022 AI Day event will be held in Palo Alto, California. At that time, Tesla will bring sci-fi into reality. Bot is expected to debut. In addition, the latest progress in Tesla's self-driving technology and Dojo supercomputer may also be announced on the same day.
Looking back at the AI Day event on August 20 last year, the release of Tesla Bot can be described as a big Easter egg. Tesla says it is 1.72 meters tall and weighs 56.6 kilograms, with a screen on the face that displays information, human-level hands, and forceful feedback sensing for balance and agile movements. Just a year later, Tesla will bring this Easter egg to life.
Not long ago, in an article published in the magazine "China Netcom", Musk wrote: "The Tesla robot was originally positioned to replace people in repetitive and dangerous work. But the long-term goal is to make it serve Thousands of households, such as cooking, mowing lawns, caring for the elderly, etc.”
"The height and weight of the Tesla robot is close to that of an adult, it can carry or carry heavy objects, and it can also walk in small steps. The screen on its face is an interactive interface for communicating with people," Musk said. "You may be curious. , why do we design this robot with legs? Because human society is based on the interaction of bipedal humanoids with two arms and ten fingers. So if we want robots to adapt to the environment and do what humans do He would have to have roughly the same size, shape, and abilities as a human being."
Musk said: "The practicality of humanoid robots will increase year by year as production scales up and costs fall. In the future, a home robot may be cheaper than a car. Maybe in less than a decade, People can buy a robot for their parents as a birthday present."
In order to achieve ultra-high computing power for artificial intelligence training, while expanding bandwidth, reducing latency, and saving costs, Tesla built the Dojo supercomputer. A single training module of the Dojo supercomputer consists of 25 Tesla's self-developed neural network training chips, the D1 chips. Since each D1 chip is seamlessly connected, the delay between adjacent chips is extremely low, and the training module achieves bandwidth retention to the greatest extent, with Tesla's own high-bandwidth, low-latency connection of 9 PFLOPs (9 petaflops) of computing power and 36TB/s of I/O bandwidth in less than 1 cubic foot.
Thanks to the independent running ability and unlimited linking ability of the training module, the performance expansion of the Dojo supercomputer composed of it is theoretically unlimited, and it is an uncompromising "performance beast". In practice, Tesla will assemble 120 training modules into ExaPOD, the world's premier artificial intelligence training computer. Compared with other products in the industry, its performance is 4 times higher at the same cost, 1.3 times higher at the same energy consumption, and 5 times less occupied.
Since the release of the Dojo supercomputer last year, the development of Tesla's FSD (Full Self-Driving Capability) has continued to accelerate. At this year's shareholder meeting, Musk revealed that research on the FSD 10.13 version has been going on for some time, and Tesla has made "some very important architectural improvements", such as improving the complexity of left turns. He said that Tesla is currently widely deploying the FSD beta version in North America, and will start large-scale testing of the fully autonomous driving capability (FSD) software this year. The cumulative mileage of the beta software has exceeded 40 million miles, and it is even expected by the end of the year. over 100 million miles.
Musk said at last year's AI Day event: "What we're trying to do at Tesla is make artificial intelligence that people love and that works...and there's no doubt that it's going to be good artificial intelligence." This year, Musk said. What surprises will it bring? Please let us wait and see.
0 Comments