Nvidia is surprised, and Microsoft is also confused: obviously Huawei's #Pangu model 3.0# is just different, but why is the jackal, tiger and leopard on the AI rivers and lakes trembling at this moment? It turns out that the direction of progress in the AI era has turned a watershed-like bend because of Huawei's unprecedented exploration!
Microsoft once imagined that OpenAI pointed a way to AIGC; Nvidia also imagined that a gas station was also opened on the AIGC road, not afraid that no one would come to drive. Sure enough, as expected by Microsoft and NVIDIA, a number of large manufacturers and startups followed the road with algorithms and PPT, and one after another large models similar to chatGPT came out for a ride... Unexpectedly, the Pangu model 3.0, which appeared during the #2023 World Artificial Intelligence Conference#, insisted on announcing "no poetry" and "no chat", only to say "the main palace does not scratch its hands and make gestures", "the main palace will not be windy and snowy night"!
"Don't write poetry" and "don't chat" are okay, it's a big deal that other AIGC models and Pangu models go their own way. Unexpectedly, just when big manufacturers and startups were thinking about how their respective chatGPT could break the door at the public level, Pangu big model has broken the door silently! Knock on the doors of finance, government affairs, and railways, knock on the doors of manufacturing, coal mines, and ports, but unexpectedly knock on the door of meteorological and research and development like the ceiling for the first time in history!
What does this mean? Poetry painting and codeword programming can be a major direction of AIGC's large model, and if all kinds of chatGPT come to chant for painting and codeword programming, this is actually out of direction.
In fact, Huawei's vigor, NVIDIA will not be aware of it in advance, but NVIDIA thought of the beginning of the story, but did not expect the turn of the story: its AI computing power chip A100 can not come to China, but unexpectedly China's special version of A800 can not come to China. As a result, Huawei's AI computing chip Ascend 910 released four years ago almost equals NVIDIA's current mainstream similar model A100/H100 in terms of performance and power consumption.
Similarly, Huawei's vigor, Microsoft will not be aware of it in advance, but Microsoft also thought of the beginning of the story, but did not expect the back of the story: after being instantly shocked by the Pangu model, Microsoft OpenAI and a number of AIGC large model manufacturers, Ben also wanted to turn a big bend and move forward along the trajectory of Huawei's Pangu large model, but before further thinking was found, this is also very frustrating: a year to the left, no Huawei has taken root and is particularly familiar with finance, government affairs, railways, manufacturing, coal mines, ports, the "legion" of AI application scenarios in meteorological and R&D industries; Looking to the right, there is no Huawei's full-stack self-developed, independent and controllable da Vinci architecture and its Ascend AI computing power chip supporting large models.
You know, AI application industries or application scenarios such as finance, government affairs, railways, manufacturing, coal mines, ports, meteorology, and research and development are either asset-heavy fields or high-value targets, and have never and will never be low-scattered and short-poor industries.
And what does this mean? Pangu large model to the place where the AI application is no man's land. So, when Huawei is running all the way, are the opponents in a hurry or thinking about how to turn?!
