|
But Wang Xiaochuan also has common problems left by typical entrepreneurs in the Internet era. Technological entrepreneurship is different from Internet entrepreneurship. After , most Internet entrepreneurship projects were about model innovation and seizing the market, while technology entrepreneurship is often on the sidelines, with little regard to market factors but to firm up its own technical direction. , observing its product route for one year, we will find that Baichuan, which originally took the open source route,
also began to launch closed source models for commercialization in the later Malaysia Phone Number Data period, and began to quickly enter the commercial deployment stage. "Baichuan was a bit carried away and did not have a main line to others. Whether it is large parameters or long texts, they are all directions that have been verified by the market. He wants to do both large models and AI applications in B and C, but there are no outstanding features. ." Feedback from a large model entrepreneur. In contrast, starting from , Zhipu has firmly chosen to become a loyal believer in OpenAI. Zhipu
was founded in , and began to discover and choose to follow the OpenAI route in But even now, All in large-scale parameter models are an extremely risky thing, especially in , when AI is bleak. However, in terms of technical path, Zhipu did not choose to blindly follow Google's BERT architecture or the GPT architecture selected by OpenAI. Instead, it proposed a GLM algorithm framework based on the two routes, which can complete the continuation and writing while understanding. fill in the blank. Speaking of which, Zhipu is a company formed by "Tsinghua Doctors" with roots in Miaohong.
|
|