The rationale why China AI Unicorn gave up on model development and switched to ML Ops

-

(Photo = X, Lee Kai-fu)

Ikaipu, a famous AI expert and founding father of 01.AI, has revealed the rationale why he stopped developing its own large language model (LLM). It also argued that Open AI recently asked the US government to ban using Chinese models.

In an interview with the South China Morning Post, the founding father of Ikaipu 01.AI told the twenty second that it’s specializing in providing solutions that tremendous -tune open source models resembling deep chic slightly than model development.

He’s a pc scientist from the University of Carnegie Melon and is a Google China president and is one in all AI experts who’ve gained international popularity. In 2023, Time was also included within the 100 most influential AI figures.

To start with, due to popularity of open source models resembling Deep Chic, developing a pre -training model that costs high in startups has grow to be an inefficient way.

“The way to recuperate the GPU costs, how much business profit growth, and whether there’s a possibility of achieving profits,” he said. The six tigers are China’s leading AI startups, including 01.AI, no tranfu AI, Moonshot AI, Minimax, Stepfun, and Biadan.

“The pre -training model is meaningful provided that we collect lots of of hundreds of thousands of users,” he said.

That is an amazing popularity of 01.AI, which he founded in late 2023 as an open source model called ‘YI’, but then switched to ML Ops as a substitute of model development.

He has modified his company to an answer provider since last yr, and is currently in control of optimizing the corporate consistent with the deep chic boom in China. Now, only 12 Chinese corporations have growth and similar capabilities as ML Ops in China’s top class.

Consequently, the corporate expects to attain sales of 100 million yuan (about 20.2 billion won) in the primary quarter, which is such as the overall revenue last yr. Currently, there are about 200 employees in 01.AI, so it has not yet reached the break -even point.

(Photo = 01.AI)
(Photo = 01.AI)

It is extremely rare that startups, which were unicorn as a frontier model after eight months of founding, gave up their model development. Last yr, the price of LLM development soared to the sky, and startups resembling Cohery said they might give attention to the event of small language models (SLM), and the influence AI, which handed over major technicians to Microsoft, focused on ML Ops.

Nonetheless, the Chinese model is difficult to enter the worldwide market, and it’s difficult to compete if greater than 300 LLMs are poured out in Korea.

The founder identified that this is just not only a matter of China. “I saw loads of individuals who canceled the chat GPT subscription because Deep Chic is free,” he said. “The most important nightmare for Sam Altman is that the competitor is free.”

As well as, the Open AI asked the Trump administration to ban using Chinese models resembling Deep Chic, which was called “national control” and “the start of paranoia.”

“Because someone built the identical house at no cost, they’re watching that the home made with the cards they arrange is beginning to collapse.”

Meanwhile, the founder appeared on Bloomberg TV last week, and the AI ​​leader was compressed into 4 places: Open AI, Google, Antroic, and XAI, and only three Deep Chic, Alibaba, and Bite Dance in China would survive.

As well as, it has been identified that China is meant to scale back the profits of open AI to pour out the open source model, and the founder also identified that Open AI will likely be hit by deep chic.

By Dae -jun Lim, reporter ydj@aitimes.com

ASK ANA

What are your thoughts on this topic?
Let us know in the comments below.

0 0 votes
Article Rating
guest
0 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments

Share this article

Recent posts

0
Would love your thoughts, please comment.x
()
x