Reference News Network, November 28 report: The U.S. "Fortune" magazine website published an article titled "The Momentum of China's Open-Source AI Models is Obvious Beyond the US and Europe" on November 25. The author is Jeremy Kahn, the editor of the AI section at the magazine. The following is an excerpt from the article:

Last week, I attended the Fortune Innovation Forum held in Kuala Lumpur, Malaysia, and hosted several panel discussions on artificial intelligence (AI) and its impact. One of the insights from this trip to Kuala Lumpur was that I became more aware of how companies outside the European and American regions are eager to develop based on open-source AI models and prefer Chinese open-source AI models.

Open-source models have multiple advantages

Several weeks ago, my colleague Beata Nolan briefly mentioned this phenomenon in an article published in this magazine, but after visiting Southeast Asia, I truly understood it: Although the United States has the most advanced AI models, it may likely lose the AI race. As Chen Yibang (音), Director of Southeast Asia and India at Singha Venture Capital Holdings, said during a panel discussion, the root cause lies in the fact that U.S. AI companies "pursue perfection," while Chinese AI companies "pursue accessibility."

We sometimes hear U.S. executives (such as Airbnb CEO Brian Chesky) say that they favor Chinese open-source AI models because these models perform well and are very affordable. However, at least for now, this attitude remains rare. Many of the European and American executives I have spoken to expressed a preference for the performance advantages of proprietary models from companies such as OpenAI, Anthropic, or Google.

This view is completely different from what I encountered among executives in Asia. Here, people focus more on controlling data and costs. In terms of these indicators, open-source models often perform better. Yuan Jinhu, co-founder and CEO of Silicon Base Flow, a leading AI cloud hosting service provider in China, stated that the company has developed various technologies that can run open-source models more economically and efficiently, meaning the cost of completing tasks using open-source models is much lower than using proprietary AI models. In addition, he stated that most customers found that by fine-tuning open-source models with their own data for specific application scenarios, the performance could exceed that of proprietary models and completely avoid the risk of leaking sensitive data or trade secrets.

Chen Yibang of Singha Venture Capital also emphasized this point. He reminded that although proprietary model providers also offer customized services based on a company's own data, they usually promise not to use these data for broader training, but "you never know what's happening behind the scenes."

Using proprietary models also means giving up control over key costs. He advised startups who consulted him: If they are developing an application that is crucial to their competitive advantage or core product, they should build it based on open-source technology. He said, "If startups are developing AI-native applications and selling them as core services, they must have full control over the technology stack, and the best way to achieve this control is to use open-source technology."

Cynthia Xianta, CEO of Dyna.AI, an AI application developer in financial services in Singapore, also stated that she believes some Chinese open-source models perform better in local language processing.

But how to evaluate the view that open-source AI is less secure? Cassandra Goh, CEO of fintech company Silverlake Axis, pointed out that models need to be secured within the system. She emphasized that this principle applies regardless of whether the underlying model is proprietary or open-source.

"Bridge Powers" Draw Attention

This conversation made me deeply realize that OpenAI and Anthropic, which are accelerating their global expansion, may face resistance, especially in middle-income countries in Southeast Asia, the Middle East, North Africa, and Latin America. This further confirms that the U.S. urgently needs to build a stronger open-source AI ecosystem beyond Meta. So far, Meta remains the only major player in the U.S. in the field of open-source frontier models. Although IBM has some open-source foundational models, their performance is far behind the leading models of OpenAI and Anthropic.

The insights from this trip to Asia go beyond this. The blueprint for building AI infrastructure in the region is also very attractive. Especially, Johor State in Malaysia is striving to become a data center hub, not only serving Singapore but also radiating throughout Southeast Asia. Talks to share data center capacity with Indonesia have already started.

Johor State's planned data center projects over the next few years will consume almost all of the state's current power generation capacity. The state and the entire country of Malaysia plan to significantly increase power generation by 2030 through gas power plants and large solar power plants.

In the increasingly intense geopolitical competition between China and the U.S. over AI technology, how regional participants choose has become a hot topic. Many countries seem eager to find a path that allows them to utilize the technologies of both superpowers without taking sides or becoming vassals. However, whether they can walk this tightrope remains a major unresolved issue.

This week, earlier this week, 30 policy experts from the Mirai-Quebec Artificial Intelligence Institute (founded by AI "father" and Turing Award winner Yoshua Bengio), the Oxford-Martin AI Governance Program, and institutions in Europe, East Asia, and South Asia jointly published a white paper, calling for middle-income countries to jointly develop and share AI capabilities and models, thereby reducing their dependence on Chinese and U.S. AI technologies to some extent.

However, whether this non-aligned movement in the AI field can be realized in diplomatic and commercial terms is still highly uncertain. But one thing is certain: officials in these bridge countries will certainly consider this idea carefully. (Translated by Liu Baiyun)

Original: https://www.toutiao.com/article/7577636748772131355/

Statement: The article represents the views of the author. Please express your opinion by clicking the buttons below [upvote/downvote].