news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

【Text/Observer Network Qi Qian】On July 23rd, local time, Facebook's parent company Meta released a new artificial intelligence (AI) open source model "Llama 3.1". In an interview report released by Bloomberg on the same day, Meta CEO Zuckerberg refuted the claim that the AI ​​open source model would be used by Chinese companies. He said that the idea that the United States is several years ahead of China in AI is "unrealistic" and that technological blockades will hinder technological development.

In the interview, Zuckerberg said that Llama 3.1 is the "most advanced" open source AI model launched by the company so far, which can be used withOpenAIIt's comparable to similar models from rivals like Google.

Zuckerberg mentioned that Meta's AI investment is huge, and the Llama 3 model has cost "hundreds of millions of dollars" in training computing power so far, and the cost is expected to be higher in the future. He said that "it will be billions of dollars in the future", but Meta is still willing to invest in the AI ​​"arms race".

“I think a lot of companies right now are overbuilding, and when you look back you think, ‘Oh, we probably all spent a few billion more than we should have,’ ” Zuckerberg said. But “on the other hand, actually, I think all the companies that are investing are making rational decisions. Because if you fall behind, you won’t be able to master the most important technologies in the next 10 to 15 years.”

According to Zuckerberg, after all investments are completed, Meta will open the technology behind Llama to the public for free, as long as users comply with the "Acceptable Use Policy". He hopes that through the open source strategy, Meta will become the foundation for other successful startups and products and have a greater influence in the development of the industry. However, Meta will still keep the data set used to train Llama 3.1 confidential.

Bloomberg said that some critics believe that Meta's open source AI model may be abused, and they are also worried that technology companies from the United States' "geopolitical rivals" such as China will use its technology to keep up with their American counterparts.

Zuckerberg is more concerned, though, that isolating U.S. AI technology from the rest of the world could ultimately backfire.

“There’s this idea that, ‘Well, we need to lock everything down,’ ” Zuckerberg said. “I think that’s wrong because the United States thrives on open and decentralized innovation… So I think locking everything down is going to hold us back from continuing to make progress and ultimately make us more likely to lose our leadership position.”

He added that the idea that the United States would be years ahead of China in AI was “also unrealistic” and that the United States would gain a clear technological advantage through the “accumulation” of smaller leads.

sinceChatGPTSince it became popular, there have been calls for it to be open source. However, after the release of ChatGPT2, OpenAI chose closed source development. Alibaba Cloud, Zhipu, Tsinghua EKG, Baichuan Intelligence, etc. chose open source, while Huawei chose closed source for data privacy and commercial benefits. In this global big model arms race, the debate on open source development and closed source development continues.

Zuckerberg expressed the same view in an open letter released on the same day (23rd).

Zuckerberg pointed out that open source will ensure that more people around the world can enjoy the benefits and opportunities brought by AI, power will not be concentrated in the hands of a few companies, and the technology can be more evenly and safely applied to the whole society. He believes that the best strategy is to build a strong open ecosystem, so that industry leading companies work closely with governments and allies to ensure sustainable first-mover advantages in the long run.

Regarding China, he mentioned in his open letter that "some people believe that the United States must adopt closed source to prevent China from obtaining these models," but "this will not work and will only put the United States and its allies at a disadvantage."

In recent years, Chinese universities and enterprises have developed rapidly in the field of AI. In early May this year, the Center for Security and Emerging Technologies (CSET) at Georgetown University in the United States released research results showing that in more than half of the hottest AI fields in the world today, "China's research is ahead of the United States." The data shows that Chinese institutions are among the top in terms of both the total number of AI research papers and the number of highly cited AI research papers.

"China is absolutely a world leader in artificial intelligence research and is likely to be a world leader in many areas," said Zachary Arnold, director of analysis for the CSET team, adding that China is now active in a range of AI research areas, including an increasing amount of basic research.

On July 3, the United Nations World Intellectual Property Organization (WIPO) stated that China has made great progress in the fields of chat robots, etc.Generative AIThe United States is far ahead of the rest of the world in terms of (AI) inventions, having applied for more than 38,000 generative AI patents in the past decade, six times as many as the United States (6,276).

Christopher Harrison, patent analysis manager at the World Intellectual Property Organization, said: "This (generative AI) is a booming field and an increasingly fast-growing field... The data shows that this field will have a profound impact on many different industrial sectors in the future." He pointed out that the scope of Chinese patent applications is very wide, coveringAutopilot, publishing, to document management.

This article is an exclusive article of Observer.com and may not be reproduced without authorization.