China may have access to powerful AI chips because of partnership with Saudi Arabia, claims report

China may be using its close ties with Saudi Arabia in developing AI and ML find its way around US-imposed tech sanctions, and get access to cutting-edge AI chips from NVIDIA, and other chip manufacturers.

Collaboration between Saudi Arabia and China in the field of artificial intelligence has raised concerns at Saudi Arabia’s leading academic institution, which fears that such ties could potentially threaten the university’s access to US-manufactured chips crucial for powering this cutting-edge technology, as per a report by the Financial Times.

Professor Jinchao Xu, an American-Chinese mathematician at King Abdullah University of Science and Technology (Kaust) in Saudi Arabia, has initiated a project called AceGPT, an Arabic-oriented large language model, in partnership with the Chinese University of Hong Kong, Shenzhen (CUHK-SZ), and the Shenzhen Research Institute of Big Data.

This development aligns with Saudi Arabia’s endeavours to take the lead in the regional advancement of AI technology, encompassing the construction of powerful supercomputers and the deployment of Large Language Models (LLMs), which serve as the foundation for generative AI systems, including chatbots. In collaboration with the United Arab Emirates, Saudi Arabia is striving to compete with AI firms and establish tailored models specifically designed for Arabic speakers.

Western officials have expressed long-standing anxieties regarding the increasing technology transfer between their traditional allies in the Gulf and China. The United States has expanded export licensing requirements for graphics processing units produced by companies such as Nvidia and AMD, effectively restricting access to these state-of-the-art chips for Chinese entities involved in building generative AI models. However, the Biden administration has not gone so far as to block exports to West Asia.

Nevertheless, individuals at Kaust are of the opinion that limiting cooperation with China is essential to ensure a secure supply of these chips. As one source familiar with the matter noted, “Many people involved have raised their concerns to leadership about the Chinese relationships jeopardising the supercomputer. They don’t want to upset the US government.”

AI has emerged as a focal point in the competitive landscape of China-US relations within the Gulf, where Riyadh and Abu Dhabi are determined to expand their flourishing trade relations to include technology transfer, while simultaneously preserving their ties with Washington, their primary security partner.

Abu Dhabi has introduced a new version of its Falcon model, which claims to be over twice as powerful as Meta’s Llama 2, previously considered the most advanced “open source” model. Additionally, G42, a company under the control of the UAE’s national security adviser Sheikh Tahnoon bin Zayed al-Nahyan, has collaborated with China on vaccines and laboratory testing and has released an Arabic-language LLM.

US officials are intensifying efforts to entice Gulf states away from China, which includes support for a rail and shipping corridor connecting India and Europe through West Asia.

This latest foray into AI by Gulf nations coincides with Kaust President Tony Chan’s outreach to China, marked by agreements with universities in Shenzhen to exchange research capabilities and talent. Chan co-authored an essay last month in Foreign Policy titled “America can’t stop China’s rise. And it should stop trying.”

Kaust has reported that its exchange with China is thriving, evident in the growing presence of Chinese academics at the university, comprising 20% of students, 34% of postdoctoral researchers, and 9% of faculty members. According to a Kaust staff member, Chinese universities offer more cost-effective interns and students capable of handling routine tasks, such as testing GPUs and sifting through data.

Kaust’s AI initiative, under the leadership of German computer scientist Jurgen Schmidhuber, is working on developing a more potent supercomputer, Shaheen III, with the aim of delivering computing power twenty times greater than its existing system.

Regarding inquiries about the new LLM, Kaust has described the AceGPT project as “an individual research project by one of our professors” that “had not been run on Kaust supercomputer Shaheen II.” This chatbot is trained in Arabic, Chinese, and English.

Kaust has emphasized its global relations and clarified that its investments in GPUs are driven by the needs of its academic community to undertake projects of significance for the Kingdom it serves. The university also asserted that it adheres to US export control regulations and maintains a monitoring framework to ensure compliance with safeguarding regulations, enabling it to operate Shaheen III.

It is important to note that physical and system software access to Shaheen III is restricted to the Kaust Core Labs system administrator and Hewlett Packard Enterprise teams.

from Firstpost Tech Latest News https://ift.tt/F7XzePw