The struggle over who controls the course of key technology, and how much, is as old as the tech sector itself. The ongoing geopolitical scuffle over the control of emerging Artificial Intelligence (AI) technologies increasingly involves both software and hardware which are key to developing advanced capabilities. The next that lines up in the battleground is access to the complex algorithms running AI systems— the Large Language Models (LLM). LLMs are high-level algorithms that are used to train generative AI tools, e.g. ChatGPT. LLMs provide the building blocks to develop applications. LLMs help AI systems understand the way that humans write and speak. OpenAI is part of a wave of LLM startups that includes AI21 Labs, Anthropic, and Cohere.
Worldwide, there is an accelerated focus on investing in next-generation AI modeling technologies. Several big tech corporations have jumped into developing Large Language Models (LLMs). Nvidia, an American Graphic Processing Unit (GPU) maker, has invested huge stocks in five big AI companies (Arm Holdings, Sound Hound AI, Recursion Pharma, Nano-X imaging, and TuSimple) banking on their prospective leadership in the generative AI segment. Similarly, Google Cloud recently committed to invest billions of dollars in Character.ai, Mistral AI, and Anthropic.
The US’s sanctions on Chinese chip companies have further instigated its accelerated drive to develop indigenous capabilities. As a result of hardware bans, China is facing serious challenges in training its AI models. As a result, there is a rush among Chinese venture capitalists to invest in AI modeling technologies aimed at developing independent software and hardware to support its AI development. The Chinese government itself, in the last two years, has approved over forty LLMs and related AI applications. Several other locally developed LLMs are flooding the Chinese market. As Chat-GPT is officially unavailable on the mainland, several start-ups like Moonshot AI and Baichuan are touting themselves as more accurate alternatives to OpenAI.
Moreover, the Chinese tech company is yet again launching a new price war, this time in the AI sector. Recently, Alibaba’s cloud computing unit has slashed its fees up to 97 percent for using its LLMs. On the same day, Baidu made its LLMs free to use. This was also followed by Tencent as it also offered huge discounts for using its LLMs while making one of its models free to use. ByteDance’s recent launch of its own model is also significantly lower than its competitors. The growing anxiety among Chinese AI companies is not only aimed at securing leadership in the huge domestic market but also at showcasing Chinese global leadership and advancement in a field that is critical to the future of AI technologies.
However, there is a doubt whether China’s rush to create its own LLMs makes many prospects to support its indigenization drive. One of the major challenges is its large-scale reliance on the US for the AI technology stack- hardware, software, data, and talent. The success of American LLM projects comes from the amount of quality data, access to the best hardware, huge capital ventures, and lastly vast potential scope for commodification. The Chinese AI industry, which desperately tries to play catch-up looks confused with attempts to hoard US tech hardware. The US’s war on China’s chip sector has restricted the import of high-end GPUs from the US. Reports say that despite the US’s revised chip ban released in October 2023, the American chipmaker’s remodeled chips have been able to reach the Chinese market. However, at the same time, the delivery of Nvidia’s downsized chips has received mixed responses from the Chinese AI companies as their efficiency is said to be reduced to support the training of LLMs. Even if Chinese chipmakers, with state-backed financial resources, develop LLMs akin to their Western counterparts like OpenAI, it will end up compromising its performance edge due to US restrictions on the export of AI hardware.
Another challenge concerning the software segment is also posed by the US’s bid to curtail China’s AI development. Generative AI modeling involves the use of set architectures that handle computational tasks. Various Chinese companies resort to open-source software like RISC-V. However, America’s anticipated move under its “chokepoint strategy” is to terminate the open-source nature of various design architectures, most prominently the RISC-V. Open-source developers make software available free of charge. They also enable programmers to modify and share the underlying source code and create their own apps. Currently, technologies like RISC-V fall out of the scope of US tech sanctions on China. If the US succeeds in reshaping the course of open-source software, it will critically endanger China’s access to key tools supporting AI development.
Third, China’s access to domestic applications to support the development of LLMs also remains short. For this reason, even Baidu’s co-founder Robin Li has called the market launches of LLMs “a waste of resources”. He added that in Chinese AI market, “there were too many LLMs “but too few AI-native applications based on those models,”. The founder of Baichuan also doubts that a price war among the domestic tech companies is hardly enough in helping them to create a competitive edge in the market.
The problem does not singularly lie in the failure to replicate a foreign technology or high-end talent or inability to achieve breakthroughs, but rather a combination of all these factors. China’s realization of the fear of missing out on the ongoing AI revolutionary developments in the US and Europe, especially in the field of generative AI propels its AI anxiety. While the Chinese chip industry seems to navigate the US’s stricter sanctions with an accelerated indigenous development drive, its course in the development of algorithms is likely to go through more murky waters. The geopolitical scuffle over these emerging technologies is expected to face more regulations and restrictions, especially to derail the level of China’s AI growth. Since the AI market is evolving rapidly, it is hard to say whether and how the Chinese AI companies sail through their rising challenges in generative AI development. China’s capability to catch up with the West in the generative AI game is largely dependent on the extent to which the Chinese companies succeed in the self-reliant drive amid the global tech war. Nevertheless, Should the US succeed in curtailing the access of open-source technologies to China, the emerging algorithmic war is likely to intensify with serious implications for China’s future AI growth.
[Image by Tung Nguyen from Pixabay]
The views and opinions expressed in this article are those of the author.
Megha Shrivastava is a final year doctoral candidate and a Dr. TMA Pai Fellow at the Department of Geopolitics and International Relations, Manipal Academy of Higher Education, India.
