Once synonymous with a copycat culture, China has transformed itself into a prominent contender within the AI domain. Despite geopolitical obstacles, China has actively challenged the US in the AI race. Interestingly, this very culture could prove instrumental in China’s pursuit of becoming an AI leader.
When Google launched the Bert language models introduced in 2018, the following year, Baidu in China responded with Ernie. Interestingly, the character name Bert hails from the renowned children’s show ‘Sesame Street,’ and Baidu also named their model after a character from the same show, reflecting the ongoing AI competition between the two nations.
Currently, the US is once more leading an AI revolution with generative AI. San Francisco-based OpenAI has ushered in a groundbreaking transformation in the AI landscape with the introduction of ChatGPT. But China too has responded with Ernie Bot, a competitor to ChatGPT and Alibaba, the e-commerce giant, has released Tongyi Qianwen, its own generative AI product. Tencent too, is working on a proprietary Large Langauge Model, which reportedly, could be the best to come out of China. In fact, Reuters reported that in the last three years, Chinese entities introduced 79 large-language models (LLMs) within the nation.
AI finesse
China already holds a prominent position in the realm of computer vision, despite initial research started out of the US in the early 60s. China’s heavy focus on this subset of AI is primarily due to the government’s extensive use of state surveillance systems. The Chinese government has identified computer vision as a strategic area and has allocated substantial funding to support advancements in this field. In 2016-2018, Gartner found VC investment in computer vision firms quadrupled to exceed $8 billion. Examining 1400+ deals, they revealed that Chinese investors led the field, contributing over 56% of total investment and dominating eight of the top ten deals.
China is also a leader when it comes to AI research. The country was responsible for approximately one-third of the global output in terms of both AI research papers published and AI citations in 2021. Moreover, over the years, we have seen many products coming out of China, gaining traction worldwide.
Take Titok for example, it was the most downloaded app last year, with around 672 million downloads and became the only second app in the world to cross 3 billion downloads. Tiktok was so successful that it forced other social media platforms such as Meta-owned Facebook, Instagram and Google-owned Youtube to launch similar short-video features.
However, Tiktok’s secret sauce is its algorithm. TikTok’s global popularity is rooted in its AI-powered algorithm, considered a standout in social media. Continuously refined, it drives the platform’s meteoric rise, adapting through user feedback and behavior. Designed for content discovery, it crafts addictive, personalised experiences, enhancing engagement and user satisfaction.
Making strides in generative AI
Moreover, brands such as Baidu, Alibaba and Huawei, have become household names in China, and these are the very companies investing heavily in generative AI. These companies are already shipping their AI products to the world. For example, Baidu’s Apollo project focuses on autonomous driving, and its technologies have been used by companies outside of China, including in partnerships with BMW and Ford. Similarly, Huawei’s AI chips have been used in smartphones and other devices, while its cloud services, including AI capabilities, are utilized by businesses worldwide.
Although GPT-4 currently holds the title of the most advanced AI model, Baidu’s Ernie Bot stands out for its tailored understanding of the complexities of the Chinese language and culture. In fact, reports indicate that Ernie 3.5, the latest iteration of the Ernie AI model, has outperformed the widely acclaimed OpenAI chatbot, ChatGPT, across various pivotal metrics.
Until now, OpenAI, while primarily concentrating on Artificial General Intelligence (AGI), has confirmed their absence of plans for GPT-5 or any models surpassing GPT-4. On the other hand, Google could unveil its work on Gemini, a potential advancement beyond GPT-4’s capabilities. Yet, it remains entirely conceivable for China to unveil the next significant breakthrough in this realm.
A state-driven quest to become an AI superpower
About six years ago, China laid out a development plan to become the world leader in AI by 2030. So far, the Asian superpower has already laid down a solid foundation to support its AI economy. China quest to AI supremacy is intertwined with economic growth. In fact, China is preparing for an AI-powered future. In terms of investments, China is also second to the US, according to a 2022 report by Stanford University. China is also expected to more than double its investment in AI to nearly USD 27 billion by 2026, according to an IDC report
Paul Scharre, the author of the book ‘Four Battlegrounds: Power in the Age of Artificial Intelligence’ believes China’s AI labs are only a year and a half behind the foremost research labs in the Western world. Additionally, the nation holds an advantage in terms of effectively implementing AI across various aspects of society.
China’s significant strides in AI is making many in the US jittery. For example, venture capitalist Vinod Khosla, whose firm has invested in OpenAI, said that the US cannot afford to moderate the rate of progress in AI in the country, as it could lead to China’s advantage. Sam Altman, the CEO of OpenAI, acknowledged the role China could play in solving alignment. He said that China has some of the best AI talent in the world and given the difficulties around solving alignment for advanced AI systems, it would require the best minds from around the world.
The data advantage
Another big advantage China has is data. China has more internet users than the US and Europe combined. With over 1.4 billion Chinese hooked to the internet, the data being generated in China is massive and foundational models thrive on such data. Nearly everyone in China uses WeChat for text messaging, video conferencing, video games, and mobile payments, among other things. Enormous volumes of data are being generated, and unlike the West, China’s data privacy regulations are comparatively less stringent.
This potentially offers Chinese LLM-building labs a notable edge. Additionally, with lenient data privacy laws, Chinese AI labs could also have easier access to valuable public datasets crucial for training LLMs. While data is gold and China has it in abundance, but most of it is in Chinese.
Models such as the GPT series caught the world’s imagination because of its ability to converse in English, and all countries in the world have a section of people that can converse or understands English, however, that is not the case for Chinese. Nonetheless, LLMs launched by Chinese labs such as Erniebot are trained both in English and Chinese datasets.
AI intertwined with US geopolitical strategy
Amidst the imposition of export controls on AI-enabling hardware by the US, China continues to make remarkable advancements in the AI domain. The Joe Biden-led administration aims to curtail NVIDIA’s sales of AI chips to China, a pivotal supplier of these chips that fuel technologies like generative AI. A report by the Centre for the Governance of AI, a British think-tank, found that more than half of the AI labs in China rely on NVIDIA for processing prowess.
Even though China is working hard to become self-reliant in this space, it will take time. For now, China is finding workarounds. In March, The Financial Times revealed that SenseTime, a firm blacklisted by the US, was employing intermediaries to circumvent export controls. Reports also suggest that certain Chinese AI companies are leveraging NVIDIA’s processors via cloud servers located in different countries to navigate the restrictions.
Autocratic government and talent exodus
Another challenge for China is AI talent. Despite having them in abundance, and given some of the top AI scientists in the world are Chinese, yet of most of them are working outside of China. A report by McKinsey revealed that China should face a significant shortage in terms of AI talent by 2030.
Moreover, Time Magazine, in an article, stated that more than half of China’s top AI undergraduate students opt to pursue their graduate studies in the US, and a significant majority choose to remain there after completing their Ph.D. Surprisingly, the primary beneficiary of this influx of Chinese talent is not China itself, but rather the US.
To mitigate this, the Chinese government has launched programmes designed to bring the talent back into the country. Nonetheless, the reason many Chinese prefer working outside of China is because of the autocratic government. “Talent exodus is a major hindrance to China’s authoritarianism in that it drives people away. China’s top AI scientists leave and it’s not just that they go abroad to study and work, they prefer a more democratic way of life,” Scharre said.
China is, in fact, one of the strictest censorship regimes in the world. From political dissent and discussion of sensitive historical events in China to information about its leaders and high-level officials, the government censors almost everything through the Great Firewall of China programme. Given the AI model will be as good as the data it is being trained on, censoring data could prove to be a problem.
The post China’s Copycat Culture Now Dominates AI appeared first on Analytics India Magazine.
