Characters, Data, Stable Power May Give China the Edge Over US in AI Stakes

China and the United States are unquestioned top dogs in the global artificial intelligence (AI) arena, with the scrap between them growing ever-fiercer by the day. Seemingly innocuous forces are, however, quietly tipping the scales. The idiosyncrasies of the Chinese language and China’s edge in data and unlimited, stable power resources are slowly but steadily forging the sword for the country’s eventual triumph in the AI melee.?

These two elements - the one an heirloom of cultural wisdom, with the other two forming a fertile bed to spur the growth of AI - are undoubtedly mighty weapons for China to wield to win the future AI contest.

China’s data hoard: unearthing cultural gems

China has the world’s largest cohort of internet users, and the data traffic they generate each day is a river in full spate, much like the vast streams whose hydropower makes up almost one-fifth (and counting) of total national energy generation. Together these rich sources - endless data and boundless power - present a veritable cornucopia to nourish AI development. The richness and complexity of Chinese information provide a huge space for AI learning unrivaled in its length and breadth.?

Chinese, an antediluvian language with profound cultural connotations and a unique evolutionary trajectory, also holds signal advantages in natural language processing:

“Text normalization is a method for standardizing text to prepare it for the tokenization, vectorization and classification steps. With [English], the?first step?would be to convert all text to lowercase. Because Chinese characters are?not capitalized?to begin with, there’s no need for that data cleaning step. Next comes?stemming?or?lemmatization.Compared to English, there is also?no concept of a stem?in Chinese. Therefore, there is?no need to perform this step either!So far, it seems like that preprocessing Chinese text data requires?less steps?than English text data; making the process (surprisingly) a little easier,” data scientist Sidney Kung? wrote on the Medium platform of news portal Towards Data Science.

Its diverse forms of idiomatic expression and complex grammatical structures still make Chinese challenging for AI to understand and generate, and to apply its linguistic logic - just as for the rest of us - but it also hides huge potential. Once AI fully overcomes the problems of crunching Chinese, this will clear the field for China in the realms of intelligent voice assistants, intelligent customer service, machine translation, and other such fields of endeavor.

Noteworthy in this respect is WuDao 2.0, “the world’s largest language model, with 1.75 trillion parameters - besting previous record holder, Google Brain… unveiled at the 2021 Beijing Academy of Artificial Intelligence (BAAI) Conference [and] capable of simulating conversational speech, writing poems and understand[ing] pictures. Google's Switch Transformer, announced in January, featured 1.6 trillion parameters,” AI Business reported at the time. “Wu Dao is closer than any of its peers to reaching?artificial general intelligence (AGI)?and achieving human-level thinking,” AIMultiple stated in January. WuDao 2.0 is also a fully-fledged chatbot, able to hold its own against ChatGPT and all other comers, per the report.?

From ancient times to the present, the Chinese language has always borne the lifeblood of Chinese civilization - even traditional ink painting uses calligraphic brush strokes - and every Chinese character contains a rich cultural connotation since first being engraved onto an animal bone or turtle shell plastron during the Shang dynasty (ca. 1766-1046 BCE) - China’s first - as part of a shamanistic, runic magic ritual. This particularity sets Chinese apart as not only a language, but also a heritage.?

In this era of AI, the huge reserve of China’s data affords it the opportunity to explore the infinite possibilities of AI. By mining its inexhaustible vein of domestic data, Chinese AI systems can not only edge ever closer to human thinking, but also bestow troves of valuable information on scientific researchers to galvanize AI’s growth.

Power supply: hidden technical support

AI development demands powerful computing, and the astronomical energy resources needed to fuel it cannot be overlooked. “With the most abundant hydropower resources in the world, China is leading the world in terms of power generation output, cumulative installed capacity and newly added capacity [and] has led the world in hydropower production since 2004,” China Global Television Network reported in 2022. If its rivers ever do run dry, China can turn to the 13 percent of global coal supplies - the world’s fourth-largest national reserves - according to statistics supplier Worldometer. These China can extract and burn far cheaper than its rivals, thanks to its more lenient regulation.

By providing a stable power supply - China’s largely underground electricity infrastructure makes unplanned outages virtually unheard of - for AI data and supercomputing centers, and through its forward-looking energy planning, the country not only ensures a steady, abundant power feed for the AI industry, but also promotes its swift development via policy initiatives more plodding Western governments find hard to match.

As the basic support for the development of AI, electricity is used not only to drive computing power, but also to support the ecosystem of the entire AI industry. China’s steady development and innovation in the electric power field has provided a strong impetus for the vigorous development of AI. In the future, as AI tech continues to evolve, the importance of energy resources will become increasingly prominent. Kyle Corbitt, co-founder and chief executive of AI startup OpenPipe, wrote in an X (née Twitter) post that he had recently spoken with a Microsoft engineer responsible for the GPT-6 training cluster project, who griped that deploying InfiniBand-level links between GPUs across regions had been a hard slog, market intelligencer TrendForce reported in March.

“Why not just co-locate the cluster in one region?” Corbitt asked. The Microsoft engineer replied, “[W]e tried that first. We can’t put more than 100K H100s in a single [US] state without bringing down the power grid.”

China is under no such disability. The nation’s full utilization of its energy resources will thus bolster its domestic AI industry, fortifying it to rack up ever-greater breakthroughs as other contenders languish in its dust.

Policy guidance: New win-win for industry

The Chinese government has demonstrated an unbending stance on its development strategy for AI. Through comprehensive arrangements in policy support, talent training, infrastructure construction and other means, Beijing has cultivated a thriving ecosystem for the transformation of the two major elements of Chinese language and electricity into AI competitiveness. Happily, at the same time, the country continues to promote structural adjustment and green development of the power industry to ensure the sustainable development of the AI industry, per the Medium report. Policy guidance and industrial chain integration are jointly erecting China’s competitive fortress amid the AI wilderness.

As an important guarantee for the development of AI, policy guidance and support can not only improve dependent industrial chains, but also provide more potentialities for the integration of Chinese and electricity factors into developing AI. The government’s continued attention and support in AI policies have propelled rapid expansion of the domestic AI industry and laid a firm foundation for transnational cooperation. In the future, with the further optimization of the policy environment, China’s competitive advantage in the AI sphere will become ever-more pronounced.

Mixing innovation, development: Comprehensive positioning of AI industry chain

Chinese language, data, and energy advantages give China a unique advantage in the field of AI, but these factors alone clearly do not suffice to stand out in a global competition. China also needs to continue to work hard in algorithm innovation, hardware manufacturing, setting standards, ethics, international cooperation and in other respects to link together a complete AI industry chain. Above all in key areas - e.g., data security and privacy protection - the country must continue to explore and improve to ensure the reasonable development of AI tech. Only on the basis of all-round innovation-driven and coordinated development can China’s AI sector ever hope to truly reach for the stars.

Factors such as algorithms, hardware, ethics, and cooperation are all crucial in this AI epoch. As an emerging AI power, China must anneal its ancillary industry chain and form a complete and comprehensive ecosystem. While formulating global standards and pushing tech innovation, the nation also needs to steadily muster its soft power, integrate with international standards, encourage salubrious advances in AI tech, and hatch more win-win opportunities.

Chinese language, data, and electric power resources, as China’s three major advantages in AI competition, have indeed reserved a space for it on the global stage. However, to have the last laugh in this savage free-for-all, China still needs more comprehensive innovation and development, policy guidance, and support.?

With the relentless advances of sci-tech, and amid intensifying global competition, China must give full rein to its own advantages, while seeking greater cooperation and innovation to hold its lead in AI and the arc of its supporting industry development.