Domestic giants, startups compete in large language model development, keeping up with ChatGPT
A Microsoft employee (right) demonstrates the integration of Microsoft Bing search engine and Edge browser with OpenAI in Redmond, Washington state, the United States, in February. (PHOTO / AP)
Aspecial battle is going on in China as established tech heavyweights and startups scramble to develop their own artificial intelligence large language models (LLMs) to pounce on the opportunities created by AI-generated content.
The move comes as AIGC is expected to become the largest technology transformation since Internet 1.0, which marks the dawn of a new AI era when computer algorithms can produce new text, images, code, video, and audio in a humanlike fashion.
Large language models are computer algorithms fed with huge amounts of textual data for use in a variety of tasks, ranging from natural language processing to machine translation.
Goldman Sachs Research said that breakthroughs in generative AI have the potential to bring about sweeping changes to the global economy.
As tools that use natural language processing work their way into businesses and society, they could drive a 7 percent, or almost $7 trillion, increase in global GDP and lift productivity growth by 1.5 percentage points over 10 years, the research house predicts.
“Despite significant uncertainty around the potential for generative AI, its ability to generate content that is indistinguishable from human-created output and to break down communication barriers between humans and machines reflects a major advancement with potentially large macroeconomic effects,” Goldman Sachs economists Joseph Briggs and Devesh Kodnani wrote in a report.
Eager to turn the rosy prospects into reality, Chinese companies are either stepping up their push in AI or coming up with new initiatives. Local governments are also rolling out favorable policies to speed up AI development.
As of May, China had developed at least 79 AI large language models, or rivals of ChatGPT — an AI chatbot developed by US-based AI research company OpenAI — according to a report released by the Institute of Scientific and Technical Information of China.
As an acronym for Chat Generative Pre-trained Transformer, ChatGPT has wowed the world by showcasing its versatility in writing essays, checking program bugs and making business plans, which differentiates it from previous narrow AI projects that excel in only one area.
Zhao Zhiyun, head of ISTIC, said China’s large language model technology has developed rapidly along various technical routes in recent years, adding that some pre-trained versions with industry influence have emerged in various fields including natural language processing and machine vision.
Chinese tech heavyweights such as Baidu Inc, Alibaba Group Holding Ltd and Tencent Holdings Ltd are all betting big on the technology, and a string of startups are also emerging to launch their LLMs.
China’s general-purpose LLM application is expanding from office and entertainment to fields such as healthcare and industry education, according to ISTIC.
With ChatGPT leading a new wave of global LLM technology development, China is pooling AI innovation resources to carry out research in LLM algorithms and key technologies.
SparkDesk, iFlytek's in-house AI large language model, was unveiled in May. (WANG GANG / FOR CHINA DAILY)
According to a plan released by Beijing in May to promote the innovation and development of AI, the core industry scale of AI in the city is expected to reach 300 billion yuan ($41.4 billion) by 2025.
Shanghai released a series of measures to support the participation of private companies in the construction of AI infrastructure such as data and computing power.
Shenzhen in South China’s Guangdong province has proposed to integrate funding resources in AI with a scale of 100 billion yuan to shore up its intelligent computing power cluster supply, innovation ability of key core technologies and products, and industrial agglomeration.
Jerry Liu, head of China internet research at UBS, said a new computing cycle will emerge every 10 or 15 years, and AIGC will be the beginning of the next cycle.
Liu said many companies in the Asia-Pacific region are behind their counterparts in developed markets in terms of digitalization. However, policy support for digitalization, coupled with the lower entry barrier provided by AI technology, could present significant growth opportunities for businesses in the region.
China has more users of generative AI and more data compared to other economies, so there are many opportunities, he added.
Liu Qingfeng, chairman of iFlytek, a Chinese AI pioneer, said: “Chinese companies must devote big, real money to research and development in fundamental science to achieve breakthroughs in AI. The impact of this generative AI technology is no less important than that of the birth of the PC or the internet. We need to do our best to learn from ChatGPT and even seek to surpass it.”
The company unveiled its in-house AI LLM SparkDesk at a launch event in May, which iFlytek said aims to exceed ChatGPT’s capabilities in understanding Chinese and reach its standards in English.
Liu from iFlytek held a live demonstration of its SparkDesk chatbot in both consumer-oriented and enterprise-focused applications in May. He used voice-recognition prompts to ask SparkDesk in Chinese and English to assess student essays and write hypothetical stories about Confucius’ attendance at the 2008 Summer Olympics in Beijing.
Shanghai-based AI company SenseTime also unveiled its self-developed LLM SenseNova in April, introducing a variety of foundation models and capabilities in natural language processing, content generation, automated data annotation, and custom model training. It also demonstrated a series of generative AI models and applications, such as text-to-image creation and 2D/3D digital human generation.
Xu Li, chairman and CEO of SenseTime, said: “In the era of artificial general intelligence, the three elements of data, algorithms and computing power are undergoing a new evolution. The number of model parameters will increase exponentially, and the volume of data will grow massively with the introduction of multimodalities, leading to a continuous surge in demand for computing power.
“We have built the infrastructure for the AGI (artificial general intelligence) era with SenseCore and named our foundation model set as ‘SenseNova’, implying ‘constant renewal, daily renewal, and further renewal’. We hope to continuously update the models’ iteration speed and their problem-solving capabilities, unlocking more possibilities for (AGI).”
Market consultancy Gartner predicts that by 2025, generative AI will account for 10 percent of all data created, compared with less than 1 percent last year, and it could be used for activities such as creating software code, facilitating drug development and targeted marketing.
The AI boom is also expected to spur demand for computing power, such as domestic AI servers, cloud computing and chips, as Washington restricts the export of US company Nvidia’s most advanced AI chips to China, experts said.
To help accelerate domestic breakthroughs, Beijing and Shenzhen announced favorable policies in May to support the development of domestically developed AI chips.
iFlytek said it has built first-class data centers at its headquarters in Hefei, Anhui province. Specifically, its deep-learning computing platform is supported by seven data centers in four cities, laying a good hardware computing foundation for the development of its AI large language model.
The company also said it has developed AI training and reasoning solutions powered by domestic computing platforms.
Shen Yang, a professor of journalism and director of the Metaverse Culture Laboratory at Tsinghua University, said generative AI has made significant progress, but it also raises concerns about ethics, copyright protection and privacy, Shen said.
ChatGPT-like applications can free human creators from tedious tasks, but they will not replace humans, Shen added.
This is in line with what ChatGPT wrote when prompted by Israeli President Isaac Herzog in February with the task: “Write an inspirational quote about the role of humanity in a world of superhuman technology.”
The answer generated by ChatGPT was: “Let us not forget that our humanity is what makes us truly special. It is not the machines that will shape our destiny but rather our hearts, minds and determination to create a brighter tomorrow for all humanity.”
