Chinese cloud and social media conglomerate Tencent has announced that its Hunyuan large language model (LLM) will be available for enterprise use, as companies in the country stake their claim as technology industry and generative AI leaders.
“A war of a hundred models has begun,” said Tencent's vice president, Jiang Jie, in comments published by Reuters. He noted that by July, there were more than 130 LLMs in China. These include the AI models that were recently unveiled by China’s largest search engine, Baidu, and AI software company SenseTime.
Hunyuan has more than 100 billion parameters and was trained with more than 2 trillion tokens, while OpenAI's GPT-3 AI model, backed by Microsoft, contained 175 billion parameters when it was released in 2020 and Meta's Llama-2 model had 70 billion parameters when it was released this year.
Tokens are the units into which a piece of text is divided for processing, while parameters refers to how neural networks are configured.
More than 50 products and services provided by Tencent, which owns the WeChat social media app, are now running on Hunyuan, Tencent said. The company claims that its product is better than ChatGPT at writing long text and solving math problems.
The model also reportedly experiences 30 per cent less hallucinations — the phenomenon where AI models make up answers or presents false information as facts when faced with questions it can’t accurately answer — compared to Llama 2.
China seeks position as AI leader
In April, China became one of the first countries to regulate AI, with the Cyberspace Administration of China (CAC) unveiling a number of draft measures for managing generative AI services, including making providers responsible for the validity of data used to train generative AI tools.
Under Chinese law, AI firms will be required to submit security assessments to the government before launching AI tools to the public, and any content generated by generative AI must be in line with the country’s core socialist values.
Failure to comply with the rules will result in providers being fined, having their services suspended, or facing criminal investigations. In May, Chinese authorities detained a man in northern China for allegedly using ChatGPT to write fake news articles.
However, the escalating US-China trade war may dampen the country’s tech prospects, as US President Joe Biden continues to sign executive orders restricting the sale of certain AI-related exports to China.
In August, Biden signed an executive order further restricting the exportation of semiconductors and microelectronics, quantum information technologies, and AI systems such as software that uses AI for military or intelligence operations.
“This program will seek to prevent foreign countries of concern from exploiting US investment in this narrow set of technologies that are critical to support their development of military, intelligence, surveillance, and cyber-enabled capabilities that risk US national security,” Biden wrote in a letter to Congress.