In defiance of stringent government regulations and persistent challenges in sourcing hardware, Chinese tech giant Tencent has officially launched its “Hunyuan” artificial intelligence (AI) system, a multimodal large language model (LLM) akin to OpenAI’s ChatGPT.
The grand unveiling occurred at the Global Digital Ecosystem Summit in Shenzhen on September 7, marking Tencent’s venture into the realm of “foundational models.” These models serve as the basis upon which associated AI APIs can operate.
Tencent envisions Hunyuan as a comprehensive suite of AI tools, designed to serve a multitude of functions ranging from image creation, copywriting, text recognition, to customer service. These applications are poised to have a profound impact on various sectors, including finance, public services, social media, e-commerce, transportation, gaming, and more.
What sets Hunyuan apart is its seamless integration into Tencent’s extensive ecosystem of applications and services. It seamlessly connects with Tencent Cloud, Tencent Marketing Solutions, Tencent Games, Tencent fintech services, Tencent Meeting, Tencent Docs, Weixin Search, and QQ Browser, enhancing the overall user experience.
The launch of Hunyuan coincides with the backdrop of frosty relations between the United States and China, characterized by the export ban on specific computer chips, crucial for AI system development and training, which the Biden administration imposed in October 2022. Diplomatic talks between the two nations are ongoing, but no concrete timeline for the ban’s termination has emerged.
Tencent’s announcement underscores its commitment to fostering open collaboration within the ecosystem, aiming to provide domestic businesses access to high-quality model services while extending an opportunity for international enterprises to tap into the Chinese market.
It’s noteworthy that ChatGPT is presently unavailable in China. While Hunyuan isn’t the first LLM introduced in the Chinese market, its launch carries significant weight. Tencent Holdings, as China’s leading tech company, wields a nationwide influence, with products reaching every corner of the country. Given the omnipresence of WeChat and its affiliated apps, Hunyuan holds the potential to penetrate the domestic market in ways that OpenAI and Microsoft might struggle to replicate with ChatGPT in Western markets.
Hunyuan is reported to be comparable in power and capability to GPT-3, OpenAI’s core model as of 2022, in terms of raw numerical specifications, boasting a staggering 100 billion parameters and 2 trillion tokens.
However, it’s important to note that parameter and token counts don’t necessarily define the capabilities of a multimodal AI system. Hunyuan’s advantage lies in its extensive training on a vast corpus of Chinese language text, theoretically positioning it ahead of models primarily trained on non-Chinese datasets when operating within the Chinese language environment.