Tencent is testing an AI model as AI buzz intensifies

August 6, 2023 4:18 pm

Tencent is internally testing an artificial intelligence-powered chatbot, joining in a race with other Chinese internet conglomerates like Alibaba, Baidu to develop a local version of ChatGPT.

Tencent has recently announced the successful integration of its foundational AI model called "Hunyuan" into various products and services. This AI model has been seamlessly incorporated into Tencent Cloud, Tencent Meeting, and Tencent Docs, among others. By leveraging the capabilities of "Hunyuan," Tencent aims to enhance the performance and functionality of these products, offering users more efficient and advanced AI-driven features.

An anonymous Tencent employees said on Chinese microblogging site Weibo that he has received internal emails inviting them to experience the "Hunyuan" large model through a webpage or a mini-program. It is understood that the "Hunyuan" supports text and voice inputs, and can generate content such as short video scripts, social media marketing texts, and WeChat moments posts. It also supports AI-generated content (AIGC) art in styles such as cyberpunk, pixel art, and mosaics.

With the "Hunyuan" large model, employees can use it for various work scenarios, such as writing emails, outlining interview questions, creating PowerPoint outlines, and designing research questionnaires. In programming scenarios, employees can experience domain name recommendations, code generation, code explanations, and SQL (Structured Query Language) statement generation. Users can use the model to plan their travel itineraries, fitness routines, or food guides.

China's internet sector has recently witnessed a significant surge in AI investment, with companies like Baidu leading the charge. In March, Baidu unveiled its major response to ChatGPT with Ernie Bot, sparking a competition among other companies such as Alibaba, ByteDance and SenseTime Group Inc. to introduce their own competing platforms. While local Chinese media outlets have reported that ByteDance is developing its own large language models called “Grace”.

"Grace" was developed within Feishu, ByteDance's enterprise collaboration platform known as Lark overseas. Currently, the testing phase is limited to a specific group of employees. When employees access the service, they will be presented with a pop-up message indicating that it relies on multiple extensive language models.