Tencent, China’s largest tech giant, said on Thursday it had started internal testing of its self-developed artificial intelligence (AI) model, and the platform is now being integrated with several internal services and products. The company’s foundation AI model, Hunyuan, has been integrated with various products, including Tencent Cloud, Tencent Meeting, and Tencent Docs.
The development of Hunyuan is part of Tencent’s AI strategy to make the technology widely used in its products and services, a key initiative of founder and chairman Ma Yun. Tencent has also set up an AI lab with a research focus on areas such as natural language processing, computer vision, and speech recognition, which are essential for the firm’s social networks QQ and WeChat. The lab has around 70 scientists and 300 engineers.
It has integrated Hunyuan into its advertising system for AI-based content creation and optimization. According to Lau, it will be rolled out to more lines of business in the future, such as gaming and finance.
Tencent is a big believer that developing its own AI models is essential because the firm can better understand users’ needs and preferences, allowing it to tailor products and services better to meet those requirements. In addition, developing its models will help it maintain a competitive advantage against other companies with different technical resources and expertise.
One such example is Tencent’s new AI conferencing service Tencent Meeting, launched in March this year to compete with global players such as Zoom and Microsoft Teams. The service supports various work habits, from flexible call-ins through smartphones, personal computers, and Weixin mini-programs to real-time screen sharing and instant text messaging that facilitate collaboration during meetings.
A new feature introduced in the service is video analysis that can automatically identify and categorize faces in videos, recognize the background of the scene, detect and remove unwanted objects from scenes and enhance images by removing noise and improving clarity. In a blog post on Thursday, Tencent said the new features are designed to support the company’s new work style of combining traditional office and remote working practices.
Besides its new AI-based conference services, Tencent already has several other AI products. Last year, the firm launched its DI-X platform, which provides a one-stop shop for its customers’ AI development and is built on top of Tencent Cloud’s computing and storage capabilities. It is equipped with mainstream deep learning frameworks such as Caffe, TensorFlow, and Torch, allowing customers to develop their AI algorithms and models quickly.
It was also the first platform to launch an Intel DL Boost product tailored explicitly to deep learning and helped accelerate machine learning training by up to two times. DI-X has been widely adopted by many of Tencent’s large customers as they build AI applications with the platform.
