Tencent Teams Up to Make AI Better for Elderly, Kids

Millions of elderly people and children in China turn to AI chatbots for companionship and guidance, but the technology hasn't always understood their unique needs.

Now Tencent is stepping up to change that. The tech company's research team is building specialized datasets to teach AI models how to better respond to vulnerable groups, including China's 69 million "left-behind children" whose parents migrate to cities for work.

Lu Shiyu, a senior researcher at Tencent Research Institute, has been developing these datasets since 2024. His team worked with elderly volunteers who contributed thousands of real questions and answers, creating an "elderly dataset" that helps AI understand how seniors communicate and what support they need.

The researchers tested major AI models from both the US and China, including Tencent's own Hunyuan system. Every single model fell short when handling sensitive topics like sex education and other issues important to rural children living without their parents nearby.

The Ripple Effect

What started as one company's project could soon benefit AI users across China's 1.4 billion population. Tencent isn't keeping these datasets to itself but actively seeking partnerships with other major AI developers who serve millions of users daily.

The timing matters. Vulnerable groups have become progressively more dependent on AI chatbots for emotional support and health guidance, especially in areas where human resources are stretched thin.

By collaborating with Chinese nonprofits that work directly with these communities, the research team ensures the datasets reflect actual needs rather than assumptions. These organizations understand the daily challenges facing elderly users navigating technology and children growing up without parental guidance.

The specialized training happens before AI models launch to the public. Better datasets during the pre-training phase mean smarter, more compassionate responses when real users ask for help.

This work represents a shift in how tech companies think about AI development, moving beyond general knowledge toward understanding the specific ways different groups communicate and seek support.