Tencent Executive Addresses AI Progress Criticism, Cites Computing Power Shortage
Tang Daosheng, Tencent's Senior Executive Vice President and President of its Cloud and Smart Industry Group, responded to criticism that the company has been slow in AI development. Writing in Tencent's internal publication Zhidian, he acknowledged fierce competition in the global large model market and said the Hunyuan model underwent multiple reconstructions, with the Hy3 architecture emphasizing training data quality. However, severe computing power shortages slowed model training and product development.
Tang Daosheng, Senior Executive Vice President of Tencent Group and President of its Cloud and Smart Industry Group, published an article in Tencent's internal publication Zhidian, responding to external criticism that "Tencent has been slow in AI." Tang said competition in the global large model market is intense, with news of new model releases and expanding parameter scales arriving daily, drawing close attention from enterprise customers. Given that peers have held an advantage in AI momentum over the past few years, it would be unrealistic to claim a complete absence of anxiety.
Tang noted that the Hunyuan large model has gone through different stages of development. After multiple reconstructions, the Hunyuan Hy3 architecture simplified complexity and placed greater emphasis on training data quality, delivering strong results among large models of comparable parameter scale. However, the company's overall computing power remains severely insufficient, which has slowed model training and product development and produced a significant impact.