TikTok Parent ByteDance Ropes In Former Alibaba AI Expert To Chase Its ChatGPT Ambitions

Former leader of Alibaba Group Holding Limited's BABA M6 model DAMO Academy Intelligent Computing Lab joined TikTok parent ByteDance's AI Lab to participate in the R&D of language generation model.
Yang Hongxia quit Alibaba in September 2022 for personal and family reasons, Pandaily reports.
Reportedly, Hongxia is in a leading position within the language generation model team and reports directly to Yang Zhenyuan, vice president of ByteDance.
Also Read: Baidu's ChatGPT Counterpart Ernie Bot Impresses Analysts
Hongxia was the technical director of M6, a super-large-scale multimodal pre-training model in Alibaba's DAMO Academy.
M6 is a universal AI model with multimodal and multi-task capabilities, proficient in designing, writing, and developing questions and answers has broad application prospects in e-commerce, manufacturing, literature and art, scientific research, and other fields.
Alibaba first released M6 in March 2021, and it was the first sizeable multimodal model with 100 billion parameters in China.
Subsequently, it changed to trillion parameters with 480 GPUs. It successfully trained a 10 trillion model within ten days based on 512 GPUs with the same parameter scale as GPT-3, with an energy consumption of only 1%.
By the end of 2022, Alibaba Cloud Moda Community had opened up over 300 AI models, including the M6 and Tongyi.
Hongxia's joining of ByteDance suggests the company's attempts to nail large model development. The ByteDance Search Department leads the company's language model team.
Currently, the team size is about ten people. The intelligent creation team under the Product R&D and Engineering Architecture Department is leading the picture model team.
Reportedly, the image generation model's progress is progressing well, and ByteDance proposes combining it with Douyin, CapCut, and other products.
Photo by olivier-bergeron via Unsplash