Shenzhen, China – ByteDance, the renowned Chinese tech company known for its popular social media platforms such as TikTok and Douyin, has recently taken a significant step towards advancing its AI capabilities by establishing a Large Model Research Institute. According to AI Technology Review, ByteDance has been quietly preparing for this move and actively recruiting top AI talent.
The establishment of the Large Model Research Institute marks a strategic shift for ByteDance, which has been investing heavily in AI research and development. Insiders reveal that several renowned AI experts have already joined the new research institute, reporting directly to Zhang Yiming, ByteDance’s CEO.
In August 2023, ByteDance launched its self-developed foundational large model, Yun Que, followed by the introduction of the AI dialogue product Douba. This year, ByteDance further expanded its AI offerings by launching the Douba Large Model family, initiating a price war with claims of being 99.3% cheaper than the industry.
The move to establish the Large Model Research Institute comes as no surprise, considering ByteDance’s growing focus on AI. AI Technology Review exclusively reported that the company has been gradually disclosing its progress in large model research since last year.
In November 2023, the Flow project was established within ByteDance, focusing on AI application research and development. The project is led by the company’s Technical Vice President, Hong Dingkun, and the Large Model Team Leader, Zhu Wenjia.
In December 2023, another product, Coze, was launched overseas, offering users a platform for rapid AI application development, with the ability to generate AI Bots in 30 seconds and integrate a suite of plugin tools to meet personalized needs.
The domestic version of Coze, named Kouzi, was officially launched in early 2024. Both the overseas and domestic versions of the product share similar product forms and functional positioning, aiming to facilitate the rapid creation, debugging, and optimization of AI chatbots.
One of the key differences between ByteDance’s Large Model Team and other leading AI startups is the team’s composition. Historically, talent in the Large Model Team has primarily come from internal business lines, such as search, Douyin, XiXi, and TikTok. However, the new Large Model Research Institute is taking a different approach by investing heavily in external talent recruitment.
Last week, Huang Wenhao, co-founder of Zero-One Everything, joined ByteDance. Huang, who holds a Ph.D. in Computer Science from Peking University, was one of the earliest individuals in China to engage in Deep Learning research. After graduation, Huang worked at Microsoft Research Asia on Agent-related research, focusing on chatbots capable of completing tasks.
Following his tenure at Microsoft, Huang joined the Tsinghua University KEG Lab as the Technical Director of the Health Computing Research Center and the Innovation Application Laboratory, focusing on AI for Life Science research. His primary research areas include natural language processing based on biometrics and the integration of natural language processing methods with structural research.
At Zero-One Everything, Huang served as the Algorithm Vice President, responsible for the training and release of multiple open-source models, including Yi-34B. His work with the company has earned recognition on LMSYS and huggingface llm leaderboard.
As ByteDance continues to invest in AI research and development, the establishment of the Large Model Research Institute is a clear indication of the company’s commitment to becoming a leader in the AI industry. With the recruitment of top talent and the development of innovative AI products, ByteDance is poised to make significant advancements in the field of large models.
Views: 0