ByteDance’s Pragmatic AI Strategy: A Factor in Its Lag Behind OpenAI and Its Approach to Catching Up
April 24, 2024 β 00:28
Investment Research Team
Since June 2023, authorities in China have registered 546 generative artificial intelligence algorithms. As the pace of registration accelerates, the breadth of their applications is also expanding. However, the overall capabilities of general purpose large-scale AI models in China still lag behind American leaders such as OpenAI and Anthropic.
There have been four batches of registration for these AI algorithms, falling under two Chinese regulations: the "Internet Information Service Deep Synthesis Management Regulations," passed in November 2022, and the "Interim Measures for the Management of Generative Artificial Intelligence Services," released in July 2023.
The two latest batches in January and February 2024 saw 129 and 266 algorithms registered respectively, compared to 41 and 110 registered in June and August 2023, respectively.
Deep synthesis service providers refer to organizations or individuals who offer deep synthesis services;
Deep synthesis service technology supporters refer to organizations or individuals who provide technical support for deep synthesis services.
Some AI algorithms could fall under both of these categories, leading to potential duplications in the count of registered AI algorithms. Baidu’s Ernie Bot large-scale AI model can be used for external technical support as well as for their own products.
Among the 546 registered algorithms, there are a total of 70 large-scale AI model algorithms. However, the majority of these are used by enterprises for their own business purposes, with only 19 providing external technical support. Moreover, these include multiple large-scale AI models tailored to specific industries, and the number of general-purpose large-scale AI models is relatively small.
China’s major tech companies lead in terms of the number of registered AI algorithms
Tencent and Netease each have 10 algorithms registered, Alibaba’s Damo Academy has seven, Baidu, Huawei Cloud, and SenseTime each have six, while ByteDance has five.
Even though Tencent has maintained a relatively low profile in this wave of AI enthusiasm, it ranks first in the list of AI algorithm technical supporters category. Besides its formidable technical prowess, Tencent has AI algorithms in multiple fields including facial fusion, voice synthesis, drawing, and image editing.
Similar to Tencent, NetEase has algorithms covering multiple industries such as image generation, voice generation, text generation as a technical supporter. As a service provider, its algorithms serve its own applications like its Youdao Dictionary and mobile games.
However, when it comes to the widest range of algorithms used in its own applications, no company surpasses Alibaba. Possibly due to organizational restructuring, Alibaba’s algorithms for external technical support all come from its Damo Academy.
However, Alibaba’s own applications such as Taobao, DingTalk, and Tmall each have their own registered AI algorithms. Most of the technology used in Taobao and DingTalk is developed in-house, although there are some instances where technology is provided by the Damo Academy or developed jointly.
Baidu’s Ernie Bot large-scale AI model is currently a leading player among Chinese general purpose large-scale AI models. It has algorithms in various fields such as digital human, speech synthesis and recognition, text-to-text, image-to-text, and facial fusion.
Huawei Cloud, on the other hand, has fully embraced the "B2B" route. Currently, it has registered six algorithms, all listed as technical supporters.
More Industry-Specific AI Algorithms
It appears that China’s generative AI algorithms are more tailored to vertical industries, according to the list of algorithms registered with authorities.
For example, e-commerce is a key industry for the development of large-scale AI models, with Alibaba’s e-commerce platforms being particularly active in embracing AI. In the short video sector, generative AI technology is widely used, with platforms like ByteDance’s Douyin and video app Kuaishou leading the way. The applications in this field mainly focus on special effects, synthesis, and face swapping.
Even traditional industries are actively embracing generative AI. In the healthcare sector, applications are mainly concentrated on health consultations and medical content generation. Home appliance giant TCL uses its own intelligent text generation algorithm in its app to generate product descriptions and usage guides based on user-entered text content.
In the education sector, TAL Education Group has filed a total of 8 algorithms, some for internal use and some for external output. In the maternity and baby industry, the Babytree Parenting app utilizes large-scale generative AI algorithms to generate content related to pregnancy and childcare based on user-input questions and providing text responses.
Another important application area is the automotive industry. BYD recently launched a new vehicle intelligent architecture "Xuanji" and its AI large model "Xuanji AI Large Model" to pursue smart driving in electric vehicles.
Related News:
