
Beijing backs Chinese chatbot developers
Beijing’s Municipal Bureau of Economy and Information Technology will support enterprises in building large AI models comparable to ChatGPT. This was reported by The Register.
According to the announcement, the department aims to strengthen infrastructure, accelerate the provision of base data, and foster collaboration among firms, universities and open-source software developers.
The bureau believes that Beijing is a suitable place to create an analogue of ChatGPT. The city accounts for 29% (1,048) of China’s AI companies.
Unlike many Western web resources such as Google, Facebook and Wikipedia, ChatGPT is not blocked in mainland China. However, Chinese users find it difficult to create an account on the service, as a Chinese phone number cannot be used for registration.
Local tech giants have also stepped up efforts to keep ChatGPT-based mini-programs off their platforms. Attempts to locate OpenAI’s chatbot on WeChat have been unfruitful.
ChatGPT supports more than 90 languages, including Chinese. It can also operate as a translator. However, the bot’s performance will vary, as the app is primarily designed for English.
Chinese firms have already joined the global race of chatbots. Earlier, the web giant Baidu said it would launch testing of a conversational AI in March 2023.
Alibaba said its research institute, the Damo Academy, is conducting internal tests of an AI chatbot. The company did not specify when a product based on it would be released.
Tencent Holdings, NetEase and JD.com are also developing their own equivalents.
In addition, less well-known platforms are emerging, some of which even copy ChatGPT branding.
Peking University sociology professor Jiang Ruxian attempted to explain why China lags in the development of conversational AI. In his view, it all comes down to the structure of Chinese science-and-technology startups.
In September 2022, researchers found that Chinese authorities subjected to political censorship Baidu’s ERNIE-ViLG text-to-image generator.
In June 2021, Chinese developers introduced a language model WuDao 2.0 with 1.75 billion parameters. This is ten times larger than GPT-3.
Рассылки ForkLog: держите руку на пульсе биткоин-индустрии!