China’s search engine pioneer, Sogou, has released an open-source large language model (LLM) called the Sogou Open Language Model (SOL) to rival OpenAI. The SOL is based on the Transformer architecture and is trained on a massive dataset of Chinese text, including news articles, social media posts, and other online content. The model is designed to help developers create natural language processing applications in Chinese, such as chatbots, machine translation, and speech recognition.
Sogou founder Wang Xiaochuan had previously stated that “China needs its own OpenAI.” The release of this LLM brings him closer to that goal. The SOL is available on GitHub and is expected to be useful in a wide range of applications.
Baichuan-Inc, an AI venture set up by Wang Xiaochuan, had previously unveiled a large language model called baichuan-7B with 7 billion parameters. The launch comes two months after the establishment of the new AI startup. Seen by Wang as Baichuan’s first milestone, baichuan-7B is open-source and has achieved “excellent results” on the three most influential Chinese language assessment benchmarks.
Wang Xiaochuan announced his foray into large-scale models by co-founding an artificial intelligence company called Baichuan Intelligence with former Sogou COO Ru Liyun. The objective is to develop a Chinese version of OpenAI’s fundamental large-scale model and innovative upper-layer applications. Wang and his associates have already provided $50 million in funding to the new company.
In addition to Sogou, other Chinese AI firms have developed powerful alternatives to Western models. A Chinese language model called GLM-130B performs better than OpenAI’s GPT-3 and Google’s PaLM. Large AI models for language, code, and images play a central role in the current proliferation of artificial intelligence.
The release of the Sogou Open Language Model (SOL) brings China closer to its goal of creating its own OpenAI. The SOL is expected to be useful in a wide range of applications, and Sogou’s founder Wang Xiaochuan is continuing to invest in the development of large-scale models through Baichuan Intelligence.