As promised, OpenAI started blocking access to its application programming interface (API) from Chinese IP addresses this week, but that has not stopped developers from using GPT – yet. Developers in mainland China are turning to virtual private networks and third-party services like Microsoft Azure to continue using GPT models. But the writing may be on the wall, and some are making backup plans that include turning to domestically trained models. They are certainly not starved for choice. Seemingly every name in China tech, large and small, is involved in the artificial intelligence (AI) race somehow. The leading generative AI infrastructure service providers right now are TikTok owner ByteDance, US-sanctioned AI firm SenseTime and Alibaba, owner of the Post. These companies not only have their own models available, they run the cloud infrastructure for enterprises to build their own applications using large language models (LLMs). Alibaba and SenseTime in particular boasted of making big gains in their AI services at the World Artificial Intelligence Conference (WAIC) last week. SenseTime released its new SenseNova 5.5 model, and Alibaba said downloads of its Tongyi Qianwen LLMs doubled to 20 million in the last two months. Alibaba Cloud also on Monday released a new ad campaign featuring the founder of Moonshot AI, one of China’s four “AI tigers” – the start-ups that are supposed to be the country’s best bet at producing its own OpenAI. Since OpenAI released ChatGPT at the end of 2022, Chinese companies have been pouring resources into making their own LLMs. Now they are seeing some results. The latest open-source Qwen model from Alibaba was ranked third globally in a recent benchmark test, behind only the latest models from OpenAI and Anthropic, the maker of Claude. Without access to the most cutting edge AI hardware – most notably the advanced chips that Nvidia is barred from exporting to China – domestic companies are still seeing gains from the so-called scaling laws, which show improved performance from larger models, more data, and greater computational resources. Some Chinese AI firms are adding more chips to make up for using older technology. https://www.scmp.com/tech.. #🇨🇳 ChinA.I. 🤖🧠🦾🤖 #Expats in China