China’s SenseTime Shows AI Chat, Image Tools

Chinese artificial intelligence firm SenseTime on Monday showed a live demonstration of its latest range of AI-powered tools, including its answers to OpenAI’s ChatGPT chatbot and Dall-E image generator.

The company, which has been sanctioned by the US, joins Chinese tech giants Baidu and Alibaba as well as dominant US search firm Google in releasing the so-called generative AI tools, which can produce human-like text or images from prompts.

SenseTime chief executive Xu Li demonstrated the SenseChat chatbot writing an email, producing a story about a cat catching a fish and writing computer code in response to prompts.

The chatbot, along with other capabilities including an image generator, a digital avatar creation platform and a pair of 3D modelling tools, are based on a new version of the firm’s SenseNova foundation model set.

Chat applications

The latest SenseNova model has been in development for the past five years, according to co-founder Wang Xiaogang.

SenseChat, a large language model (LLM) with hundreds of billions of parameters, can conduct complex dialogues with humans and comprehend extensive texts, SenseTime said.

The firm showed LLM-powered applications including a programming assistant, a health consultation assistant and a PDF file reading assistant that can extract and summarise information from complex documents.

The company showed the SenseMirage tool, which can create images from text prompts, as well as SenseAvatar, a 2D/3D digital avatar generator, and SenseSpace and SenseThings, 3D content-generation platforms that can create large-scale 3D scenes and detailed objects for metaverse and mixed-reality applications.

Avatar generation

“We can generate a digital scene for a livestreaming room. The product creation and interactive content are all generated by AI, and the livestreaming room runs non-stop 24 hours a day,” said SenseTime co-founder chief executive Xu Li.

“In fact, you may not even be able to tell whether it is a real person doing the broadcasting at night.”

SenseTime has put on a US trade blacklist for allegedly developing facial recognition tools to facilitate government surveillance of ethnic Uyghurs in the Xinjiang region. The company strongly denies wrongdoing.

The sanctions limit SenseTime’s ability to obtain high-end AI accelerator chips, but at the event Xu said the company has enough graphics processing units to support its SenseNova operations.