Shanghai-based SenseTime demos chatbot SenseChat, available in Chinese and English, image tool Miaohua, and video tool Ruying, powered by its SenseNova LLM
Context & Ripple Effects
SenseTime is jumping into the generative-AI race from Shanghai with a full stack demo: the chatbot-and-content-tool pattern other Chinese companies would later ship publicly, here split across a bilingual SenseChat assistant, an image generator (Miaohua), and a video tool (Ruying), all running on one homegrown foundation model, SenseNova. The timing matters — Alibaba Cloud's unveiling of Tongyi Qianwen and plans to embed it across DingTalk and Tmall Genie landed the very next day, marking April 2023 as the moment China's big platforms stopped watching ChatGPT and started answering it.
The lineage runs deeper than this news cycle: Microsoft's [[a:831533|Xiaoice had already proven Chinese-language conversational AI at 20M registered users back in 2015]]. What changed is that the underlying models are now large enough to power text, image, and video from a single lab — and for SenseTime, whose stock was down more than 80% since its 2021 Hong Kong IPO at the time, SenseNova became the narrative it would keep building on, culminating in the 30%+ share surge when the latest SenseNova version debuted a year later.
First-order effects
- SenseTime now competes head-to-head with Alibaba Cloud's Tongyi Qianwen in the same week, and unlike Alibaba it has no flagship consumer product like DingTalk to distribute through — SenseChat, Miaohua, and Ruying must find their own audience.
Second-order effects
- Alibaba's plan to push its chatbot into every product sets the distribution benchmark rivals must answer: standalone demos like SenseChat risk losing to assistants embedded where users already work.
Third-order effects
- The September approvals that let Baidu, ByteDance, and peers roll out chatbots publicly show Beijing acting as gatekeeper for who reaches consumers — labs like SenseTime that build state-compatible models are positioned to survive that filter.
The trend: China's AI labs are converging on ChatGPT-style assistant suites built on proprietary LLMs, with regulatory approval becoming the decisive competitive moat over raw model quality.