Unique Research · Insight / Unique Friend Q&A · January 22, 2024 · Guest: Xue Chong, Founder, Quanzhen Medicine (全诊医学)
Inventory type remains insight (Unique Friend series — not partner welcome).
Guest
Xue Chong — M.D. in surgery, Peking Union Medical College; postdoc, Johns Hopkins; distinguished researcher, Beihang Hangzhou Innovation Institute; formerly clinical work at Zhejiang University Second Affiliated Hospital. Founded Quanzhen Medicine in January 2016 for medical AI R&D and application. Core team mixes M.D.s and computer scientists; 40+ invention patents and software copyrights on self-developed products.
Q&A
1. Product and core scenes?
Quanzhen AI is generative AI for automatic medical text — a doctor’s AI assistant tackling heavy documentation across three scenes: pre-visit patient dialogue → pre-consultation notes; understanding doctor–patient dialogue → outpatient/inpatient notes; doctor↔AI dialogue → ward-round or operative notes. It also offers clinical decision support: diagnostic suggestions and further test/treatment recommendations.
2. 2023 productization / market?
Product is mature and market-validated — among the largest doctor-user bases and globally leading in its niche. Users span independent clinic doctors and large hospitals. Biggest 2023 challenge and win: productizing for Class-III Grade-A hospitals. Co-developed with Zhejiang Provincial People’s Hospital and Sir Run Run Shaw Hospital (ZJU) — inpatient full-flow AI notes and outpatient AI notes respectively — then to nearly 30 Class-III Grade-A hospitals within a year.
3. GenAI / LLM opportunities vs past waves?
GenAI meets content-generation demand across industries; LLMs cut AI R&D cost and push computers toward human-like language — the symbol that fuses hearing and vision, enabling AI Agents that perceive and act. Differences vs past: (1) AI startups run leaner — they save customers’ headcount and use AI on their own ops; (2) deep industry natives move faster on real domain opportunities.
4. China landing challenges?
Weakest link: domestic general LLMs still lag overseas commercial and open-source stacks. Two modes emerge — iOS-like closed OS (OpenAI, Gemini) vs Android-like open (Llama). AIGC startups must pick models by need and vertical.
5. Competing with giants’ “dimensionality reduction”?
Since GPT-3.5, output capability between giants and startups has leveled. Moats: customer scale + high-quality closed data; hit pain points, productize fast, and recirculate product data into a new barrier.
6. Are models still too weak?
Depends on the problem and how far you need to solve it. Even if the base model is short, product overlays can reach customer-perceived value people will pay for.
7. Human–machine collaboration paradigm?
Like Agents, collaboration goes multimodal — ear/mouth (listen/speak) and eye/hand (read/write), the way WeChat chats with real friends. ChatGPT already made AI collaboration feel like messaging a friend. Quanzhen AI lets doctors write notes in multiple modes, not only keyboard typing.
8. 2024 tech to watch?
(1) Multimodal model progress as the substrate for broad Agents; (2) AI fused with machines — EVs, medical robots — coupling “brain” and “muscle” intelligence.
1 CN image; attach on ingest. Signup CTA in CN omitted. Title framing: models/AI amplify existing medical resource potential.
— END —