Dialogue with Wang Xiaochuan: After Straying from AGI’s Mainstream Path 对话王小川：离开通用人工智能的主干道之后

The Strategic Shift: From Mainstream Race to Niche Focus

The article highlights a pivotal moment for Baichuan Intelligence. While the AI industry was engaged in a frenetic "arms race" to release new general-purpose large models every few days, Wang Xiaochuan made a counterintuitive decision: to drastically downsize the general model team, shut down multiple industry-specific lines, and commit entirely to the healthcare vertical.

• Background and Pressure: This decision came during a period of existential uncertainty for the company. Despite having a substantial team, Baichuan was stretched thin, simultaneously pursuing general models, healthcare, and commercialization. Wang describes this as not knowing "what we were really doing or what value we were creating."
• The Non-Consensus Choice: Choosing healthcare was seen as abandoning the mainstream, high-visibility path of general AI. This led to departures of key personnel and resistance from investors eager for a quicker IPO and commercial returns. Wang acknowledges the "loneliness" of this path but argues that continuing the general model race would have brought a different, equally profound type of anxiety.
• Core Philosophy: Wang’s logic rests on a distinction between following trends and pursuing a deeply believed problem. He returned to his original entrepreneurial vision of creating "life models" and AI doctors, viewing the advent of models like ChatGPT not as a goal in itself, but as a powerful tool to enable that long-held mission.

The "Why" and "How" of an AI Doctor

Wang's bet is not just on a technology, but on a specific solution to a systemic problem: China's severe shortage of quality medical supply.

• Patient-Centric, Not Just Doctor-Efficiency: The common approach in health-tech—helping doctors be more efficient—makes less sense in China, where a doctor may already see 50-80 patients a day. Instead, Baichuan aims to "create more doctors" by building AI agents that act as family doctors, managing patient health proactively.
• Product Vision: "Bai Xiao Yi": This AI agent is designed for continuous, long-term companionship, not just episodic Q&A. Integrated via WeChat and an app, it can prepare medical summaries for doctors, analyze prescriptions, manage medical history, and send medication reminders. Its "persistent memory" system stores health data across a patient's lifetime, a feature crucial for medical care but often missing in general chatbots.
• Deep Integration into the Medical System: A key part of the strategy is to embed AI into the existing healthcare infrastructure. Baichuan's AI pediatrician has been deployed in Beijing Children's Hospital for multi-disciplinary consultations and is being expanded to over 150 county-level hospitals, indicating a path toward scalable, systemic impact through AI-assisted tiered diagnosis.

Commercialization and The Road Ahead

The article challenges the common notion that healthcare is a "longer, slower path." Wang rejects this as a "inertia of the era."

• The New Logic of Speed: Wang points to the rapid rise of Coding Agents as evidence that traditional boundaries can be shattered. The speed of innovation in AI applications changes the calculus for how quickly value can be delivered and, consequently, monetized in other domains.
• The "Water to a Canal" Theory: Wang’s core commercial belief is that if a company can deliver sufficiently critical value to users, commercial success will follow naturally. For healthcare, this value is not just convenience but improved health outcomes and access.

本文通过对王小川的深度访谈，揭示了百川智能在AI行业狂热期内一次与众不同的战略抉择，其背后蕴含着对技术趋势、商业本质和创业初心的深刻思考。

一、战略转变：从“卷通用”到“All in医疗”

在行业大厂与创业公司争相发布通用大模型的背景下，百川智能在一年前进行了“大刹车”：

• 行业现状：通用大模型进入“轰炸式更新”的同质化竞争阶段，平均3天就有新版本。
• 百川的选择：大幅缩减通用模型团队，关闭多条行业线，All in医疗大模型。
• 直接成果：近期发布医疗大模型M4及Agent产品 “百小医”。
• 代价与动荡：团队规模缩小，部分合伙人离开，原定上市节奏延迟。

二、路径选择：为何是医疗？逻辑何在？

王小川的选择并非偶然，其背后有一套清晰的逻辑链：

1. 对主流道路的反思：他认为，即使留在通用模型赛道并成功上市，焦虑也不会减少，因为公司需明确自身创造的独特价值。
2. 回归创业初心：从创立之初，王小川就想做“生命模型”、造AI医生。ChatGPT的出现被视为实现该目标的助力，而非方向的改变。
3. 独特的商业判断：
   • 否定“医疗是更长更慢的路”的观点，认为这是时代惯性。
   • 以Coding Agent的快速崛起为例，论证AI能打破旧有行业边界。
   • 核心理念：在AI时代，只要交付足够重要的价值，商业化将水到渠成。

三、产品逻辑：以患者为中心，造“更多的医生”

百川的医疗路径并非泛泛而谈，其产品设计体现了鲜明的理念：

• 核心理念：不做单纯的医生提效工具，而是以患者为中心，旨在增加医生供给。
• 产品形态：
  • B端：AI儿科医生已在北京儿童医院“上岗”，会诊吻合率达95%，并开始向基层下沉。
  • C端：“百小医”定位为 “AI家庭医生” ，通过App和微信生态提供主动、全周期的健康管理。
• 技术亮点（M4模型）：
  • Agent架构：从对话走向临床，能主动执行复杂工作流。
  • 强大能力：幻觉减少、循证能力增强、提问能力提升（每提升2点提问能力可增加1点诊断准确率）。
  • 长期记忆：拥有基于患者全生命周期的数据存储和管理能力。

四、深层思考：非共识道路的孤独与价值

文章透露出王小川选择背后的深层心理与行业启示：

• 应对孤独感：转型面临内外部（团队、投资人）的巨大不理解和曲解，但王小川表示，比这更难忍受的是公司缺乏清晰的价值锚点。
• 拒绝“矩阵化”：他反思之前同时开拓多条业务线（通用模型、医疗、商业化）的错误，认为在单点未突破前，矩阵化管理危险且低效。
• 中国医疗的特殊性：
  • 中国医生已极度繁忙（日诊50-80人），单纯的“提效”空间有限。
  • 核心矛盾是优质医疗资源严重不足。
  • 因此，百川的AI医生定位为融入体系，做分级诊疗的前置环节和主动健康管理。
• 创业哲学的彰显：王小川的选择代表了一种创业思路——在喧嚣中寻找真正相信的问题，并愿意用足够长的时间去回答它。这不仅是业务选择，更是价值观的体现。

总结

王小川与百川的故事，是AI淘金热中一则关于“选择”与“坚信”的寓言。在众人涌向看似最宽阔的通用大模型赛道时，他们选择了一条狭窄但认为更深的道路。这背后，是对技术终极价值的追问，是对**医疗领域核心