Health & Wellness | 健康知识库

by 大牙哥

治愈所有人类疾病与健康和科技的未来 | Mark Zuckerberg & Dr. Priscilla Chan

Curing All Human Diseases & the Future of Health & Technology | Mark Zuckerberg & Dr. Priscilla Chan

17分钟阅读

治愈所有人类疾病：Chan Zuckerberg Initiative与健康技术的未来

摘要

Meta首席执行官Mark Zuckerberg与Chan Zuckerberg Initiative联合创始人Priscilla Chan博士探讨了他们的慈善使命——在21世纪末治愈、预防或控制所有人类疾病。CZI通过资助前沿基础科学、构建开源软件与硬件工具，以及将AI应用于海量生物学数据集来推进这一目标。对话还涵盖了社交媒体对心理健康的影响、VR/AR的未来，以及AI如何改变日常生活。

核心要点

  • CZI的核心策略并非直接治愈疾病，而是为科学家提供更好的工具以加速发现——正如望远镜的出现先于天文学的重大突破。
  • 细胞是理解疾病的基本单元；绘制人体约37万亿个细胞及其状态（健康vs.疾病）的完整图谱，是CZI方法论的核心。
  • 单细胞测序已经揭示了全新的细胞类型（例如，囊性纤维化中此前未知的肺细胞类型），从而改变了我们对疾病的认识。
  • **AI与大型语言模型（LLM）**可作为假设生成器，处理海量生物学数据集，发现任何研究人员都无法独自察觉的规律。
  • 虚拟细胞项目旨在基于人类细胞图谱构建一个AI驱动的人体细胞模拟系统，实现快速、低成本的计算机模拟实验。
  • CZI芝加哥Biohub正在将纳米级传感器嵌入工程化组织（以皮肤为起点），以实时测量细胞通讯和炎症信号。
  • CZI纽约Biohub正在将免疫细胞改造为体内传感器——在体内巡航，检测动脉斑块等问题，并有望将其清除。
  • 罕见病是了解正常生物学的窗口——研究7,000余种罕见病，可以阐明与所有疾病相关的基本机制。
  • 社交媒体对健康的影响在很大程度上取决于使用方式：主动与他人建立连接与幸福感相关；被动消费负面内容则不然。
  • 技术本身并无好坏之分，但设计选择——包括引导用户跳出内容循环的算法机制以及青少年家长控制功能——对健康结果有显著影响。

详细笔记

CZI的使命与战略

  • CZI由Mark Zuckerberg与Priscilla Chan博士于2015年创立，明确目标为：在本世纪末治愈、预防或控制所有人类疾病。
  • CZI并不打算自行攻克疾病，而是致力于加快科学进步的步伐，为全球研究人员提供更好的工具。
  • 灵感来源于科学史：重大发现通常以新工具的出现为先导（例如，望远镜推动了天体物理学的发展；显微镜推动了生物学的发展）。
  • 三种运作模式：
    1. 资助科学家——包括激励跨学科合作，以及推动开放科学与预印本共享
    2. 构建工具——开源软件（如CELLxGENE、Napari）和硬件（如电子显微镜）
    3. 开展科学研究——通过Biohub研究所网络推进

以细胞为中心理解疾病

  • 人体估计含有37万亿个细胞，每个细胞对DNA指令的解读（通过mRNA）略有不同。
  • 现代医学在很大程度上理解了从基因突变到疾病的路径（例如，CFTR基因突变→囊性纤维化），但对中间的细胞层面步骤尚不清楚。
  • Chan博士的类比：“我们知道食谱里有个错别字，而且蛋糕很难吃——但我们不知道中间发生了什么。”
  • 单细胞测序发现了一种受囊性纤维化突变影响的全新肺细胞类型，从而重塑了对该疾病的理解。
  • CZI的CELLxGENE工具允许任何研究人员输入一个基因，即可获得显示哪些细胞类型表达该基因的热图——支持跨器官假设生成（例如，一个心脏病相关基因在胰腺中同样活跃）。

人类细胞图谱与AI

  • CZI是人类细胞图谱的主要资助方——这是一个自2017年起开发的、覆盖人类、小鼠和果蝇的单细胞水平细胞类型近完整图谱。
  • 该数据集规模庞大，仅凭人工分析远远不够；这正是**大型语言模型（LLM）**发挥关键作用之处。
  • LLM是在海量数据集上训练的模式识别系统；当以细胞生物学数据而非语言数据进行训练时，它们可以预测细胞状态和相互作用——类似于AlphaFold（DeepMind）解决蛋白质折叠问题的方式。
  • CZI正在建设全球规模最大的非营利生命科学AI计算集群之一（约1,000个GPU），用于在人类细胞图谱数据上训练模型。
  • 重要注意事项：LLM会产生幻觉——输出结果必须由科学家进行验证。其最佳用途是作为假设生成器，而非提供最终答案。
  • 长远目标：构建虚拟细胞——一个人体细胞的AI模拟系统，使研究人员能够以低成本、快速的方式开展计算机模拟（in silico）实验。

Biohub网络

旧金山Biohub（斯坦福大学、加州大学伯克利分校、加州大学旧金山分校） ——专注于单细胞生物学和传染病。

芝加哥Biohub（伊利诺伊大学厄巴纳-香槟分校、芝加哥大学、西北大学）

  • 研发组织传感器：将超薄传感器嵌入整个工程化皮肤组织
  • 读取细胞分泌物、细胞间通讯方式，以及炎症的最早期信号
  • 炎症导致约50%的死亡
  • 还研究神经肌肉接头——这对肌萎缩侧索硬化症和衰老至关重要（传导减慢会导致老年人跌倒）

纽约Biohub

  • 将免疫细胞改造为体内传感器（“细胞内窥镜”）
  • 细胞在体内巡航，检查冠状动脉中的动脉粥样硬化斑块、卵巢癌和胰腺癌，以及神经退行性疾病
  • 第二阶段：改造细胞，使其能够主动治疗所发现的问题（靶向免疫细胞疗法）
  • 预计需要10至15年的开发周期

罕见病与转化研究

  • CZI的Rare As One项目组合资助患者倡导团体，帮助其建立生物登记库，并与研究人员及药物开发者合作。
  • 超过7,000种罕见病共同影响着数百万人——每一种都为了解正常人体生物学提供了一扇精准的窗口。
  • CZI不直接从事药物开发——这一职能留给生物技术初创企业和公司；部分公司从CZI资助的Biohub工作中衍生而来。

社交媒体、技术与心理健康

  • 研究表明，技术对健康的影响并非一概而论——使用方式至关重要。
  • 通过平台进行的主动社交连接与幸福感乃至长寿相关（与关于社会关系与健康的更广泛文献相一致）。
  • 被动消费负面内容（例如，持续不断的负面新闻）则不具有同样的益处。
  • Meta的设计干预措施包括：
    • 16岁以下账户默认启用隐私设置
    • 家长监护工具
    • 内容循环提示：当青少年陷入内容循环时，系统会提示其拓展内容多样性
    • 采用长期反馈信号（而非仅凭点击量）来对抗标题党内容
  • Zuckerberg提醒应避免在内容设计上过度家长主义——用户偏好差异显著。

CZI背后的个人动机

  • Priscilla Chan博士是中越裔难民（“船民”）的女儿，其家人在越战后逃离越南；家人愿意为了更美好的未来冒一切风险，这是她乐观主义精神的根基。
  • CZI正式成立于他们第一个孩子出生的当天——Zuckerberg在医院产房里编辑着宣告成立的公开信。
  • 有了孩子改变了他们的时间视野，促使他们迫切地将长期以来的抱负付诸行动。

相关概念

  • 人类细胞图谱
  • 单细胞测序
  • mRNA
  • 大型语言模型
  • 人工智能
  • 蛋白质折叠
  • 囊性纤维化
  • 炎症
  • 神经肌肉接头
  • 肌萎缩侧索硬化症
  • 动脉粥样硬化
  • 癌症
  • 罕见病
  • 计算机模拟研究
  • 社交媒体与心理健康
  • 肠脑轴
  • 免疫系统

English Original 英文原文

Curing All Human Diseases: The Chan Zuckerberg Initiative & the Future of Health Technology

Summary

Mark Zuckerberg (CEO of Meta) and Dr. Priscilla Chan (co-founder of the Chan Zuckerberg Initiative) discuss their philanthropic mission to cure, prevent, or manage all human diseases by the end of the 21st century. CZI pursues this goal by funding cutting-edge basic science, building open-source software and hardware tools, and applying AI to massive biological datasets. The conversation also covers the role of social media in mental health, the future of VR/AR, and how AI stands to transform everyday life.

Key Takeaways

  • CZI’s core strategy is not to cure diseases directly, but to give scientists better tools to accelerate discovery — mirroring how the telescope preceded astronomy breakthroughs.
  • The cell is the fundamental unit of disease understanding; cataloging all ~37 trillion human cells and their states (healthy vs. diseased) is central to CZI’s approach.
  • Single-cell sequencing has already revealed entirely new cell types (e.g., a previously unknown lung cell type in cystic fibrosis) that change our understanding of disease.
  • AI and large language models (LLMs) can function as hypothesis generators — processing enormous biological datasets to find patterns no human researcher could detect alone.
  • The virtual cell project aims to build an AI-powered simulation of human cells trained on the Human Cell Atlas, enabling fast, cheap, in silico experimentation.
  • CZI’s Chicago Biohub is embedding nanoscale sensors into engineered tissue (starting with skin) to measure real-time cellular communication and inflammation.
  • CZI’s New York Biohub is engineering immune cells to act as internal sensors — navigating the body, detecting problems like arterial plaques, and potentially clearing them.
  • Rare diseases are windows into normal biology — understanding the ~7,000+ rare diseases can illuminate fundamental mechanisms relevant to all disease.
  • Social media’s effect on health depends heavily on use pattern: active connection with others is associated with well-being; passive consumption of negative content is not.
  • Technology is not inherently harmful — but design choices, including algorithmic nudges away from content loops and parental controls for teens, significantly shape health outcomes.

Detailed Notes

The CZI Mission and Strategy

  • Founded in 2015 by Mark Zuckerberg and Dr. Priscilla Chan with the stated goal: cure, prevent, or manage all human diseases by the end of the century.
  • CZI does not aim to cure diseases itself — it aims to accelerate the pace of science by equipping researchers worldwide with better tools.
  • Inspired by the history of science: major discoveries are typically preceded by new tools (e.g., the telescope enabled astrophysics; the microscope enabled biology).
  • Three modes of operation:
    1. Fund scientists — including incentivizing cross-disciplinary collaboration and open/pre-print science sharing
    2. Build tools — open-source software (e.g., CELLxGENE, Napari) and hardware (e.g., electron microscopes)
    3. Do science — through a network of Biohub institutes

The Cell as the Center of Disease

  • The human body contains an estimated 37 trillion cells, each interpreting DNA instructions (via mRNA) slightly differently.
  • Current medicine largely understands the path from genetic mutation → disease (e.g., a CFTR mutation → cystic fibrosis), but not the intermediate cellular steps.
  • Dr. Chan’s analogy: “We know there’s a typo in the recipe and the cake is awful — but we don’t know what happens in between.”
  • Single-cell sequencing revealed an entirely new lung cell type affected by the cystic fibrosis mutation, reshaping disease understanding.
  • CZI’s CELLxGENE tool allows any researcher to input a gene and receive a heat map showing which cell types express it — enabling cross-organ hypothesis generation (e.g., a heart disease gene also active in the pancreas).

The Human Cell Atlas and AI

  • CZI has been a major funder of the Human Cell Atlas — a near-complete map of cell types across humans, mice, and flies at the single-cell level, developed since 2017.
  • The dataset is too large for human analysis alone; this is where large language models (LLMs) become critical.
  • LLMs are pattern recognition systems trained on massive datasets; when trained on cell biology data instead of language, they can predict cell states and interactions — similar to how AlphaFold (DeepMind) solved protein folding.
  • CZI is building one of the largest non-profit life sciences AI clusters (~1,000 GPUs) to train models on Human Cell Atlas data.
  • Key caveat: LLMs hallucinate — outputs must be validated by scientists. Their best use is as hypothesis generators, not final answers.
  • Long-term goal: a virtual cell — an AI simulation of human cells that allows researchers to test interventions cheaply and rapidly (in silico).

The Biohub Network

SF Biohub (Stanford, UC Berkeley, UCSF) — focused on single-cell biology and infectious disease.

Chicago Biohub (UIUC, University of Chicago, Northwestern)

  • Engineering tissue sensors: ultra-thin sensors embedded throughout engineered skin tissue
  • Reads out what cells are secreting, how they communicate, and the earliest signals of inflammation
  • Inflammation drives approximately 50% of all deaths
  • Also studying the neuromuscular junction — critical in ALS and aging (slowed transmission causes falls in elderly)

New York Biohub

  • Engineering immune cells as internal sensors (“cellular endoscopes”)
  • Cells navigate the body to inspect coronary arteries for plaques, ovarian and pancreatic cancer, and neurodegenerative disease
  • Second phase: engineer cells to actively treat what they find (targeted immune cell therapy)
  • 10–15 year development horizon

Rare Disease and Translation

  • CZI’s Rare As One portfolio funds patient advocacy groups to build bioregistries and collaborate with researchers and drug developers.
  • Over 7,000 rare diseases collectively affect millions — and each offers a precise window into normal human biology.
  • CZI does not pursue drug development directly — that role is left to biotech startups and companies; some spin out from CZI-funded Biohub work.

Social Media, Technology, and Mental Health

  • Research suggests technology’s effect on health is not uniformly good or bad — use pattern matters significantly.
  • Active social connection via platforms correlates with well-being and even longevity (consistent with the broader literature on social relationships and health).
  • Passive consumption of negative content (e.g., relentlessly negative news) does not carry the same benefits.
  • Meta’s design interventions include:
    • Under-16 accounts default to private settings
    • Parental oversight tools
    • Content loop nudges: teens are prompted to diversify content if stuck in a loop
    • Long-term feedback signals (not just clicks) used to combat clickbait
  • Zuckerberg cautions against excessive paternalism in content design — user preferences vary widely.

Personal Motivations Behind CZI

  • Dr. Chan is the daughter of Chinese-Vietnamese refugees (“boat people”) who fled Vietnam after the war; her family’s willingness to risk everything for a better future is a foundational source of her optimism.
  • CZI was formally launched the same day their first child was born — Zuckerberg was editing the announcement letter in the hospital delivery room.
  • Having children shifted their time horizon and created urgency to act on long-held ambitions.

Mentioned Concepts

  • Human Cell Atlas
  • single-cell sequencing
  • mRNA
  • large language models
  • artificial intelligence
  • protein folding
  • cystic fibrosis
  • inflammation
  • neuromuscular junction
  • ALS
  • atherosclerosis
  • cancer
  • rare disease
  • in silico research
  • social media and mental health
  • gut-brain axis
  • immune system

关系图谱