A discussion on the evolution of AI technology, from pre-training to reinforcement learning and now recursive self-improvement, with agents becoming a reality in 2025.
要点 · TL;DR
AI 正进入递归自我改进的新时代，发展速度加快。 AI is entering a new era of recursive self-improvement, accelerating development velocity.
信任对 AI 采用至关重要；政府和公司必须合作进行安全测试。 Trust is crucial for AI adoption; governments and companies must collaborate on safety testing.
数据是新石油；政府应将医疗和国家安全数据视为战略资产。 Data is the new oil; governments should treat healthcare and national security data as strategic assets.
核心观点 · Key points
AI 正进入递归自我改进的新时代，开发速度正在加速。 AI is entering a new era of recursive self-improvement, accelerating development velocity.
智能体现已真正发挥作用，从编码开始扩展到个人智能体。 Agents are now working in earnest, starting with coding and expanding to personal agents.
信任对 AI 采用至关重要；政府和公司必须合作进行安全测试。 Trust is crucial for AI adoption; governments and companies must collaborate on safety testing.
数据是新的石油；政府应将医疗和国家安全数据视为战略资产。 Data is the new oil; governments should treat healthcare and national security data as strategic assets.
在公共服务中部署 AI 可以展示切实利益并建立信任。 AI deployment in public services can demonstrate tangible benefits and build trust.
组织必须成为 AI 优先，自上而下指令与自下而上成功案例结合。 Organizations must become AI-first, with top-down mandate and bottom-up success stories.
反共识 · Contrarian takes
缩放定律持续；关于收益递减的怀疑是错误的。 Scaling laws continue; skepticism about diminishing returns was wrong.
递归自我改进标志着超越预训练和强化学习的新范式。 Recursive self-improvement marks a new paradigm beyond pre-training and reinforcement learning.
经过多年炒作，智能体终于在 2025 年底开始真正工作。 Agents are finally working after years of hype, starting in late 2025.
印度的消费 AI 初创公司比美国还多，挑战了西方主导的观念。 India has more consumer AI startups than the US, challenging the notion of Western dominance.
技术领导力取决于部署而非发明；印度是典范。 Technology leadership depends on deployment, not invention; India exemplifies this.
政治领袖必须将 AI 视为当务之急，而非明日议题。 Political leaders must treat AI as an immediate priority, not a tomorrow issue.
本期章节 · Chapters（共 21）
引言：政企合作Introduction: Government and Industry Collaboration
加速与信任Acceleration and Trust
儿童与 AI 焦虑轶事Anecdote about children and AI anxiety
全球 AI 态度与信任鸿沟Global attitudes towards AI and trust gap
政府与 AI 安全机构角色Role of governments and AI security institutes
用 AI 重塑政府服务Reimagining Government Services with AI
开场与介绍Ingredients for turning AI ambitions into reality
智能体走向现实Opening and Introduction
信任与政府角色Agents becoming real
公共部门 AI 用例Trust and Government Role
技术扩散与部署策略AI Use Cases in Public Sector
国家 AI 战略核心要素Technology Diffusion and Deployment Strategy
企业与政府 AI 采纳类比Core elements of a national AI strategy
自上而下与自下而上结合Parallel between enterprise and government AI adoption
展望：AI 加速扩散Mix of top-down and bottom-up for AI transformation
个人与群体超级智能Looking ahead: accelerating pace and diffusion of AI
AI 峰会与政府角色Personal and Interpersonal Superintelligence
AI 发展时代AI Summit and Government Role
AI 信任与焦虑Eras of AI Development
构建 AI 产品信任Trust and Anxiety Around AI
Building Trust in AI ProductsBuilding Trust in AI Products