OpenAI 联合创始人 Sam Altman 和 Greg Brockman 讨论他们的关系、公司早期经历，以及如何共同应对戏剧性事件和成功。
OpenAI co-founders Sam Altman and Greg Brockman discuss their relationship, the early days of the company, and how they've navigated drama and success together.
要点 · TL;DR
推出优秀产品是展示 AI 价值的关键。 Shipping great products is key to showing AI's value.
AI 将普及专家级服务，如超人医疗建议。 AI will democratize expertise like superhuman medical advice.
通过真实世界反馈的迭代部署比秘密开发更安全。 Iterative deployment with real-world feedback is safer than secret development.
核心观点 · Key points
最重要的是推出优秀产品，让人们直接感受 AI 的价值。 The most important thing is to ship great products so people can feel AI's value directly.
AI 将大幅提升所有人的底线，例如通过智能手机免费获得超人类医疗建议。 AI will raise the floor for everyone, e.g., free superhuman medical advice via smartphone.
通过实际使用进行迭代部署和安全验证，优于秘密开发。 Iterative deployment and safety through real-world use is better than secret development.
未来是了解你的背景并代表你行动的个人 AGI。 The future is personal AGI that knows your context and acts on your behalf.
算力是利润中心而非成本中心；更多算力创造更多价值。 Compute is a profit center, not a cost center; more compute enables more value.
反共识 · Contrarian takes
美国在硬件和机器人领域远远落后；只有 AI 驱动的机器人才能追赶。 The US is far behind in hardware and robotics; only AI-driven robots can catch up.
关于危险模型的恐惧营销往往是控制 AI 访问权限的策略。 Fear-based marketing about dangerous models is often a tactic to control AI access.
OpenAI 早期的 AGI 路径是通过竞争性多智能体模拟，而非语言模型。 OpenAI's early AGI path was via competitive multi-agent simulations, not language models.
与埃隆·马斯克的诉讼是讲述 OpenAI 真实故事的机会。 The lawsuit with Elon Musk is an opportunity to tell OpenAI's true story.
写作质量难以通过强化学习评判；个性化将使 AI 为每个用户成为优秀写作者。 Writing quality is hard to judge via RL; personalization will make AI a great writer for each user.
本期章节 · Chapters（共 34）
开场与介绍Opening and Introduction
算力雄心与长期愿景Compute Ambition and Long-Term Vision
安全沟通分歧Disagreement on Safety Communication
AI 韧性与迭代部署AI Resilience and Iterative Deployment
安全沟通演变的反思Reflections on Safety Communication Evolution
对 AI 与人类目的的恐惧Fears about AI and human purpose
赞助商插播：BrexSponsor break: Brex
公众认知差距与直觉 AIPublic awareness gap and intuitive AI
AI 的顿悟时刻Aha moments with AI
AI 未来的兴奋与不确定性Excitement and Uncertainty about AI's Future
主持人的怀疑与更新Host's Skepticism and Update
写作质量与技术视角Writing Quality and Technical Perspective
奖励信号与个性化挑战Challenges in Reward Signal and Personalization
确保 AGI 惠及全人类Ensuring AGI Benefits All of Humanity
世界的三种未来Three futures of the world
美国硬件劣势对华US hardware disadvantage vs China
向智能体转型与核心使命Transition to Agents and Core Mission
各垂直领域迎难而上Rising to the Moment Across Verticals
模型是产品的一部分，而非产品本身Models as Part of the Product, Not the Product Itself
重点领域：智能体平台、计算机工作、个人 AGIFocus Areas: Agentic Platform, Computer Work, Personal AGI
消费与企业界限模糊及降级项目Consumer vs Enterprise Blurring and Deprioritized Projects
Sora 与算力分配Sora and compute allocation
超级应用愿景与身份基础设施Super App Vision and Identity Infrastructure
执行对比：OpenAI vs AnthropicExecution Comparison: OpenAI vs Anthropic
强大模型与恐惧营销Powerful Models and Fear-Based Marketing
神话与网络安全框架Mythos and Cybersecurity Frameworks
政府压力与公平性Government Pressure and Fairness
政府关系与国家安全Government Relations and National Security