Jared Kaplan

Jared Daniel Kaplan 是一位理论物理学家和人工智能研究员，他塑造了现代人工智能的发展。作为约翰霍普金斯大学的副教授以及 Anthropic 的联合创始人兼首席科学官，他连接了基础物理学与前沿机器学习。Kaplan 关于神经语言模型缩放定律的研究改变了行业处理 AI 训练的方式，表明性能随着计算量、数据和模型规模的增加而可预测地提升。

Kaplan 在 2020 年共同撰写了里程碑式的 GPT-3 论文，引入了一个 1750 亿参数的模型，引发了一波大型语言模型的采用浪潮。在 Anthropic，他开创了宪法式 AI（Constitutional AI），这是一种通过自我监督而非大量人类反馈来使模型与人类价值观对齐的技术。他还担任 Anthropic 的负责任扩展官，负责制定发布强大 AI 系统的安全协议。

在转向 AI 之前，Kaplan 在理论物理学领域工作了 15 年，研究量子引力、全息原理和共形场论。他的学术道路包括赫兹奖学金、斯隆研究奖学金和美国国家科学基金会职业奖。Kaplan 将深刻的物理学洞察与 AI 专业知识独特结合，使他成为通向人类级人工智能道路上的领先声音。[编辑]

Jared Kaplan 从小对理论物理学有浓厚兴趣，这促使他在斯坦福大学攻读物理学和数学学士学位。他于 21 世纪初毕业，随后在哈佛大学获得物理学博士学位，其论文《全息原理的若干方面》由 Nima Arkani-Hamed 指导。

大约在 2009 年完成博士学业后，Kaplan 成为斯坦福直线加速器中心和斯坦福大学的博士后研究员，专注于量子引力、AdS/CFT 对应和宇宙学。这些岁月建立了他对复杂理论系统的深厚专业知识，后来证明对理解神经网络非常宝贵。他于 2012 年加入约翰霍普金斯大学担任副教授，教授深度学习基础，同时继续他的物理学研究。[编辑]

Kaplan 的突破性进展出现在 2020 年，他发表了《神经语言模型的缩放定律》，该研究表明语言模型的性能随着计算量、数据集大小和模型参数的增加而可预测且平滑地提升。这一见解从根本上改变了公司分配 AI 训练资源的方式，使扩展成为该领域的核心策略。

同年，他共同撰写了《语言模型是少样本学习者》，引入了 GPT-3，一个拥有 1750 亿参数的模型，在各种任务中实现了显著的少样本性能。GPT-3 成为一个商业里程碑，为 ChatGPT 等产品提供动力，并推动了大型语言模型的广泛采用。在 Anthropic（他于 2021 年与 Dario Amodei 共同创立），Kaplan 开发了宪法式 AI，一种使用一套原则指导模型行为的技术，减少了对人类反馈的依赖。2024 年 10 月，他被任命为负责任扩展官，负责前沿 AI 系统的发布前安全评估。[编辑]

Jared Kaplan 在媒体中被描绘为一位严肃、低调的科学家，他连接了严谨的理论物理学与务实的 AI 开发。报道强调他的智慧、对 AI 风险的谨慎态度以及基础性贡献，而非个人魅力。他在论文、演讲和政策文件中的语气是技术性、精确且注重安全的。没有组织化的粉丝群体、官方粉丝名称或粉丝文化仪式；他主要在 AI 研究和有效利他主义社区中受到尊重。他没有可识别的标志性美学风格或社交媒体声音，且没有维护任何公开的社交媒体账户。[编辑]

Jared Kaplan currently holds no placement in the iFANN Global Top 1,000 rankings or any scope-specific list, a status confirmed by the August 2026 evaluation cycle. While his influence reshaped the trajectory of artificial intelligence, he remains outside the published fame hierarchies that track celebrity culture and mass-market consumption signals. The absence of a formal ranking reflects his position as a technical authority rather than a pop-culture icon; his reach is defined by deep professional impact within physics and machine learning communities rather than broad public adoration. Consequently, there is no documented fandom name for his supporters, and no organized group refers to themselves with a collective moniker. His audience consists primarily of physicists, AI researchers, and students who engage with his work through academic channels like lectures and peer-reviewed papers rather than social media campaigns or fan rituals. Unlike entertainers who inspire chart pushes or charity drives, Kaplan's following interacts with him strictly through the lens of scientific discourse and industry development. There are no records of fan-funded projects, subway advertisements, or viral cultural myths surrounding his person. The stories told about him are not folklore but verified accounts of his warnings regarding existential risks and his strategic pivots at Anthropic. These moments serve as professional lore for those in the field, marking critical decision points in AI safety rather than whimsical legends associated with athletes or musicians.[编辑]

Jared Kaplan co-founded Anthropic alongside Dario Amodei, a fellow Hertz Fellow and former OpenAI VP, bridging their shared background in theoretical physics with pragmatic AI development. His seminal scaling laws paper has accumulated over 10,000 citations as of 2026, cementing its status as a foundational text in the field. He resides in Pacifica, California, a coastal town beloved by surfers, where he maintains an exceptionally private life away from the spotlight. Before his transition to artificial intelligence, Kaplan dedicated fifteen years to theoretical physics, focusing on quantum gravity and holography. In recent years, his warnings about the 2030 decision point have become a recurring theme in industry discourse, framing the era between 2027 and 2030 as a critical window for aligning AI systems before they can train themselves without constraint. This stance, often cited as the "ultimate risk," has evolved into a defining narrative for the AI safety movement. Additionally, the mid-2026 release of the model named "Claude Mythos" marked a turning point in security protocols, though the title refers to the system's capabilities rather than a legend about Kaplan himself. His contributions to GPT-3 remain a cornerstone of modern large language model adoption, introducing a 175-billion-parameter architecture that sparked a global wave of LLM development.[编辑]

Kaplan 有一个儿子，其姓名和出生细节未公开披露。他居住在加利福尼亚州帕西菲卡。他保持低调的个人形象，公开资料中未提供关于其父母或兄弟姐妹的信息。[编辑]

Kaplan 经常与他的 Anthropic 联合创始人 Dario Amodei 合作，并与 OpenAI 团队在 GPT-3 和 Codex 上密切合作。他的博士导师是著名理论物理学家 Nima Arkani-Hamed。[编辑]

Jared Kaplan 没有公开的约会历史或已确认的关系。包括维基百科、福布斯和 Anthropic 官方简介在内的可靠来源未提及任何浪漫伴侣、婚姻或离婚。他保持低调的个人形象，不公开讨论私生活。截至 2026 年，Kaplan 被报道为单身，无确认伴侣。[编辑]

Heading into 2026, Jared Kaplan's net worth is estimated at $3.7 Billion according to iFANN's editorial analysis. The vast majority of this wealth stems from his equity stake in Anthropic, which reached a valuation between $18 and $20 billion following multiple funding rounds. Kaplan co-founded the company in 2021 and serves as its chief science officer, driving its strategic direction. His previous tenure at OpenAI was solely as a researcher and did not include an equity component, contributing minimally to his financial standing. He also earns a salary as an associate professor at Johns Hopkins University, typically ranging from $200,000 to $300,000 annually for that role. No information is available regarding real estate holdings, endorsements, or other ventures beyond these primary income sources. Kaplan's net worth has surged rapidly since Anthropic's founding, mirroring the soaring valuations seen across the artificial intelligence sector.[编辑]

Kaplan获得了多项著名学术奖项。他于2005年获得赫兹奖学金用于物理学研究生学习，2010年代获得斯隆研究奖学金用于早期职业研究，2015年因AdS/CFT和量子引力方面的工作获得NSF CAREER奖。[编辑]