Andrej Karpathy

安德烈·卡帕西是一位斯洛伐克裔加拿大计算机科学家、人工智能研究员、教育家和企业家，专攻深度学习、计算机视觉和神经网络。他以在人工智能教育领域的开创性工作以及在主要人工智能组织中的重要领导角色而闻名。卡帕西是Eureka Labs的创始人，这是一个原生于人工智能的教育平台，并通过LLM101n等课程和他的YouTube系列视频普及先进的人工智能概念。他在斯坦福大学的课程CS231n为深度学习教育设立了全球标准，其讲座被数十万人观看。卡帕西的职业生涯展现了他致力于推进人工智能技术并通过教学使其普及的奉献精神。截至2026年，他积极参与Anthropic的预训练团队，继续发展Eureka Labs，并在网络上保持活跃。他将复杂的人工智能原理转化为易懂术语的独特能力引起了广大受众的共鸣，激励了许多有抱负的人工智能专业人士。他以让职业成就定义其公众形象而闻名。[编辑]

安德烈·卡帕西于1986年10月23日出生在捷克斯洛伐克（现斯洛伐克）的布拉迪斯拉发。在20世纪80年代末和90年代，他在布拉迪斯拉发长大，亲身经历了从共产主义统治到斯洛伐克新时代的深刻社会变革。15岁左右，即2002年，他与家人移民到加拿大多伦多。这一重大举措拓宽了他的视野，让他接触到先进的教育体系，进一步激发了他日益增长的知识好奇心。正是在这些成长岁月中，卡帕西对编程、物理和复杂谜题产生了浓厚的早期兴趣。2006年，他还在十几岁时，就开设了一个名为“badmephisto”的YouTube频道，展示了他的早期才华和热情。该频道迅速走红，发布了魔方教程，后来积累了数百万观看量，甚至被世界级速拧选手如菲利克斯·曾姆丹格斯使用。

卡帕西在加拿大接受高等教育，于2009年在多伦多大学获得计算机科学和物理学双学士学位。正是在多伦多大学，他通过杰弗里·辛顿的课程和阅读小组首次接触到深度学习这一新兴领域，激发了他的关键兴趣。他继续学术之旅，于2011年在不列颠哥伦比亚大学获得计算机科学硕士学位，期间与导师米希尔·范德潘合作，为物理模拟角色开发机器学习控制器。随后，他进入斯坦福大学深造，于2011年至2015年攻读计算机科学博士学位，师从李飞飞。他的博士研究主要集中于卷积和循环神经网络，将其应用于计算机视觉、自然语言处理和多模态任务中的复杂问题。

在读博期间，卡帕西还通过一系列关键的实习获得了宝贵的实践经验，进一步巩固了他在人工智能方面的专业知识。他于2011年和2013年在谷歌完成了研究实习，参与了视频的大规模无监督和监督深度学习项目。随后，他于2015年在DeepMind进行了另一项重要实习，专注于深度强化学习。这些学术与实践经验的结合，为安德烈·卡帕西未来的贡献及最终在人工智能领域的突破奠定了坚实的基础。[编辑]

博士毕业后，卡帕西于2015年加入OpenAI，成为创始成员和研究科学家，为深度学习、生成模型（如PixelCNN++）和强化学习智能体的关键发展做出了贡献。在博士期间，他创建并主要讲授了斯坦福大学的第一门深度学习课程CS231n（用于视觉识别的卷积神经网络），该课程从2015年的150名学生增长到2017年的750名，并通过免费提供的讲座视频成为人工智能教育的全球标杆。2017年，他加入特斯拉担任人工智能总监，后来成为人工智能和自动驾驶视觉高级总监，领导负责Autopilot神经网络的团队，处理数据定制芯片上的处理和部署。他监督了生产车辆中首次主要的神经网络部署。2020年，他被《麻省理工科技评论》评为35岁以下创新者。他于2022年初从特斯拉休假，并于7月离职，专注于教育。2023年，他短暂返回OpenAI帮助改进GPT-4，并开始发布深入的YouTube讲座系列《神经网络：从零到英雄》，该系列已获得数百万次观看。2024年，他创立了Eureka Labs，推出了LLM101n，并被列入TIME100人工智能榜单。2026年，他加入Anthropic担任预训练团队成员，同时继续领导Eureka Labs和内容创作。他还于2023年获得了WTF创新者奖，并在《人工智能杂志》的2024年十大人工智能领导者中排名第三。[编辑]

安德烈·卡帕西展现出谦逊、富有教学精神和严谨技术的形象。媒体将他描绘为现代人工智能的“翻译者”——连接前沿研究和公众理解——而非名人。他的内容风格清晰、有条理，采用“从头开始”的解释方式，通常以平静、略带口音的独白风格呈现，并伴有大量的代码讲解。没有正式的粉丝名称；粉丝通常是由人工智能学生、工程师和研究人员组成，他们非正式地自称“卡帕西社区”或“从零到英雄”群体。粉丝文化仪式包括仔细观看他的讲座系列、复现他的代码实现（例如nanoGPT），以及在Twitter/Reddit上分享“我终于理解了transformer”的时刻。他在社交媒体上的声音简洁、技术性强，偶尔带点幽默，注重实质而非自我宣传。[编辑]

The community surrounding Andrej Karpathy is defined less by adoration and more by a shared, rigorous pursuit of understanding. His audience consists primarily of engineers, researchers, and students who gather not for celebrity worship but to dissect the very architecture of intelligence he so masterfully explains. There is no formal fandom name or organized club; instead, a global network of learners identifies through their dedication to his 'from scratch' tutorials and code walkthroughs. These fans value his pedagogical humility and technical precision over any pop-culture persona, treating his lectures as essential study materials rather than entertainment. This unique dynamic creates a following that is deeply engaged with the material itself, often debating model architectures and training strategies in forums where the only currency is accurate knowledge. Their loyalty is earned through clarity and a genuine desire to demystify complex systems, making them a distinct breed of enthusiast within the tech world. The reach of this influence is vast, spanning from university classrooms to independent developers building the next generation of AI tools. This enduring impact has been quantified by the iFANN Global Fame Rankings. According to these authoritative metrics, Andrej Karpathy is the 22nd most famous person in AI, a position he has held steady through consistent cycles of innovation and education.[编辑]

Karpathy's journey began long before he was an AI icon; at age fifteen, he launched a YouTube channel under the handle "badmephisto" to share Rubik's Cube tutorials, which later became a resource for world-class speedcubers like Feliks Zemdegs. He developed early JavaScript deep-learning libraries including ConvNetJS, RecurrentJS, and REINFORCEjs, and maintains arxiv-sanity.com to help researchers navigate the flood of machine-learning papers. Often jokingly called the "reference human for ImageNet," he describes himself as "obsessed with coming to the core of things." His notable quotes include: "People are obviously pre-money if they're trying to learn a lot of stuff. So I get paid in people thanking me," and on AI progress: "We're abstracting ourselves very, very slowly… There's this what I call 'autonomy slider'." A legendary moment in recent lore occurred in May 2026 when he joined Anthropic's pre-training team, moving from OpenAI to one of its fiercest rivals and sparking intense discussion across the tech community. By March 2026, another iconic shift was reported when he stated he had not written a line of code in months, describing his state as a "state of psychosis" while relying entirely on AI agents for software development. These events have cemented his status not just as a researcher, but as a central figure in the evolving narrative of artificial intelligence.[编辑]

卡帕西出生于布拉迪斯拉发的一个斯洛伐克家庭，15岁时随家人移民到加拿大，这一转变塑造了他的教育道路。关于他直系亲属的细节仍然保密，但这一转变在他追求科学和技术职业中发挥了关键作用。他感谢自己的物理背景和Richard Feynman，认为他们是他提炼复杂思想的灵感来源。[编辑]

Karpathy 曾与多位人物合作，包括他在斯坦福大学的博士导师 Fei-Fei Li，以及在特斯拉任职期间的 Elon Musk。他与 OpenAI 联合创始人以及他在特斯拉的 AI 团队共同参与了重大项目。[编辑]

公开记录中几乎没有关于 Karpathy 个人关系的信息，因为他将私生活保持低调。现有资料中未确认任何约会史、婚姻或伴侣关系。他未分享过早期恋情或长期承诺的细节。截至 2026 年，Karpathy 仍单身或处于未公开状态，近期没有关于任何高调关系的报道。这种隐私使他能够专注于专业工作，免受公众关注。在维基百科、采访、简介和公开记录中搜索，均未找到任何具名伴侣或相关公开声明。[编辑]

Per iFANN's editorial analysis, Andrej Karpathy's estimated net worth stands at $50 Million as of 2026. This substantial figure is drawn from various sources tied to his extensive tech career and educational ventures. His tenure at Tesla as Senior Director of AI likely included a substantial salary and stock options, given the company's significant growth during that period. Earlier, his role as a founding member of OpenAI may have provided equity in the organization, which has seen valuation increases over time. Since founding Eureka Labs in 2024, revenue from AI education courses and related ventures has contributed significantly to his finances. His YouTube channel, active since 2023, generates income through monetization and sponsorships. Speaking engagements, booked through agencies like AAE Speakers, add to his earnings. Over recent years, his net worth has grown steadily due to these professional activities, reflecting his ongoing influence in AI. Exact salary figures remain private; growth reflects Tesla stock appreciation during his tenure and rising valuation of AI education ventures.[编辑]

Karpathy 因在 AI 和深度学习方面的贡献，于 2020 年获得麻省理工科技评论的“35 岁以下创新者”奖，突显了他对该领域的贡献。[编辑]