Character.AI 公测

2022 年秋天，当大多数对话式 AI 还在强调「助手」身份时，一个叫 Character.AI 的产品悄悄上线了。它的创始人 Noam Shazeer 是 Transformer 论文作者之一，Daniel De Freitas 也来自谷歌。这两个人没有做另一个通用聊天机器人，而是做了一个让用户「和角色聊天」的平台——虚构人物、历史名人，或者自己捏一个角色，都能聊。

产品逻辑简单得近乎直觉：既然大模型已经学会「扮演」某种人设，为什么不把这种能力做成产品？用户可以创建角色的性格、背景与说话方式，然后和它进行开放的、不受任务约束的多轮对话。上线后的增长快得惊人，数月内就积累了庞大的用户群，成为 2022-2023 年间最热门的消费级 AI 应用之一。

Character.AI 的走红，折射出大模型价值判断的一次转变。在此之前，行业讨论的焦点是「AI 能帮你做什么」——写代码、答问题、提效率。而 Character.AI 证明，大量用户真正想要的是「AI 能陪我说什么」——一种情感上的、娱乐性的、非功利的关系。AI 伴侣从科幻概念变成了真实的大众需求。

这个转变带来的连锁影响至今仍在延伸：AI 陪伴、虚拟角色、情感计算成为独立赛道，也催生了关于成瘾、依赖与虚拟关系伦理的大量讨论。无论褒贬，它都证明了一件事——大模型的能力里，藏着人类对「被理解、被陪伴」的渴望，而这种渴望本身就是一门巨大的生意。

回看 Character.AI 的公测，它的意义不止于一个成功产品，而在于它第一次让「情感」成为 AI 消费的核心变量。当后来者蜂拥进入 AI 伴侣赛道时，2022 年 9 月那个让用户和虚构角色聊天的小平台，早已把答案写在了前面。

In the autumn of 2022, while most conversational AI products still insisted on the "assistant" identity, a product called Character.AI quietly launched. Its founder Noam Shazeer was one of the authors of the Transformer paper; Daniel De Freitas also came from Google. Rather than build another generic chatbot, these two made a platform where users "chat with characters"—fictional figures, historical names, or characters they invented themselves.

The product logic was almost intuitive: if large models had learned to "play" personas, why not make that ability into a product? Users could craft a character's personality, background, and way of speaking, then hold open-ended, task-free conversations with it. Growth was startlingly fast. Within months it had amassed a huge user base and become one of the hottest consumer AI apps of 2022-2023.

Character.AI's rise reflects a shift in how the industry valued large models. Until then, the conversation centered on "what AI can do for you"—code, answers, efficiency. Character.AI proved that a huge number of users actually wanted "what AI can say with me"—an emotional, entertaining, non-utilitarian relationship. AI companions went from science fiction to mass demand.

The ripple effects still extend today: AI companionship, virtual characters, and affective computing became standalone sectors, along with extensive debates about addiction, dependency, and the ethics of virtual relationships. Whatever the verdict, it proved one thing—hidden inside large-model capability is a human hunger for being understood and accompanied, and that hunger is itself a vast business.

Looking back at Character.AI's public launch, its significance goes beyond a successful product. It was the first time "emotion" became a core variable in AI consumption. When later entrants flooded the AI companion space, the answer had long been written by that small platform in September 2022 that let users chat with fictional characters.