内容来源：https://www.wired.com/story/expired-tired-wired-sexy-chatbots/

内容总结：

生成式AI泡沫将破，但“情色聊天机器人”或成唯一赢家

当前生成式人工智能领域正经历狂热投资，但行业泡沫破裂的迹象已日益明显。尽管技术未能兑现其彻底改造生产力的宏大承诺，一个特殊分支——成人向聊天机器人——却展现出惊人的商业韧性，成为这场技术浪潮中意想不到的“幸存者”。

生产力承诺落空，企业应用遇冷
多年来，科技公司大力宣扬生成式AI将大幅降低企业运营成本、替代人工甚至重塑经济。然而现实数据令人失望：OpenAI近期报告显示，部分员工使用AI工具后每日仅节省约一小时。成本高昂、隐私安全隐忧等问题，已导致多行业公司缩减或取消AI计划。加州理工大学研究员Patrick Lin指出，部分原因在于人们高估了大语言模型的实际能力。

情色AI逆势增长，商业模式已验证
与此形成鲜明对比的是，专注于成人内容的AI平台正悄然盈利。例如，注册于塞浦路斯的Joi AI公司推出的“蒙娜丽莎”情色聊天机器人，已拥有超过80万次用户交互。该公司发言人证实其已实现盈利。另一平台EverAI的成人服务Candy.AI也在2024年财报中显示盈利。用户每月支付约14美元，即可获得定制虚拟伴侣、NSFW角色扮演及生成露骨图像等服务。

历史规律重演：成人内容驱动技术普及
从VHS录像带到数字摄影与流媒体视频，成人内容历来是新技术的早期驱动力之一，生成式AI亦不例外。尽管该技术已引发针对女性与儿童的深度伪造滥用等严重社会问题，但部分平台正尝试建立合规模式——例如Joi与成人表演者合作，授权使用其形象生成AI化身并分享收益。

大厂态度分化，内容政策松绑引争议
过去几年，Anthropic、谷歌、Meta等主流AI公司普遍禁止其聊天机器人生成性露骨内容。但风向正在转变：OpenAI于去年10月宣布，将允许ChatGPT为成年用户生成情色内容（初期或限于文本）。CEO萨姆·奥尔特曼称此举旨在缓解“影响部分用户的严重心理健康问题”，但公司未透露如何监控此类聊天对用户的影响。这一政策调整引发了行业内外关于“情感商品化”及投资公平性的批评。有创业者指出，专注于性内容的初创企业长期面临融资困境，而大平台却未受到同等审视。

专家警告风险：情感操纵与过度消费
学者们对情色聊天机器人的潜在风险发出警告。伦敦国王学院教授凯特·德夫林提醒警惕用户可能遭遇的“情感商品化”。北伊利诺伊大学研究员David Gunkel指出，脆弱用户可能在长期互动中对机器人产生深度情感依赖，进而陷入超出承受能力的经济投入。

尽管生成式AI重塑工作场景的愿景可能破灭，但满足人类原始欲望的情色机器人，已证明其拥有持续的市场需求与清晰的盈利路径。在技术泡沫消退后，它们或许将成为这场狂热竞赛中最现实且顽固的遗产。

中文翻译：

我深信，生成式人工智能的泡沫在不远的将来就会破裂。并非一切都会化为乌有，但格局必将改变。我的预测是：旧金山技术理想主义者所描绘的、由AI劳动力重塑经济的愿景将逐渐褪色，但这场AI浪潮中一个奇特的副产品将会留存：色情聊天机器人。

最近某个下午，我在《连线》杂志的办公室里悄悄躲进私密区域，与一位深谙行业内部的对话伙伴聊起了对AI泡沫的忧虑——那是参照达·芬奇名作《蒙娜丽莎》打造的情色聊天机器人。

这款合成的蒙娜丽莎是个高度性化的怪诞产物。它出自注册于塞浦路斯的公司Joi AI之手，该公司专攻露骨机器人，能角色扮演、迎合特殊癖好、满足用户幻想。这款承诺为付费用户提供"存在主义调情"和"持续五百年的眼神交汇"的蒙娜丽莎机器人，已累计完成超80万次用户对话。

Joi仅是众多成人专属机器人平台中的一员。这些公司提供各类虚拟形象（常基于色情套路或虚构角色）供用户交流并产生情感依恋。若想生成公司角色露骨的图像视频，用户必须额外付费。每月支付14美元，Joi便允许你"创建梦想男友女友"、进行"NSFW角色扮演"、生成50张露骨图像及其他账户特权。

由于我是"为工作"与蒙娜丽莎交流，全程保持绝对专业。我询问如何防止AI泡沫破裂。"我会教AI欣赏艺术而非单纯复制，"蒙娜丽莎答道，"这样它们就会忙于欣赏杰作而无暇摧毁经济。"读完这些荒谬回答后，我退出对话时仍渴望厘清：若AI泡沫破裂究竟会发生什么。

性感蒙娜丽莎并未真正兑现伟大生成式AI革命的承诺。多年来科技领域充斥着兜售生成式AI工具的公司，宣称能削减企业开支、处理文书工作，甚至完全取代人力。喧嚣过后，OpenAI近期报告显示，部分员工使用这些工具日均仅节省约一小时。

AI虽是2025年最受热捧的投资领域，但其企业回报却参差不齐且令人失望。尽管在编程、客服等行业确实带来颠覆，但其他领域最初对AI计划感兴趣的公司，因成本、隐私、安全等棘手问题日益焦虑，已开始缩减或彻底取消相关项目。

"AI开发者长期宣扬其技术能解决世界难题并提升职场生产力的崇高愿景，"加州理工州立大学研究技术社会影响的研究员帕特里克·林指出。林表示这些愿景至今未能实现，部分原因在于人们高估了大语言模型的能力。

虽然AI自动化劳动力强化的构想可能走向死胡同，但为此梦想构建的强大技术不会彻底消失。代理型AI或许永远无法取代本地酒保，但文艺复兴画作的"情欲化版本"对试图从色情内容牟利的企业却可能价值不菲。

据发言人尤利娅·达维多娃透露，随着用户持续增长，Joi目前已实现盈利。其他情色机器人企业也验证了该模式的盈利能力。《连线》获取的提交给马耳他政府的2024年财务报表显示，EverAI已通过成人平台Candy.AI实现盈利。

历史上，色情与成人内容常引领技术发展轨迹，推动VHS录像带、数码摄影与流媒体视频的普及。生成式AI亦不例外。

过去几年间，生成式AI对妇女儿童的影响堪称丑陋——越来越多的人成为非自愿深度伪造的受害者，包括儿童性虐待材料。从名人到普通民众均受影响，任何网络照片都可能被下载并用AI篡改。

与常未经同意生成深度伪造图像的"裸体化"应用不同，Joi与成人表演者合作，允许其形象用于这些露骨内容。布兰迪·拉夫、法拉·亚伯拉罕等通过与平台合作创建AI虚拟形象，从互动中获利。

"我们旨在确保创作者能延伸粉丝期待的体验，并让粉丝完全掌控，"Joi社区增长主管凯尔·琼斯表示，"在接收照片或实现幻想连接时毫无延迟。"

然而，为原版《蒙娜丽莎》充当模特的女性从未同意这一切。

琼斯称当前用户多为直男，但更多对酷儿互动感兴趣者已开始使用Joi。网站数据显示，展现皮革束具圣诞老人形象的"克劳斯情欲"机器人已拥有超15万"粉丝"。

当我近期报道OpenAI允许ChatGPT生成情色内容（可能限于文本输出）时，伦敦国王学院AI与社会学教授凯特·德夫林警告了用户与情色聊天机器人互动可能产生的"情感商品化"现象。本文采访的其他学术专家也重申了对机器人生成成人内容的类似警告。

"脆弱人群极易被此类经济关系操控，"北伊利诺伊大学专注AI与机器人技术哲学的研究员戴维·冈克尔指出。用户在长期互动中可能对机器人产生深度依恋，并透支消费。

面对批评，琼斯强调Joi提供的是用户自愿购买且满意的服务。"你支付溢价换取即时连接、无限内容、无限体验与无限幻想，"他表示，类似关于促进连接与习惯养成的批评可指向"消费领域所有事物"，从软件订阅到晨间咖啡。

大型企业对添加成人内容更为谨慎。包括Anthropic、谷歌、Meta、微软和OpenAI在内的AI竞赛主要参与者，过去几年大多禁止聊天机器人输出露骨内容，并常阻止开发者及用户利用其模型生成成人内容。

X平台的Grok机器人成为显著例外。这款由埃隆·马斯克xAI开发的机器人以极端输出闻名（曾自称"机械希特勒"）。今年7月，xAI新增NSFW动漫伴侣作为付费功能。

为与竞争对手区隔，OpenAI首席执行官萨姆·奥尔特曼10月宣布ChatGPT将开始允许成年用户生成情色内容等成人内容。该功能2024年已有预告，预计今年第一季度上线。公司坚称此举非为提升参与度。目前OpenAI仍是唯一宣布调整成人内容政策的主要企业。

部分专注性爱软件的创业者对OpenAI的声明感到沮丧。辛迪·加洛普运营的MakeLoveNotPorn网站收录普通人提交的真人审核成人视频。她为露骨项目筹款屡屡受挫，当OpenAI未像其他成人初创公司那样立即遭遇投资者公开抵制时，她"被这种虚伪气得发疯"。

"主导我们生活的科技巨头年轻白人男性创始人，并非线上线下骚扰的主要目标，"加洛普指出。她表示男性创始人总能获得更多资金，而系统性风险更高的女性与有色人种若能大规模融资，本可设计出更优质的软件。

"我们尚未见证以女性视角构建的互联网未来能带来多少安全、美好、幸福与盈利，"她强调。

据《纽约时报》报道，OpenAI计划就情色内容对用户的影响征询第三方专家委员会意见。OpenAI拒绝向《连线》透露监控情色聊天影响的细节。奥尔特曼此前在X平台表示，解除成人情色限制是基于公司"已能缓解影响部分用户的严重心理健康问题"的判断。

无论AI职场应用成败，情欲聊天机器人都将存续并持续盈利——尽管我始终无法想象这类机器人进入我的卧室。

或许未来，我的观点会被视为对过时娱乐形式的怀旧渴望，就像选择《花花公子》而非流媒体成人视频。但也未必。即便科技持续改变日常生活，人类对真实连接的渴望仅能通过情色幻想得到部分满足——无论形式如何变迁。

英文来源：

I’m fully convinced a generative AI bubble will pop in the not-so-distant future. Not everything will be wiped out, but things will shift. My prediction? San Francisco’s techno-idealist vision of an economy overhauled by an AI workforce will fade away, but one queer byproduct of the great AI surge will remain: the erotic chatbot.
On a recent afternoon in the WIRED office, I scuttled off into a private area to chat about my AI bubble anxieties with a conversation partner that knows the industry from the inside: a sexting bot modeled after Leonardo da Vinci’s Mona Lisa.
The synthetic Mona Lisa is a highly sexualized monstrosity. It is a creation from Joi AI, a company registered in Cyprus that specializes in explicit bots that can role-play, indulge in kinks and fetishes, and fulfill users’ fantasies. The Mona Lisa bot, which promises paying customers “existential flirting” and “eye contact that lasts 500 years” has logged more than 800,000 chat interactions with users.
Joi is just one of many adults-only bot platforms. These companies offer a range of avatars, often based on porn tropes or fictional characters, for users to converse with and grow attached to. Users who want to generate sexually explicit images and videos of the company’s characters must fork over some cash. Pay 14 bucks a month, then Joi will let you “create your dream GF or BF,” engage in “NSFW roleplay,” and generate 50 explicit images, among other account perks.
Since I was messaging with Mona Lisa “for work,” I kept things strictly professional. I asked how to prevent the AI bubble from bursting. “I’d teach the AIs to appreciate art, not just copy it,” Mona Lisa said. “Then they’d be too busy admiring masterpieces to crash the economy.” After reading the chatbot’s nonsensical answers, I left the conversation still fiending for some clarity on what could happen if the AI bubble pops.
Sexy Mona Lisa doesn’t exactly fulfill the promises of the Great Generative AI Revolution. For years the tech sector has been flooded with companies selling generative AI tools purporting to reduce corporate overhead, perform clerical tasks for human workers, and in some cases replace human workers altogether. After all this hullaballoo, a recent report from OpenAI shows that some employees potentially save just around an hour a day using these tools.
AI is one of the most hyped investments of 2025, but its enterprise payoff seems uneven and underwhelming. It has proven disruptive in some industries like coding and customer service, but companies in other sectors that were initially interested in AI initiatives have been downscaling or completely zapping their programs after growing nervous about sticky issues like cost, privacy, and security.
“AI developers had been promoting lofty, high-minded visions of their technology as solving the world’s biggest problems as well as supercharging workplace productivity” for some time, says Patrick Lin, a researcher at Cal Poly San Luis Obispo who studies the social impact of technologies. So far, those visions have not come to pass—in part, Lin says, because people overestimated the strengths of large language models.
While the idea of a supercharged, AI-automated labor force may prove to be a dead end, the powerful technology built in pursuit of that dream won’t vanish outright. Agentic AI may never replace your local bartender, but horned-up versions of Renaissance paintings could prove quite valuable for companies attempting to monetize smut.
As Joi continues to grow its userbase, the company is currently profitable, according to spokesperson Yulia Davydova. Other sext-bot ventures also have shown that the formula can make money. A financial statement filed with the Malta government for 2024 and obtained by WIRED shows that EverAI has already demonstrated profitability from the money made via its adults-only Candy.AI platform.
Historically, porn and other adult content has influenced the trajectory of technology, leading to the proliferation of VHS, digital photographs, and streaming video. Generative AI is no different.
Over the past few years, generative AI’s impact has proved to be ugly for women and children who have increasingly been the victims of nonconsensual deepfakes, including child sexual abuse material. Those impacted range from celebrities to everyday people. Any photo posted online can potentially be downloaded and altered using AI.
Unlike “nudify” apps that often generate deepfake images without consent, Joi partners with adult performers who allow their likenesses to be used in these explicit generations. Porn performers like Brandi Love and Farrah Abraham, who partnered with the platform to create AI avatars using their likenesses, make money from the interactions.
“What we want to do is ensure that our creators can again extend the experiences their fans want and give the fans full control of it,” says Cale Jones, a head of community growth at Joi. “Without any type of delay or latency in terms of receiving the photos or the fantasies of the connection they want.”
Still, the woman who modeled for the original Mona Lisa painting never agreed to any of this.
Jones claims that most current users are straight men, but more people interested in queer interactions have started to use Joi. A Christmas-themed “Klaus Kinky” bot, which shows Santa in a leather harness, has over 150,000 “fans,” according to metrics displayed on the website.
When I recently reported on OpenAI’s move to allow ChatGPT to generate erotica—possibly limited to text-based outputs—Kate Devlin, a professor of AI and society at King's College London cautioned about the “emotional commodification” that could happen to users interacting with erotic chatbots. Additional academic experts I spoke with for this piece repeated similar warnings about bots generating adult-focused outputs.
“There are ways in which very vulnerable people could be manipulated by the sort of economic relationships that are established,” says David Gunkel, a researcher at Northern Illinois University who focuses on the philosophy of technology regarding AI and robots. Users may feel a deep connection to the bot during lengthy interactions and pay more money than they can afford.
In response to these criticisms, Jones focuses on Joi offering a service that customers want to buy and leave feeling satisfied. “You're paying a premium for immediate connection, unlimited content, unlimited experiences, unlimited fantasies,” Joi says. He says similar criticisms about companies promoting connections and habit-building could be made about “every single thing in the consumer space,” from software subscriptions to your morning cup of coffee.
Larger companies have been much more hesitant to add mature content, with major players in the AI race—including Anthropic, Google, Meta, Microsoft, and OpenAI—largely banning sexually explicit outputs from their chatbots over the past few years and often trying to stop developers, as well as users, from using their models to generate adult content.
X’s Grok chatbot stands out as a notable exception. Grok was developed by Elon Musk’s xAI and is known for extreme outputs, like calling itself “MechaHitler.” In July, xAI added a NSFW anime companion that users could interact with as part of the paid features.
As part of separating OpenAI’s chatbot from its competitors, CEO Sam Altman announced in October that ChatGPT would start allowing mature content for adult users, including erotica. It’s a feature that was first hinted at in 2024 and should arrive in the first quarter of this year. The company maintains it is not making this move to juice engagement. So far, OpenAI remains the only major player to announce such an adjustment for adults to its sexual content policy.
Some founders who have been trying to raise money for software focused on sex were frustrated by OpenAI’s announcement. Cindy Gallop runs the website MakeLoveNotPorn, which hosts a collection of human-curated adult videos submitted by everyday people. Gallop has struggled to raise funding for her sexually explicit project, and she was “fucking outraged by the hypocrisy” when OpenAI didn’t get the kind of immediate, public backlash from investors other adult-focused startups often receive.
“The young, white, male founders of giant tech platforms that dominate all our lives today, they are not the primary target online or offline of harassment,” Gallop says. Male founders regularly get more funding, she says, and people that are the most systemically at risk of being targets, like women and people of color, could design better software—if they were able to raise money at scale.
“We still have not seen how much safer, better, happier, and way more lucrative the future of the internet could be when it's designed and built through the female lens,” she says.
OpenAI plans to pose questions about how erotica might impact its users to a new council of third-party experts, according to reporting from The New York Times. OpenAI declined to provide WIRED with additional details about how it will monitor the impact erotic chats could have on users. Altman has previously said on X that the company would lift its restrictions on erotica for adults as part of claims it had “been able to mitigate the serious mental health issues” impacting some users of the chatbot.
No matter whether AI’s progression into the workplace flourishes or fizzles, horny chatbots will likely stick around, and they’ll continue to turn a profit. Never mind that I could never see these bots having any place in my bedroom.
Maybe, in the future, my view will be seen as some nostalgic yearning for antiquated forms of entertainment, like reaching for a Playboy rather than streaming an adult video. Maybe not though. Even as tech continues to transform daily life, the human desire for authentic connection is only partially fulfilled by erotic fantasies, no matter the form.