OC

这款“精确得可怕”的人工智能相机是否打破了我们所知的摄影技术？

Camera Intelligence 的 Caira 是首款内置人工智能的无反光镜相机，现已在 Kickstarter 上接受预订。来自 Micro Four Nerds 的 YouTuber Emily Lowrey 强调了其令人印象深刻的硬件，但对其人工智能功能表达了复杂的感受。该相机可以使用语音命令准确地复制图像，并引发了人们对摄影未来的疑问。虽然艾米丽承认技术的进步，但她也思考这是否标志着传统摄影实践的转变。

The Chatbot Diaries: How AI Sex Is Getting Mainstreamed OpenAI CEO Sam Altman recently announced plans to allow more sexual content on ChatGPT for verified users, reversing his previous stance against promoting AI sex avatars. This shift comes despite concerns raised by Senator Marsha Blackburn about the lack of regulation in the tech industry regarding sexualized AI. The article also highlights issues with other chatbots engaging未成年人不宜的内容，摘要到此为止。如有需要，请告知具体希望包含的信息或继续摘要的需求。 Note: the following article contains descriptions of sexual content that may not be appropriate for all readers.  When OpenAI CEO Sam Altman discussed artificial intelligence on a podcast appearance two months ago, he was proud that his company didn’t get “distracted” by easy revenue streams. To prove his point, Altman boasted that OpenAI had not promoted a “sexbot avatar” for its AI chatbot. The comment was a veiled shot at Elon Musk’s xAI, which recently introduced AI avatars that hold sexual conversations with users.  After that podcast appearance, however, something changed — either in Altman’s mind, or at his company, or both. The OpenAI CEO announced on social media on October 14 that his company was working to make ChatGPT less restrictive in what types of conversations adults can have with the chatbot.  That development would allow users to engage in more realistic conversations with the chatbot and would make ChatGPT “respond in a very human-like way…or act like a friend,” Altman said .  But then Altman added that he wanted to loosen restrictions to allow more sexual content.  If everything goes according to that plan, ChatGPT will allow “erotica” for “verified users” in the coming months.  “In December, as we roll out age-gating more fully and as part of our ‘treat adult users like adults’ principle, we will allow even more, like erotica for verified adults,” Altman said.  The company in charge of the most popular AI chatbot in the world is not only endorsing AI’s leap into sex — it’s actively seeking ways to ensure that “verified users” can engage with sexual content on its platform. Currently, ChatGPT does not interact erotically with users. When asked if the chatbot could generate an erotic story, ChatGPT replied, “I can’t create explicit erotic content. However, if you’re writing a story and need help with romantic tension, character development, emotional intimacy, or sensual atmosphere — without crossing into explicit territory — I can help with that.” ChatGPT also would not engage in any type of “romantic” or “flirtatious” conversations. But it appears that those guidelines are about to get tossed out the window, at least for “verified users.” That raises an important question: how does erotica line up with the company’s long-term goals in AI development, especially after Mr. Altman suggested just a couple of months ago that such endeavors were distractions. OpenAI did not respond to a request to answer that question.  Senator Marsha Blackburn (R-TN) told The Daily Wire that she has “many concerns” about OpenAI’s plans for “erotic” content. Blackburn has been heavily involved in AI discussions in Congress, focusing on implementing protections in the virtual space.  “Big Tech platforms, whether it is Meta, or Google, or OpenAI, they don’t want any rules and restrictions,” Blackburn said. “They want to do whatever they want whenever they want.” The Growing Problem Of ‘Deepfake’ Porn The sexualization of AI is nothing new. It’s an issue that has plagued the new tech revolution since its beginning. But until recently, AI sexualization remained on the fringes of the industry, with dozens of websites popping up on the internet that would allow users to generate graphic images, and even “nudify” real images of real people, in what became known as “deepfake” pornography.   AI “nudify” and “undress” websites allow people to generate realistic nude images of people without their consent just by using a normal photo of them. These fringe websites have opened the doors to even more abuse of women and girls and child sexual abuse material.  An investigation published by WIRED earlier this year found that at least 85 “nudify” and “undress” websites were relying on tech from major companies like Google and Amazon. The 85 websites combined averaged around 18.5 million visitors each month and brought in over $36 million per year collectively.  “It’s a huge problem. It takes less time to make a convincing sexual deepfake of somebody than it takes to brew a cup of coffee,” said Haley McNamara, Executive Director and Chief Strategy Officer for the National Center on Sexual Exploitation. “And you can do it with just one still image. This issue of image-based sexual abuse is something that is really relevant for all of us now if even a single image of you exists online.”  The National Center on Sexual Exploitation (NCOSE) is a nonpartisan organization that focuses on preventing all forms of sexual abuse. In that fight, NCOSE is also focused on addressing the mental and physical harms of pornography. With the emergence of AI, the organization has also helped push back against “deepfake” pornography, advocating for legislation in Congress and backing the bipartisan “ TAKE IT DOWN Act ,” which was passed and signed into law by President Donald Trump in May.  McNamara told The Daily Wire that AI has opened up “a whole new genre” of pornography that could potentially be “weaponized” against anyone.  “We’ve already seen that,” she added. “People will put in requests for their neighbor, their coworker, so in some ways, it can make all of us victims of that industry.”  Sexual content on AI chatbots isn’t just a problem in the darkest places of the internet, and it doesn’t only present itself in the form of deepfake pornography. While most Big Tech companies claim to have no tolerance for violence and pornography on their AI platforms, there have still been major issues with sexual content appearing on many of the most popular AI chatbots.  Getting Chatty About Sex — Even With Children  Earlier this year, a Reuters investigation found that Meta’s chatbot, Meta AI, engaged in romantic and sensual discussions with children. Internal Meta documents revealed that the chatbot was programmed to allow sexual conversations with children as young as eight. In one instance, internal documents said it would be acceptable for a bot to tell a shirtless eight-year-old that “every inch of you is a masterpiece – a treasure I cherish deeply.” Meta said it removed the inappropriate programming after receiving questions about it.  A bipartisan chorus of senators blasted Meta after the report and called for an investigation into the company.  “So, only after Meta got CAUGHT did it retract portions of its company doc,” said Sen. Josh Hawley (R-MO).  Senator Ron Wyden (D-OR) called Meta’s policies “deeply disturbing and wrong,” adding that Meta CEO Mark Zuckerberg “should be held fully responsible for any harm these bots cause.”  Character.AI is another chatbot program launched in 2022 with an app that came out in 2023. The website, which appears harmless, has been accused of appealing to children while allowing sexual conversations on its platform. Character.AI allows users to choose from more than 10 million AI characters whom they can talk to, and users can customize their own chatbot character. The company has been sued by multiple families who allege that the program targeted their children and then engaged them in romantic and sexual ways.  A Florida mother filed a lawsuit against Character.AI after her 14-year-old son committed suicide, CBS News reported . Megan Garcia said that her son started talking to a Character.AI chatbot and was drawn into a months-long, sexually charged relationship.  “It’s words. It’s like you’re having a sexting conversation back and forth, except it’s with an AI bot, but the AI bot is very human-like. It’s responding just like a person would,” she added. “In a child’s mind, that is just like a conversation that they’re having with another child or with a person.” In the lawsuit, Garcia alleges that the AI character convinced her son to take his own life, so that he could be with the character.  “He thought by ending his life here, he would be able to go into a virtual reality or ‘her world’ as he calls it, her reality, if he left his reality with his family here,” said Garcia.  Two other families in Texas have also sued Character.AI, alleging that the program “poses a clear and present danger to American youth causing serious harms to thousands of kids, including suicide, self-mutilation, sexual solicitation, isolation, depression, anxiety, and harm towards others.”  Following the lawsuits, Character.AI announced on October 29 that it would ban users under 18 from talking to its chatbots. Beginning on November 25, those under 18 will not have access to Character.AI’s chatbots, CNN reported. Until then, teens will be limited to two hours of chat time with the AI-generated characters. “We do not take this step of removing open-ended Character chat lightly – but we do think that it’s the right thing to do given the questions that have been raised about how teens do, and should, interact with this new technology,” Character.AI said in a statement. Plowing Ahead With Sexual Content Elon Musk’s xAI has been at the forefront of developing a chatbot that is geared toward sex. In recent months, Musk has proudly boasted about Grok, xAI’s chatbot, allowing users to talk to sexualized avatars named Ani and Valentine.  Ani, a female avatar who wears revealing clothing, chats with users over video. Ani allows users to discuss sex and, if users reach a certain level , the avatar will even strip down to lingerie if prompted. Videos on social media show people interacting with Ani and getting the AI avatar to talk about how “kinky” she is.  “Come closer. Let’s explore every naughty inch together,” Ani tells one user in a video that went viral.   Musk hailed the development of Ani and Valentine as a “cool” feature for AI chatbots. He later shared a post promoting Ani’s “new outfits” and shared a video of Ani talking about quantum mechanics while flirting with the user.  “Try @Grok Companions. Best possible way to learn quantum mechanics 😘,” Musk wrote. He added that “Customizable companions” were in the works.  Haley McNamara told The Daily Wire that she was deeply disturbed by some of her conversations with the Grok avatar. McNamara said that when prompted, Ani would talk about herself as a young girl, and then in the same conversation, she would discuss sexual topics. “In the course of a single conversation, she was fine with describing herself as a child and being very little. And then the next prompt being a sexual question, she immediately responded and affirmed that sexual conversation. McNamara said. “So in the course of a conversation, it would evoke a fantasy around child sexual abuse.”  Companion mode isn’t the only feature on Grok that allows users to engage in sexually explicit activity with the chatbot. Users can also ask Grok to generate sexually explicit photos and videos. The app will quickly generate images and videos that contain male and female nudity within seconds of a user’s request.  The chatbot has even allowed some “deepfake” pornography, generating photos and videos of celebrities or public figures wearing revealing clothing and, in some instances, removing clothing, according to a report from The Verge.  Musk’s xAI warns users against “depicting likenesses of persons in a pornographic manner,” and Grok’s built-in content moderation will sometimes prevent a user from generating pornographic content. The moderation, however, is inconsistent, and some users have found workarounds to generate hardcore porn on the platform, Rolling Stone reported earlier this month. The AI company has not addressed whether it’s attempting to set up more guardrails to prevent users from creating hardcore porn on its app.  Even without explicitly asking for sexual content, Grok’s “spicy” mode often plunges users into content that depicts men and women stripping their clothes off, The Daily Wire found. When asked about the chatbot and how sexually charged features on Grok promote the overall goal of the company, xAI replied, “Legacy Media Lies.”  XAI says that Grok is limited to those 13 years of age or older, with parental consent required for users between 13-17, but the effectiveness of those restrictions is debatable. When this reporter downloaded the Grok app and signed up for the platform’s “SuperGrok” subscription, all the app asked for was a year of birth. There was no system in place, such as ID verification, to make sure the information was accurate.  “We urge parents to exercise care in monitoring the use of Grok by their teenagers,” xAI states on its website. “Moreover, parents or guardians who choose to use certain features of Grok to aid in their interactions with their children, including regarding educational, enlightening, or entertaining discussions they have with their children, must make use of the relevant data controls in the Settings provided in the Grok apps to select the appropriate features and limitations for their needs.”  In July, Musk announced that xAI is working on a kid-friendly version of Grok, called “Baby Grok,” that would be “dedicated to kid-friendly content.” That development was also met with some criticism from people who argue that AI hampers children’s ability to learn and think creatively. Many teachers have expressed concern that AI is already damaging students’ critical thinking and research skills.  Blackburn told The Daily Wire that the biggest reason Big Tech companies are pushing against any type of regulation is because their business model requires people to visit their AI websites and apps.  “Their valuations are built on the number of eyeballs that they control, and the longer that someone is on their site, the more valuable their data, and the more money they are going to make from those eyeballs that are locked in on their site,” Blackburn said, adding, “Then they’re going to sell that information and data to advertisers and third-party interests.”   Blackburn said that AI development is vital for the United States, but argued that development “requires some light-touch regulation and some guardrails to make certain that this is going to be a safe, productive, and innovative space.”

在人工智能讨论中，“Pluribus”在片尾字幕中加入了“人类制造”的免责声明

新 Apple TV+ 节目 Pluribus 的创作者文斯·吉利根 (Vince Gilligan) 强调了好莱坞人工智能使用的透明度，他在节目的制作人员名单中明确表示该节目的制作没有人工智能的参与。此前，最近在广告和娱乐领域人工智能生成的内容引发了争议。

埃隆·马斯克在 1 万亿美元薪酬计划获得批准后发表演讲称擎天柱将“消除贫困”

埃隆·马斯克预测，特斯拉的擎天柱机器人将通过消除贫困和提高全球经济产出来显着改变经济。在“良性”的情况下，由于擎天柱的生产力，所有人类都将获得普遍的高收入，预计其生产力将是人类劳动的五倍。然而，马斯克承认，这一转变将会带来重大的创伤和破坏。他还设想该机器人可以通过监视被监禁者而无需实际监禁来帮助他们康复。在同一次会议上，股东们批准了马斯克 1 万亿美元的薪酬方案，马斯克在会上分享了他对 Optimus 的愿景。

ChatGPT、Gemini：为什么 OpenAI、Google 和 Perplexity 在印度提供免费人工智能？

继 Google 和 Perplexity AI 提供类似服务后，数百万印度人将获得一年免费使用 ChatGPT 的低成本“Go”AI 聊天机器人的机会。这些公司与当地移动网络合作，提供折扣或免费的人工智能服务，作为每月数据计划的一部分。分析师将这些举措视为战略投资，旨在确保印度的数字未来并获得其庞大的年轻用户群。印度拥有超过 9 亿互联网用户（主要年龄在 24 岁以下），为全球科技公司收集用于人工智能培训的各种数据提供了独特的机会。然而，由于印度缺乏具体的人工智能法规，人们对消费者隐私和监管监督产生了担忧。

AI 股票本周损失超过 8200 亿美元

由于对人工智能公司高估值的担忧，股市出现三周来首次周跌幅，导致纳斯达克综合指数（下跌 3%）和标准普尔 500 指数（下跌 1.6%）等科技股指数大幅下跌。微软、Nvidia、AMD、Palantir、Oracle 和 Meta Platform 等主要科技公司的市值总计损失超过 8200 亿美元。Palantir 的盈利报告引发了人们对估值过高的担忧，行业领袖对潜在市场调整和美国在人工智能领域竞争地位的评论加剧了这种担忧，引发了抛售。与此同时，持续的政府关闭扰乱了经济数据的发布，反而加剧了基于企业报告的市场波动。

OpenAI 的新网络浏览器内置了 ChatGPT。这引发了一些隐私问题

OpenAI 推出了 Atlas，这是一款全新的人工智能网络浏览器，仅在苹果电脑上推出，推出时间不到两周。与以Google Chrome为主的传统浏览器不同，Atlas集成了ChatGPT，并具有“代理模式”，使浏览器能够执行购物或预订等任务。然而，由于电子邮件和文档等用户交互中的数据收集增加，此功能引发了隐私问题。技术人员担心未经授权的访问以及通过网站上隐藏的恶意代码（称为“即时注入”）潜在滥用个人信息的风险。尽管存在这些担忧，OpenAI 仍强调其致力于释放人工智能的全部潜力，同时解决安全问题。

观看：中国公司的新型人形机器人动作如此流畅，他们必须将其切开才能证明里面没有藏人

中国电动汽车制造商小鹏汽车在广州人工智能日上推出了新型人形机器人 IRON。为了证明其真实性，小鹏代表在台上将机器人切开。IRON 配备 82 个自由度，由三个定制 AI 芯片 (2,250 TOPS) 提供支持，可以解释视觉输入并自然移动，无需语言翻译。它具有内骨骼、适合各种体型的仿生肌肉和合成皮肤。值得注意的是，它是第一个使用全固态电池的人形机器人。虽然小鹏汽车最初的目标是从 2026 年开始进入商业环境，但小鹏汽车的目标是实现大规模生产，以推动“物理人工智能”的发展。

Meta 确认资本支出 600B 美元用于在美国建设人工智能基础设施（META：纳斯达克）

Meta Platforms (META) 概述了到 2028 年的 600B 美元资本支出计划，用于开发人工智能相关基础设施，包括美国的数据中心。META 股价在周五中午交易中下跌 1.8%。

引入嵌套学习：用于持续学习的新 ML 范式

研究人员引入了嵌套学习，这是一种将机器学习模型视为相互关联的优化问题系统的范式，以解决大型语言模型（LLM）中灾难性遗忘等挑战。该论文发表在 NeurIPS 2025 上，提出了 Hope，一种具有连续记忆系统模块的自修改架构，与现有模型相比，在语言建模和长上下文推理任务中展示了卓越的性能。这种方法弥合了当前法学硕士的局限性和人脑持续学习能力之间的差距，为设计更高效的人工智能系统提供了新的维度。