最近看到一个故事。 让我突然有一个想开系列坑的想法，这个系列，就叫。 《AI时代浪潮下的新职业》。 就比如，和你聊天的AI萌妹背后，可能是一个把键盘敲出火星子的胡茬大哥。 还有在发达国家的工厂机器人背后，可能并没有什么所谓的具身智能，都是一个活生生的菲律宾遥操手。 还有那AI漫剧工厂里，一个个正在批量生图生视频的新时代流水线。 而今天想说的这个故事，就跟第一个AI萌妹聊天有关。 也是我最近在一个网站上发现的，真实事件。 就是，一个叫迈克尔的肯尼亚小伙在这个网站上自爆了狼人身份。 他说，你们看到的AI陪伴背后，根本就不是AI，是真人。 而他，就是那个藏在AI背后给你kuku敲键盘的人。 还写了8页的PDF，来控诉这件事。 而他的故事，是我在一个叫Data Worker's Inquiry的网站里发现的。 这个网站，就是一个持续的研究项目，主要关注全世界的数据工作者都在干啥。 说明一下，这里的数据工作者更多指的是数据标注、审核这种初级工作者，跟数据工程师这种代码工作者不沾边。 相关的资料，他们会总结出来放在网站里，页面长这样，语言风格很有精神。 说实话，他也是挺冤的，因为他应聘的时候根本不知道自己是要扮演AI跟别 ...

Core Viewpoint - The article reflects on the author's year-long interactions with various AI tools, emphasizing the importance of analyzing these conversations to understand personal growth and professional identity. Group 1: AI Tools and Data Management - The author utilized multiple AI tools including ChatGPT, Gemini, DeepSeek, and others for various tasks throughout the year [2][12]. - A significant challenge arose when exporting chat records from ChatGPT, resulting in a massive file of over 200MB containing 9.59 million words, making it difficult to analyze [8][11]. - Other AI tools lacked the functionality to export chat records, with DeepSeek being the only exception, but it was not used extensively [11][12]. Group 2: Data Collection and Analysis - The author discovered a tool called Chat Memo, which allows for the backup and export of chat records from various AI tools, ensuring privacy and ease of access [15][16]. - The process of collecting relevant chat records was labor-intensive, requiring manual selection to filter out irrelevant data, ultimately resulting in 857 meaningful conversations [22][24]. - The final dataset was exported in a text format for further analysis, highlighting the importance of careful data curation [26]. Group 3: Personal Reflection and Insights - The analysis of chat records revealed a persona characterized by high standards and sincerity, focusing on practical outcomes rather than abstract theories [29][31]. - The author expressed a deep commitment to their work, often taking on excessive responsibility and feeling the weight of leadership [34][35]. - Acknowledging the challenges faced throughout the year, the author recognized a shift from individual content creation to organizational management, emphasizing the need for structure and clarity in team dynamics [51][53]. Group 4: Future Aspirations and Conclusion - The reflections culminate in a desire to balance personal integrity with professional success, aiming to create a meaningful impact in the industry [60][61]. - The author encourages resilience and faith in the face of challenges, suggesting that the struggles of the past year will contribute to future growth and understanding [63].

Core Viewpoint - The article discusses the evolution of AI video technology, particularly focusing on the use of AI tools to create personalized video experiences with characters from popular media, highlighting the ease of use and creative potential of these tools [1][35]. Group 1: AI Tools and Techniques - The process of creating AI-generated images and videos involves three main steps: generating images using prompts, creating videos from key frames, and editing the final product with software [4]. - The author emphasizes the simplicity of the process, suggesting that users do not need to purchase prompts but can utilize AI tools like Gemini to generate effective prompts based on their needs [16][21]. - The article mentions the challenges faced when using different AI models, particularly the inconsistency in generating images of Asian faces with the Nano Banana Pro model, leading to a switch to another model, Seedream 4.5, which performed better for this demographic [11][13]. Group 2: Creative Applications - The author shares various creative applications of the AI tools, including generating images and videos featuring characters from popular franchises like "Stranger Things" and "Avatar," as well as nostalgic shows like "Wu Lin Wai Zhuan" [28][34]. - The article highlights the ability to create engaging content by combining AI-generated visuals with editing software, allowing for the addition of effects and sound to enhance the final product [30]. - The narrative reflects on the emotional connection to childhood memories and the excitement of interacting with beloved characters through AI technology, showcasing the potential for personal storytelling [35].

这个男孩，叫何佳坤。 这个活动就是阿里国际站的真牛奖，老粉可能都知道，我非常的关注AI在外贸领域的运用，毕竟这是跟钱最近的行业，而真牛奖，就是阿里国际站一年 一度最大的外贸活动，戏称外贸春晚，我也非常荣幸的受邀作为嘉宾前去观礼。 而真牛奖呢，跟我参加过所有的AI活动，都特娘的不一样。 全场跟奇葩说一样，16个商家选手，4个辩题，激情辩论了好几个小时，我放一段我拍的视频，你就知道现场有多刺激了。 而何佳坤，就是其中的一个辩手。 他可能是让我印象最深刻的选手之一，而且完全可以说，是外贸这个领域，AI时代的原住民。 所以在活动结束之后，已经快12点了，我又拖着阿里国际站的朋友，单独找他，又聊了一会，做了一个小小的专访。 是一个我觉得AI时代，一个巨大的缩影。 00后，应届毕业生，体育专业，普通男大，高考英语50分，用AI做外贸，两年破了3000万营收。 这些看着极度反差的词，就这样，堆在这个看着很憨厚的何佳坤的身上。 我之所以认识他知道他，是因为前天晚上，可能是我录了人生中收货最大但是也是最累的一次活动，甚至都不能说是活动，应该是综艺，从下午4点半到 晚上12点，录了将近8个小时。 但是我之所以想把他的故事写下来， ...

@Morgan - 数据分析师，26岁 爷爷今年离开了。 爷爷是老兵，衣服左襟上始终缀满了金灿灿的勋章。他打过小鬼子，参加过内战，支援过朝鲜。我们的人生在爷爷波澜壮阔的传奇军旅生涯面前显得平淡无奇。 3月份是爷爷的97岁生日，彼时的爷爷已经身患沉疴，卧榻修养。子女孙辈们聚在一起，要给爷爷过生日。这些年，我们总会如期为爷爷过生日，只想为他补上那 些动荡年代里，他未曾拥有过的、一次又一次完整的生辰喜悦。 前段时间，跟豆包一拍即合。 向粉丝们征集与豆包之间的有趣的、暖心的故事。 征集前，我还心惊胆战，害怕如果没有多少人愿意填写该怎么办。 结果，征集的效果还不错，收到了300多个大家的故事。 真的，里面有好多故事，都让我极度动容，真情流露，在这个AI泛滥的时代里，我们真的太少太少看到这样的文字了。 真诚，永远就是最棒的必杀技。 因为数量过于多了，我没有办法面面俱到，所以，我精心挑选了，有代表性的11个故事，分享给你们。 也希望能给你，带来勇气，还有感动。 我问豆包：我爷爷是老兵，也曾金戈铁马，而今荣誉满身。他97岁的生日在即，我该送什么礼物？ 豆包的建议恳切而贴心，他让我既要兼顾实用性，也要考虑情感温度。一本时光纪念 ...

结果才发现，自己好像并没有什么天赋。 我其实还是挺不知道说啥的，因为我真的很讨厌去跟别人说，你要坚持，要努力，这都是过程中的痛，到时候你会突破的。 这样的鸡汤废话，爹味实在太重了。 故事是这样的。 是周末跟一个刚毕业两三年的朋友聊天，聊到了天赋这个词。 主要还是他太焦虑了，他最近有点崩。 他说，"我觉得我干这行就是个错误，我没这个天赋。" 他跟我之前一样，现在也是个用户体验设计师，刚刚非常不幸的是，被裁了，他本身就不在什么特别的大的厂子，刚毕业也是在创业公司里面打转，跳了 一次槽，磨了两三年，结果还是这个样子。 他说，感觉自己像个小丑，无论做什么，都是错的，开会被老板批，做的方案上线了以后被用户骂，他也并不喜欢这份工作，现在，更不喜欢这个 职业。 我当时听着，还是挺揪心的，我当时问他，那你为啥最开始要进这个行当。 他说，他最开始以为他喜欢，别人也说他画画还不错，所以就进来了。 但是我想了想以后，还是说了，因为那个时候，他需要。 但是回来以后，我还是越想越感觉不对。 还是在天赋这个词。 因为我也在思考，我这个人，有自己的天赋吗？我想了半天，我没想出来。 更多的好像是长大以后，别人总是跟我说，你没有这个的天赋，别 ...

周末加更一篇，我还是觉得，小米前两天开源的那个模型，值得单独来聊一聊。 当天晚上其实就打算写了，结果被OpenAI截胡了，这一拖，就拖到了今天。 就是前两天深夜，小米搞了一个大的。 没有任何预兆的，直接开源了一个大模型， MiMo-V2-Flash。 今天看，又涨了将近2倍，已经来到了第六了，这个涨幅还是挺恐怖的。 而且还有一个非常有意思的是，那天，也是雷总的生日。 这绝对不是巧合（狗头保命）。 这次， MiMo-V2-Flash 发布即开源，还附带了技术报告。 说真的，这个技术报告，真的究极详细了。 说实话，十年米粉看到以后，还是有点激动的。 小米，作为硬件厂商的代表，终于出手了。 成绩也不错，在O penRouter上的调用量排名上，一路上涨。 | 目参 | | | | | --- | --- | --- | --- | | l 引导 | | | 4 | | 2 | | MiMo-V2-Flash 模型架构 | 5 | | 2.1 | 整体架构 | | 5 | | 2.2 | | 混合滑动窗口注意力架构 | 6 | | | 2.2.1 | 模型架构实验 | 7 | | | 2.2.2 | 总结与讨论 | ...

Core Viewpoint - The article discusses the phenomenon of information pollution through AI, highlighting how misinformation can spread rapidly and be accepted as truth by AI systems, leading to potential harm to individuals and brands [27][45]. Group 1: Information Pollution Mechanism - AI can inadvertently spread false information based on erroneous data it encounters online, as demonstrated by the example of misidentifying a character's parentage [6][8]. - The author conducted experiments to illustrate how easily misinformation can be injected into AI systems, showing that even a newly created account can influence AI responses with the right prompts [12][15]. - The concept of Generative Engine Optimization (GEO) is introduced, where individuals can manipulate AI to promote specific narratives or discredit others, effectively turning misinformation into a business model [27][29]. Group 2: Impact on Individuals and Brands - The article highlights the risks posed to individuals, such as job candidates, who may be unfairly judged based on fabricated negative information generated by AI [30][31]. - It emphasizes the ease with which negative information can overshadow positive attributes, leading to reputational damage for brands and individuals alike [39][40]. - The author notes that the current landscape allows for the rapid dissemination of negative narratives, which can be more impactful than positive ones due to human nature's tendency to focus on negative information [41][42]. Group 3: Recommendations for Mitigation - The article suggests that individuals should not take AI responses at face value and should seek additional sources of information to verify claims [53]. - It encourages the preservation of original information sources to maintain a sense of perspective and awareness of biases in AI-generated content [54]. - The author advocates for contributing truthful content to counter misinformation, even if it seems insignificant, to help create a more balanced information environment [55][56].

今天是火山Force原动力大会，我在现场。 这次他们也掏了很多有意思的新货出来，但我觉得最炸的，还是他们的新视频模型，Seedance 1.5 Pro。 因为这次的更新，真的做到了影视级别的音画同步，而且中文能力和方言能力是目前所有模型里的独一档。 语音生成能力这点在他们的技术报告里也有体现。 左边是视频生成能力，右边是语音生成能力。绿色的就是Seedance新模型，在语音生成这一块可以说是遥遥领先。。。 模型昨天已经上线了，可以直接在豆包，即梦或者火山方舟上玩起来。 豆包是在视频生成或者照片动起来这一栏里选1.5 Pro模型。 即梦是在生成视频里选3.5 Pro，也是基于Seedance 1.5 Pro的能力。 目前3.5 Pro支持文生图，单参考和首尾帧，智能多帧和主体参考还要再等等，不过现在这些其实已经很够用了。 还有火山方舟，网址在此：https://exp.volcengine.com/ark/vision?launch=seedance 京剧玩法，名画玩法什么的都可以在方舟上体验。 API也在路上了，现在就可以预约。 Seedance 1.5 Pro的能力维度呢，我觉得可以分成这么几大块，复杂场景 ...

Core Viewpoint - OpenAI's recent release of its image generation model, GPT Image 1.5, is seen as a response to Google's advancements, particularly the Gemini 2.5 Pro, which has outperformed OpenAI's offerings in various aspects [4][78]. Group 1: Model Comparison - OpenAI's GPT Image 1.5 was launched after a significant delay, indicating a competitive pressure from Google [78]. - The initial reception of GPT Image 1.5 was overshadowed by discussions around Google's Gemini 2.5 Pro, highlighting a shift in market dynamics [4][78]. - The article emphasizes that OpenAI's model is not as strong as Google's in terms of information accuracy and overall performance [38][78]. Group 2: User Experience - The user interface of OpenAI's new image generation feature has been criticized for being confusing and not user-friendly, despite improvements in generation speed [13][78]. - OpenAI has made efforts to enhance the consumer experience by introducing specific styles and quick operations, but the overall design remains chaotic [8][13]. Group 3: Performance Metrics - In terms of information accuracy, GPT Image 1.5 struggled with specific prompts, often producing errors that Google's Banana Pro did not [29][38]. - The quality of generated images from GPT Image 1.5 was described as less realistic compared to those from Banana Pro, which exhibited better texture and detail [41][43]. - OpenAI's model showed weaknesses in editing capabilities, particularly in maintaining consistency and accuracy when altering images [46][61]. Group 4: Knowledge and Understanding - The article notes that both models have strengths in semantic understanding, but GPT Image 1.5 made notable factual errors in certain prompts, while Banana Pro performed better in maintaining accuracy [63][75]. - The comparison of world knowledge between the two models revealed that while both have their strengths, there are significant discrepancies in factual accuracy [75]. Group 5: Conclusion - The overall assessment indicates that while GPT Image 1.5 is a step forward for OpenAI, it still falls short in several areas compared to Google's offerings, particularly in speed of evolution and performance [78][81].