以下文章来源于Vista氢商业 ，作者何愚 Vista氢商业 . 关心都市白领的消费生活 提供易懂有趣的品牌报道 本文来自微信公众号： Vista氢商业 ，编辑：卢力麟，作者：何愚，原文标题：《时薪千元挖大厂前 员工教会AI后一脚踹开，"多少数据公司正在把高级牛马当饲料榨干"》，题图来自：AI生成 从咨询公司普华永道离职后，邵逸凡收到了一封看起来几乎完美的兼职工作邀请邮件，时薪高达120 美元，居家办公，时间自由。 工作内容是"给AI出题"——但不是为了让它答对，而是为了让它答错，然后再手把手教会它，如何 像年薪百万的"高级牛马"们一样做咨询。 被AI取代之前，打工人正在牺牲 自己供养AI？ 问题到底出在了哪里？ 邵逸凡很快发现，要想真正难倒AI，光靠编数据是远远不够的。你不能抄袭，也不能使用网上的公 开案例，这一切都在暗示着，你必须把自己真正做过的项目喂给它。平台甚至会直接询问："你是否 愿意在抹掉客户信息的情况下，提供你以前做过的项目报告和模型？"但这些项目资料，是她和团队 伙伴一起熬了好几个月完成的，又是客户花了几百万买断的。 最关键的是， 看似体面的时薪，更像是一笔对你过往经验和知识的"一次性买断费" ...

以下文章来源于镜相工作室 ，作者镜相作者 AI焦虑已经从大厂渗透进了大学这座象牙塔。在一片"被替代"的忧惧声中，一群来自非北上广深、 非985高校的大学生，主动跳进了这股浪潮——他们不是来做实习的，是来兼职"拧螺丝"的，拧的是 AI大模型的螺丝。 在"AI将替代99%的岗位"和"拿AI offer月入3万"两种声音的共同影响下，这批不容易挤进大厂实习队 伍的大学生，走进了大厂数据标注的流水线，成为新的AI工人。他们每天在大厂提供的兼职平台上 抢单、标注、纠错、优化，月收入大多在1000到2000元之间。在这条看不见的AI流水线上，他们既 是训练者，也是被挑选的对象。 他们分散在成都、郑州、武汉、厦门等非一线城市的角落，在课余或下班后的时间里，登录那个决定 他们今晚"有活干"还是"白蹲守"的后台，他们标注图片，校准对话，贡献乡音，优化代码，试图找到 新的机会。 对于大厂来说，大学生群体兼具垂直领域的专业知识和充足的参与热情，是完成最基础的AI数据标 注、AI回复纠错、AI声音识别等最合适的人群。通过向大学生伸出兼职橄榄枝，他们不仅能高效率 地找到符合大模型训练要求的人，还能减轻雇佣正职员工的成本。 镜相工作室 ...

以下文章来源于镜相工作室 ，作者镜相作者 镜相工作室 . 商业世界的风向与人 来源丨 镜相工作室 （ID： shangyejingxiang ） 作者丨 马舒叶 编辑丨 卢枕 图源丨Midjourney AI焦虑已经从大厂渗透进了大学这座象牙塔。在一片"被替代"的忧惧声中，一群来自非北上广深、非 985高校的大学生，主动跳进了这股浪潮——他们不是来做实习的，是来兼职"拧螺丝"的，拧的是AI 大模型的螺丝。 在"AI将替代99%的岗位"和"拿AI offer月入3万"两种声音的共同影响下，这批不容易挤进大厂实习 队伍的大学生，走进了大厂数据标注的流水线，成为新的AI工人。他们每天在大厂提供的兼职平台上 抢单、标注、纠错、优化，月收入大多在1000到2000元之间。在这条看不见的AI流水线上，他们既 是训练者，也是被挑选的对象。 他们分散在成都、郑州、武汉、厦门等非一线城市的角落，在课余或下班后的时间里，登录那个决定 他们今晚"有活干"还是"白蹲守"的后台，他们标注图片，校准对话，贡献乡音，优化代码，试图找到 新的机会。 对于大厂来说，大学生群体兼具垂直领域的专业知识和充足的参与热情，是完成最基础的AI数据标 注 ...

Core Viewpoint - The article discusses the emergence of university students engaging in AI-related part-time jobs, particularly in data labeling and model training, as a response to the dual pressures of job market anxiety and the allure of high-paying AI positions. This trend highlights the evolving landscape of employment in the AI era, where traditional roles are being replaced by new opportunities that require different skill sets [4][25]. Group 1: AI Part-Time Jobs - A group of university students from non-first-tier cities is participating in AI data labeling tasks, earning between 1,000 to 2,000 yuan per month [4][5]. - These students are seen as ideal candidates for basic AI data tasks due to their specialized knowledge and enthusiasm, allowing companies to reduce costs associated with hiring full-time employees [5][6]. - The competition for these part-time jobs has intensified, with students needing to "抢单" (grab orders) quickly to secure tasks [6][28]. Group 2: Job Experience and Career Impact - Many students view these part-time roles as a way to enhance their resumes and gain relevant experience in the AI field, which is increasingly important for securing internships and job offers [20][22]. - The experience gained from these roles is being leveraged in interviews, with students able to discuss their contributions to AI model training and data labeling processes [21][23]. - The article notes that students are aware of the precarious nature of their roles, feeling like "饲料" (feed) for AI models, yet they continue to pursue these opportunities for future career prospects [29]. Group 3: Market Dynamics and Employment Trends - The article highlights a significant shift in the job market, with traditional roles being reduced while AI-related positions are on the rise, as evidenced by a 36% increase in AI job postings in the first half of the year [26]. - Major tech companies are increasingly relying on AI, leading to substantial layoffs in traditional roles, with reports of up to 90% of certain teams being cut [27]. - The competitive nature of AI-related part-time jobs is expected to grow, as more students enter the market seeking similar opportunities [28].

Core Insights - The rise of AI anxiety has permeated universities, leading students from non-top-tier schools to engage in part-time jobs related to AI model training, often referred to as "screwing the bolts" of AI models [1][2][19] - These students, primarily from second-tier cities, are participating in data labeling and optimization tasks, earning between 1,000 to 2,000 yuan per month, while also seeking opportunities to enhance their resumes and gain relevant experience in the AI field [1][14][19] Group 1: Student Engagement in AI Tasks - Students are actively participating in AI data labeling through platforms like "Xpert," where they can earn between 50 to 200 yuan per task, requiring no prior AI experience [3][6] - The competition for these part-time jobs has intensified, with students needing to "抢单" (grab orders) quickly, similar to securing tickets during peak travel seasons [2][4] - The nature of the tasks includes evaluating AI responses for accuracy and relevance, which allows students to feel a sense of contribution to the AI training process [12][14] Group 2: Financial Aspects and Job Market Dynamics - Despite the high expectations set by advertisements, actual earnings from these part-time jobs are often lower than anticipated, with students like Wang Lei reporting monthly incomes of 800 to 1,200 yuan after deductions [6][11] - The job market is shifting, with traditional roles being replaced by AI, leading to a 36% increase in AI-related job postings in the first half of the year, while many traditional positions are being cut [16][17] - Companies are increasingly looking for candidates with AI-related experience, making these part-time roles valuable for students seeking internships and job opportunities [14][15] Group 3: Challenges and Realities of AI Part-time Work - The work environment for these students is often challenging, with strict requirements for task completion and potential exploitation by intermediaries [11][12] - As the demand for AI training grows, the competition among students for these roles is expected to increase, leading to a more saturated market [18][19] - Students express a mix of hope and anxiety regarding their future job prospects, recognizing the need to adapt to the evolving landscape where AI plays a significant role [17][19]

Core Insights - The article discusses the launch of FinSearchComp, an open-source financial search and reasoning benchmark developed by ByteDance's Seed team in collaboration with Columbia Business School, aimed at evaluating AI's performance in financial analysis tasks [1][3][5] Evaluation Results - The best-performing model, Grok 4 (web), achieved an accuracy of 68.9% on the global dataset, which is still 6.1 percentage points behind human experts. In the Greater China dataset, Doubao (web) led with an accuracy of 53.3%, falling short by over 34 percentage points compared to human experts' 88.3% [1][11] Task Design - FinSearchComp includes three progressively challenging task categories that reflect the complexity of financial analysts' daily work: 1. Time-sensitive data fetching, focusing on real-time data like stock prices [7] 2. Simple historical lookup, requiring fixed-point fact retrieval [7] 3. Complex historical investigation, demanding multi-period aggregation and analysis [7] Data Reliability - The benchmark's quality is supported by ByteDance's Xpert platform, which provides expert knowledge and high-quality AI training data. The project involved 70 financial experts, ensuring data reliability through cross-validation from official sources and professional financial databases [9][10] Importance of Search Capability - The evaluation highlighted the critical role of search capabilities, with models equipped with web search functionality showing significant performance improvements across tasks. Models without search capabilities scored zero on time-sensitive tasks, emphasizing the necessity of real-time data access for accurate financial analysis [12][11] Industry Implications - The findings suggest that while AI can assist in financial data retrieval, it still has considerable room for improvement. The article advocates for the establishment of a comprehensive evaluation system for financial AI, akin to a "driving license" for AI products, to ensure reliability before they can fully replace human analysts [13]