图片巨头 Getty Images 则抓住了 Stability AI 的现行—— AI 不仅「吃」了它数百万张带版权的图片，甚至在生成的新图里，还心虚地「吐」出了那个标志 性的水印，像个作案后忘了擦掉指纹的笨贼。 这些案子的起因都源于那个让所有 AI 公司夜不能寐的问题： 用来训练模型的海量数据，究竟算不算「赃物」？ 过去几个月，AI 圈一桩桩版权案接连爆出。 迪士尼拉着环球影业，把 Midjourney 告上法庭，指控它在没打一声招呼的情况下，就学会了画《星球大战》和《小黄人》。 最近，两场判决落槌，似乎给出了一个答案。Anthropic 和 Meta 在版权官司中都获胜让许多人以为警报已经解除。但细看之下，这场胜利并非这么简单 ——法官在判决书里悄悄留下了一串解开谜题的钥匙。 从表面上看，案情并不复杂。一群作家指控这两家公司用他们写的书来训练大模型，这是偷窃。AI 公司则辩称，这叫「变革性使用」，是学习，不是抄 袭，属于法律允许的「合理使用」。 出人意料的是，法官采纳了 AI 公司的说法。他们认为，用书籍训练 AI，和直接复制书籍来卖是两码事，确实有「变革性」，而且原告作家们也没能拿出 有力证据，证明自 ...

Core Insights - The global PVC market is experiencing a significant supply surplus due to increased supply, weak demand, and rising international trade protectionism [2][3] - The imbalance in supply and demand is pressuring PVC prices and pushing many producers, especially in the U.S., towards unsustainable profit margins [2][4] - The U.S. is facing a dramatic shift in trade patterns, with anti-dumping measures from multiple countries hindering its traditional role as a major PVC exporter [3][5] Supply and Demand Dynamics - The current global PVC supply far exceeds demand, necessitating rationalization of production capacity [5] - Despite a projected 8% growth in domestic PVC demand in the U.S. for 2024, the local market is insufficient to absorb existing production capacity [5] - The ongoing expansion in East Asia, the Middle East, and South Asia is contributing to increased production capacity, which is not supported by demand growth expectations [5] Trade and Regulatory Environment - Anti-dumping measures are reshaping global PVC trade flows, particularly affecting U.S. exports to Europe and other key markets [3] - Recent increases in import tariffs, such as Brazil's PVC tariff rising from 8.2% to 43.7%, pose significant challenges for U.S. producers [3] - The European and UK markets are tightening access to U.S. PVC, effectively closing doors to American exports [3] Cost Structure and Profitability - European and Asian integrated PVC producers have total costs that are 20% to 30% higher than those of U.S. producers, primarily due to differences in natural gas and electricity prices [4] - Current global PVC spot prices are below the production costs for many companies, indicating unsustainable profitability across the industry [4] - The only recent positive aspect in the PVC market has been temporary shipping arbitrage opportunities due to disruptions in the Red Sea [4]

将投中网设为"星标⭐"，第一时间收获最新推送 巴菲特反复提及："能够一直 获得丰厚利润的投资者，几乎都是格雷厄姆和多德的信徒 。"他称自己投资理论 的"85%来自格雷厄姆"，甚至给自己的儿子取名霍华德·格雷厄姆·巴菲特…… 格雷厄姆是谁？他就是价值投资理论的奠基人，被誉为"华尔街教父"。他在1934年首次推出《 证券分析 》，书中 的不朽智慧，被证券市场的沉浮起落所反复印证。 1984年，巴菲特甚至专门为这本书做了一次专场演讲。出版近百年，《证券分析》依然被华尔街以及全球投资大佬 奉为"投资圣经"！你所领略过的所有投资知识，都能在这本书中找到影子。 《证券分析》中的许多细节源自另一个时代。最早的版本隐晦地描述了蒸汽时代的经济，蒸汽驱动的火车在大地 上飞驰，输送了大量的工业品。如今，全球经济的信息化特征越来越明显，有成千上万的人通过电脑和手机收 集、追踪和分析数据，致力于信息经济；而无数其他人则积极投身于蓬勃发展的服务经济。 早期版本中提到的许多企业已经被合并、重组或清算，格雷厄姆和多德使用的一些工具和方法已经过时或不再适 用。例如，账面价值对今天的投资者来说，远不如一个世纪前那么重要。格雷厄姆和多德建议投资者 ...

Core Viewpoint - The recent court rulings in the U.S. regarding AI training and copyright usage have provided a significant legal precedent for AI companies, affirming that certain uses of copyrighted materials for AI training can be considered "fair use" [1][3][12]. Group 1: Court Rulings - On June 23, 2023, Anthropic won a landmark ruling from the U.S. District Court for the Northern District of California, where the court deemed its practice of digitizing and using millions of books for training its AI model, Claude, as "fair use" [3][7]. - On June 25, 2023, a federal judge dismissed a lawsuit against Meta, ruling that its use of copyrighted works for AI training also fell under "fair use" [9][12]. - The rulings emphasize the concept of "transformative use," where the AI's output is not a direct copy of the original works but rather a new creation inspired by them [7][14]. Group 2: Legal Implications - The court's decisions provide a "safe harbor" for AI companies, potentially reducing their legal risks and costs associated with AI development [12][14]. - However, the rulings also highlight the need for clear distinctions between lawful and unlawful data sources, as seen in Anthropic's case where it downloaded over 7 million pirated books, which the court condemned as copyright infringement [8][12]. - The potential financial implications for Anthropic could be severe, with estimates suggesting it may face billions in damages due to its use of pirated materials [8][9]. Group 3: Industry Impact - The rulings are expected to accelerate the development of AI technologies by clarifying the legal landscape surrounding data usage for training [14][18]. - There is a growing concern among creators and copyright holders about the impact of AI-generated content on the value of original works, as AI can produce content at scale and lower costs [17][18]. - The balance between fostering innovation in AI and protecting the rights of creators remains a critical issue, with ongoing discussions about how to ensure fair compensation and recognition for original works [18][19].

Core Viewpoint - The company is conducting a transaction to acquire all shares of Hangzhou Bicheng Digital Technology Co., Ltd. and has engaged Shanghai Zhonghua Asset Appraisal Co., Ltd. as the evaluation agency to assess the market value of the target assets [1][2]. Group 1: Independence and Reasonableness of the Evaluation - The evaluation agency, Shanghai Zhonghua, has no related party relationships with the company or the transaction counterpart, ensuring its independence [1]. - The assumptions made in the evaluation report comply with national laws and regulations, and are reasonable according to market practices and asset appraisal standards [1][2]. Group 2: Evaluation Methodology and Purpose - The purpose of the evaluation is to determine the market value of the target assets as of the evaluation benchmark date, providing a reference for the transaction [2]. - The evaluation methods used are appropriate, and the conclusions drawn are reasonable, reflecting the actual situation of the evaluated assets [2]. Group 3: Fairness of the Transaction Price - The transaction price will be based on the evaluation report, adjusted for cash dividends during the transition period, ensuring fairness and reasonableness without harming the interests of the company and minority shareholders [2].

21世纪经济报道记者肖潇 北京报道 在不经授权的情况下用人类作品训练AI，算不算侵权？围绕AI版权的争议已持续三年，两起来自美国 的司法判决为这场讨论提供了新的参照： 6月23日，美国加州北区联邦法院裁定，AI公司Anthropic未经3位作家许可，使用其已出版书籍训练大 模型Claude的行为合法，符合"合理使用"原则； 6月25日，同一法院的另一法官发布简易判决，认为Meta公司在未经13名作家许可的情况下使用书籍训 练大模型Llama，同样构成"合理使用"。 这两起判决在整体上为AI公司开了绿灯，但也警告了红线。两份判决书都提到，"合理使用"的适用范围 有限，需要考虑数据采集方式是否侵权。 对市场或作品价值的影响是Meta案法官最看重的一点。"在很多情况下，未经授权用受保护作品训练AI 大模型是违法的，因为可能削弱创作者的创作动机和市场回报。"但在本案中，法院认为原告没有证明 这一点。 Anthropic案的判决思路与之类似。法院同样认为，训练Claude模型的过程具备"高度转换性"，并非为了 再现原作品，最终也没有向用户输出原文或类似内容，只是输出风格相似的表达，因此不能视作市场替 代品。判决书引用了 ...

Core Viewpoint - A federal judge ruled in favor of Meta Platforms, dismissing serious copyright infringement claims from a group of authors who alleged that Meta used their books without permission to train its AI systems. The judge found insufficient evidence to prove that Meta's actions diluted the market for the authors' works, thus not violating U.S. copyright law [1][2]. Group 1: Legal Context - The ruling is significant as it supports the argument that using copyrighted works to train AI systems can fall under "fair use," a principle that allows limited use of copyrighted material without permission [3][4]. - The judge, Vince Chhabria, noted that while the current case did not establish legality for Meta's practices, it highlighted that the plaintiffs failed to present valid arguments or evidence [2][3]. Group 2: Industry Implications - The decision is part of a broader trend where AI companies, including OpenAI and Anthropic, argue that their use of copyrighted materials for training is transformative and falls within fair use, which is crucial for the growth of the AI sector [3][4]. - The ruling may influence future cases as it represents the second significant federal court decision regarding the fair use of copyrighted materials in AI training, following a controversial ruling in a separate case involving Anthropic [3][4]. Group 3: Market Reactions - Meta expressed appreciation for the ruling, emphasizing that fair use is a key legal framework for developing transformative AI technologies [3]. - The case has garnered significant market attention, as it reflects ongoing tensions between traditional copyright holders and emerging AI technologies, with copyright holders claiming that AI companies threaten their livelihoods by generating competing content [4][5].

不圆 发自 凹非寺 量子位 | 公众号 QbitAI 无需原作者同意， AI可以用已出版书籍作训练数据了 。 就在最新判决的诉讼中，美国法院裁决：允许Claude背后公司Anthropic在 未经作者许可 的情况下，使用 合法购买 的已出版书籍训练AI。 法院参考了美国版权法中的"合理使用" （Fair Use） 原则，认为AI训练属于"转化性使用" （Transformative Use） ，即对原作品的新用途 未取代原作市场，且有利于技术创新和公共利益。 这是美国法院首次认可AI公司对书籍的 使用权 ，保护人工智能公司在使用受版权保护的文本训练 LLMs 时不受限制： 大大降低了AI训练数据的 版权风险 。 对此，不少网友的看法是这样的：既然人类读书并加以理解是毫无争议的，那AI读书并理解也应该合理。 怎么一回事？ 对Anthropic的指控，由三位作家在2024年8月发起。 2、原告对于Anthropic的指控是： 非法复制作品（包括盗版和扫描版）用于训练AI，侵犯版权 。 值得关注的是，Anthropic的案件不仅针对使用已出版书籍训练AI，还涉及到书籍的来源问题： 2021年，Anthropic的联 ...

当实现短期目标存在压力时，政策会加码；反之，政策会更加侧重于长期目标 。 文：国金宏观宋雪涛/联系人张馨月 去年以来，政策的核心逻辑始终围绕平衡短期目标和长期目标展开 。短期目标是稳增长、稳预期，最 直接的目标是实现5%的经济增速。长期目标是推动改革、加快转型，通过科技创新和市场化改革提高 经济要素的配置效率，打开经济潜在增长空间。去年"9.24"以来，短期目标的紧迫性明显上升，但长 期目标的重要性依然不容忽视。当实现短期目标存在压力时，政策通常会加码；反之，政策会更加侧重 于长期目标。 从当前内外部环境看，实现短期目标的压力较小 。外交方面，中美之间的不确定性从日内瓦协议到伦 敦高会谈以来持续减弱；中国与非美特别是周边国家外交取得积极进展。国内方面，上半年消费表现亮 眼，5月社零同比增长6.4%，创2024年以来新高；出口仍有韧性，5月出口同比增长4.8%；二季度经 济增速有望延续一季度5.4%的表现，全年实现5%增长目标的压力不大。 近期，消费、地产、改革、科技等领域的政策声音都传递出由短期刺激转向长期支持的信号。 （一）消费："以旧换新"强调节奏"平稳有序"，整治套补骗补 6月以来，各地陆续暂停国补，原 ...

Core Viewpoint - The company, Guangdong Zhaoqing Xinghuo Biotechnology Co., Ltd. (Xinghuo Technology), has received an inquiry letter from the Shanghai Stock Exchange regarding its 2024 annual report, particularly focusing on its cash flow from investment activities, which has shown continuous outflows over the past three years [1][2]. Investment Activities - The net cash flow from investment activities for the years 2022 to 2024 was -789 million, -859 million, and -864 million yuan respectively, indicating a consistent cash outflow [1]. - The cash payments for investments during the same period were 1.579 billion, 2.412 billion, and 1.8 billion yuan [1]. Major Investment Projects - The company has detailed its major investment projects, including the acquisition of 99.22% of Ningxia Yipin Biotechnology Co., Ltd. for a total transaction price of 5.376 billion yuan, with 4.585 billion yuan paid through share issuance and 791.6384 million yuan in cash [3][4]. - The decision to use self-owned funds for the cash payment of the acquisition was approved by the board of directors, adhering to the company's investment decision-making procedures [4]. Financial Performance - The company reported operating revenues of 1.601 billion, 1.606 billion, and 1.605 billion yuan for the years 2022 to 2024, with net profits attributable to the parent company of 120.579 million, 75.997 million, and 108.141 million yuan respectively [6]. - The net cash flow from operating activities increased from 211.975 million yuan in 2022 to 239.776 million yuan in 2024, indicating improved operational efficiency [19]. Cash Management and Investment Strategy - As of the end of 2024, the company had a cash balance of 1.759 billion yuan, a year-on-year increase of 23.97%, with interest income of 13 million yuan [9]. - The company has established a risk management system and a comprehensive financial management policy to control investment risks, including a clear approval process for investment products [8]. Debt and Financing - The company’s long-term borrowings stood at 1.721 billion yuan and short-term borrowings at 1.008 billion yuan at the end of 2024, with interest expenses of 110 million yuan [9]. - The company has been reducing its bank loan balance over the years, with a total bank borrowing of 3.391 billion yuan at the end of 2024, down from 4.188 billion yuan in 2022 [19]. Acquisition and Goodwill - The company acquired 100% of Sichuan Jiuling Pharmaceutical Technology Co., Ltd. in 2019, resulting in goodwill of 198 million yuan, which has been subject to impairment testing due to declining profitability in subsequent years [20][23].