【环球网科技综合报道】10月21日消息，据外媒CNBC报道，OpenAI与美国演员工会（SAG-AFTRA） 等共同发布联合声明，宣布将建立深度合作机制，重点防范其AI视频生成工具Sora可能引发的深度伪造 （Deepfake）滥用问题。 来源：环球网 据悉，Sora作为OpenAI推出的文本到视频生成模型，凭借其"1分钟长视频生成""物理规律模拟""多镜头 叙事"等能力，自曝光以来便引发全球关注。然而，其强大的内容生成能力也带来深度伪造风险——不 法分子可能利用该技术伪造名人影像、篡改历史事件或制造虚假新闻，对个人名誉、社会信任甚至国家 安全构成威胁。 此次合作中，OpenAI承诺在Sora的后续版本中实施严格的"选择加入"（Opt-In）政策：所有艺术家、表 演者及普通用户需明确授权后，其声音、肖像方可被用于AI生成内容。例如，若用户希望生成布莱恩· 克兰斯顿的虚拟影像，必须通过其本人或授权代表的确认，并明确使用场景与期限。此外，OpenAI将 建立快速响应机制，对涉嫌侵权的生成内容进行下架处理，并配合执法部门追溯源头。 此外，OpenAI宣布将向合作方开放部分技术细节，鼓励第三方机构开发深度伪造检测工具 ...

机器之心报道 编辑：泽南、Panda 「我很喜欢新的 DeepSeek-OCR 论文…… 也许更合理的是，LLM 的所有输入都应该是图像。即使碰巧有纯文本输入，你更应该先渲染它， 然后再输入。」 一夜之间，大模型的范式仿佛被 DeepSeek 新推出的模型给打破了。 昨天下午， 全新模型 DeepSeek-OCR 突然开源 。在该模型的处理过程中，1000 个字的文章能被压缩成 100 个视觉 token，十倍的压缩下精度也可以达到 97%，一 块英伟达 A100 每天就可以处理 20 万页的数据。 这种方式或许可以解决大模型领域目前头疼的长上下文效率问题，更重要的是，如果「看」文本而不是「读」文本最终被确定为正确的方向，也意味着大模型的 范式会发生重要的转变。 GitHub 上， DeepSeek-OCR 项目一晚收获了超过 4000 个 Star 。 因为是开源的小模型，DeepSeek-OCR 第一时间经历了整个 AI 社区的检验，很多大佬在看完论文之后纷纷发表了看法，兴奋之情溢于言表。 OpenAI 联合创始成员之一，前特斯拉自动驾驶总监 Andrej Karpathy 表示，它是一个很好的 OCR ...

Core Insights - The article discusses the impact of supervised fine-tuning (SFT) on the general capabilities of large language models (LLMs), suggesting that SFT does not always lead to a significant decline in general performance when a smaller learning rate is used [2][34] - The research challenges the long-held belief that domain-specific fine-tuning inevitably causes catastrophic forgetting of general capabilities, proposing that the choice of training strategy plays a crucial role [2][34] Experiment Details - The study utilized two domain-specific datasets, MedCalc and ESCI, which represent scenarios where open-source LLMs typically perform poorly, making them ideal for domain-specific SFT [5] - Various open-source LLMs were selected for experimentation, including Qwen3-8B and Gemma3-4B, with a focus on controlling the learning rate during SFT [6] Findings - **Finding 1**: Using a smaller learning rate (e.g., 1e-6) allows models to maintain strong performance in target domains while significantly reducing the decline in general capabilities [11] - **Finding 2**: For classification tasks, when the training objective includes only the final label, a wider range of learning rates can achieve ideal performance, as seen in the ESCI dataset [12][14] Theoretical Analysis - The research team concluded that smaller learning rates can effectively limit the decline in general performance, aligning with the experimental findings [17] - The analysis also indicated that when training targets only include final labels, the number of "hard tokens" encountered decreases, allowing for a broader acceptable learning rate range [17] Token Adaptive Loss Reweighting (TALR) - TALR is introduced as a method to dynamically adjust the loss weights of tokens based on their prediction probabilities, effectively reducing the impact of hard tokens during training [20] - The method allows for real-time updates of token weights, ensuring that the model's confidence levels guide the training process [21] Experimental Results - In experiments comparing various strategies to mitigate catastrophic forgetting, TALR demonstrated superior performance, especially under higher learning rates, maintaining domain gains while minimizing losses in general performance [26][27] Conclusion and Future Directions - The research emphasizes the continued importance of SFT in enhancing LLM capabilities, suggesting that while smaller learning rates and TALR are effective, further exploration of more robust strategies is necessary to address the forgetting problem [34][35] - Future research should focus on balancing domain-specific performance with general capabilities, particularly in specialized fields like medicine, where retaining foundational knowledge is crucial [35]

Core Insights - Meta has experienced explosive user growth following the launch of its short video feature "Vibes," with daily active users increasing from 775,000 to 2.7 million in just four weeks, and daily downloads reaching 300,000 [1][2][5] User Growth and Market Impact - The launch of "Vibes" on September 25 is closely correlated with the surge in Meta AI's user numbers, providing strong data support for the argument that AI video drives traffic [2][5] - On October 17, Meta AI's daily active users grew by 15.58%, while competitors ChatGPT, Grok, and Perplexity saw declines of 3.51%, 7.35%, and 2.29% respectively [4] Features of Vibes - "Vibes" allows users to create, discover, and share AI-generated short video content, with personalized recommendations improving over time as users engage more [7] - Users can easily repurpose content and share it across platforms like Instagram and Facebook, enhancing the potential for viral spread [7] Competitive Landscape - Meta AI's growth may also be indirectly supported by competitor Sora's recent strategies, which have drawn attention to AI video generation technology [8][9] - Sora's "invite-only" strategy may have inadvertently created opportunities for Meta AI, as users unable to access Sora may seek alternatives like Meta AI [9]

这家总部设于多伦多和旧金山的公司于周一表示，将为现有员工提供参与此次二次出售的机会，使其无 需等待公司上市或被收购即可实现股权变现。 智通财经APP获悉，人工智能(AI)初创企业Cohere日前宣布，将为有意向外部投资者出售所持股权的员 工启动一项股权转让计划。该公司上月刚以70亿美元估值完成融资，并获得英伟达公司的投资。 在团队建设方面，Cohere近期引进了两位担任关键职务的高管：前Meta(META.US)公司AI研究负责人 Joelle Pineau已加盟担任首席AI官;前优步(UBER.US)高管Francois Chadwick则出任公司首席财务官。 公司同时披露，其年化营收已突破1.5亿美元(该指标基于近期业绩对公司年度营收的预测)，较今年5月 的1亿美元实现显著增长。 截至目前，Cohere已累计融资约16亿美元。今年8月，由Radical Ventures和Inovia Capital领投，英伟达和 AMD(AMD.US)旗下AMD Ventures等现有投资方跟投，公司完成了5亿美元融资。一个月后，又通过第 二轮交割获得1亿美元注资，使公司估值达到约70亿美元。 Cohere公司成立于20 ...

上周，在 Dreamforce 2025 峰会，Anthropic 联合创始人兼 CEO Dario Amodei 说了一句引发广泛关注的 话： 我对短期内 AI 的互补性依然乐观，但我也必须坦白:两到五年内，真正的替代将开始出现。 这意味着，"AI 替代"不再是科幻电影里的威胁，而是从 Anthropic 办公室里的一行行代码开始，在真实 的生产系统里一步步落地。 但 Amodei 的警告，不止是给工程师的。 "这不会只发生在开发者身上，"他补充说，"保险、金融、医疗等行业的企业客户已经在用 Claude 执行 端到端任务。被替代的第一批人，往往是工作流程里那些最容易被自动化的岗位。" "AI 取代人类"这件事，从理论进入了倒计时。 接下来，问题不再是AI 会不会替你，而是谁会先被替？他们是怎么一步步被替的？而你，又该如何重 新定义自己的角色？ 第一节｜替人的起点：端到端能力 今天很多人说 Claude 会写代码，但 Dario Amodei 的重点根本不是写代码，而是它能做完一件完整的 事。 他观察到：以前 AI 只是帮你写几行代码，现在它可以调试系统、修复 Bug、完成整个部署流程。 换句话说，不再是你 ...

Artificial intelligence is the greatest investment opportunity of our lifetime. The time to invest in groundbreaking AI is now, and this stock is a steal! AI is eating the world—and the machines behind it are ravenous. Each ChatGPT query, each model update, each robotic breakthrough consumes massive amounts of energy. In fact, AI is already pushing global power grids to the brink. Wall Street is pouring hundreds of billions into artificial intelligence—training smarter chatbots, automating industries, and b ...

Core Insights - Artificial intelligence (AI) is identified as the greatest investment opportunity of the current era, with a strong emphasis on the urgency to invest now [1][13] - The energy demands of AI technologies are highlighted, indicating a looming crisis in power supply as AI continues to grow [2][3] Investment Opportunity - A specific company is positioned as a critical player in the AI energy sector, owning essential energy infrastructure assets that will benefit from the increasing energy demands of AI data centers [3][7] - This company is not a chipmaker or cloud platform but is described as a "toll booth" operator in the AI energy boom, collecting fees from energy exports [5][6] Financial Position - The company is noted for being debt-free and holding a significant cash reserve, which is approximately one-third of its market capitalization [8] - It is trading at less than 7 times earnings, making it an attractive investment compared to other energy and utility firms burdened with debt [10][8] Market Trends - The company is positioned to benefit from the onshoring trend driven by tariffs, as well as the surge in U.S. LNG exports under the current administration [14][7] - The infrastructure needs for energy are emphasized, with the company being capable of executing large-scale projects across various energy sectors [7][6] Future Outlook - The influx of talent into the AI sector is expected to drive continuous innovation and advancements, reinforcing the potential for growth in AI investments [12] - The overall narrative suggests that investing in AI and its supporting energy infrastructure is crucial for capitalizing on future technological advancements [11][13]

Core Insights - Artificial intelligence (AI) is identified as the greatest investment opportunity of the current era, with a strong emphasis on the urgent need for energy to support its growth [1][2][3] - A specific company is highlighted as a key player in the AI energy sector, owning critical energy infrastructure assets that are essential for meeting the increasing energy demands of AI technologies [3][7] Investment Landscape - Wall Street is investing hundreds of billions into AI, but there is a pressing concern regarding the energy supply needed to sustain this growth [2] - AI data centers consume energy equivalent to that of small cities, leading to a strain on global power grids and rising electricity prices [2][3] - The company in focus is positioned to capitalize on the surge in demand for electricity driven by AI advancements [3][6] Company Profile - The company is described as a "toll booth" operator in the AI energy boom, benefiting from the increasing need for energy infrastructure [4][5] - It is involved in the U.S. LNG exportation sector, which is expected to grow significantly under the current administration's energy policies [7] - The company is noted for its capability in executing large-scale engineering, procurement, and construction projects across various energy sectors, including nuclear energy [7][8] Financial Position - The company is completely debt-free and has a substantial cash reserve, amounting to nearly one-third of its market capitalization [8] - It is trading at a low valuation of less than 7 times earnings, making it an attractive investment opportunity compared to other energy and utility firms [10][11] Market Trends - The ongoing AI infrastructure supercycle, combined with the onshoring boom and a surge in U.S. LNG exports, positions the company favorably in the market [14] - The influx of talent into the AI sector is expected to drive continuous innovation and advancements, further solidifying the importance of energy infrastructure [12][13]

Core Insights - Artificial intelligence (AI) is identified as the greatest investment opportunity of the current era, with a strong emphasis on the urgency to invest in AI technologies now [1][13] - The energy demands of AI technologies are highlighted, with data centers consuming as much energy as small cities, leading to concerns about power grid strain and rising electricity prices [2][3] Investment Opportunity - A specific company is presented as a key player in the AI energy sector, owning critical energy infrastructure assets that are essential for supporting the anticipated surge in energy demand from AI data centers [3][7] - This company is characterized as a "toll booth" operator in the AI energy boom, benefiting from the increasing need for energy as AI technologies expand [4][5] Market Position - The company is noted for its unique position in the market, being one of the few capable of executing large-scale engineering, procurement, and construction projects across various energy sectors, including nuclear energy [7][8] - It is described as debt-free and holding a significant cash reserve, which is nearly one-third of its market capitalization, positioning it favorably compared to other energy firms burdened with debt [8][10] Growth Potential - The company also has a substantial equity stake in another AI-related venture, providing investors with indirect exposure to multiple growth opportunities in the AI sector without the associated high premiums [9][10] - The stock is considered undervalued, trading at less than seven times earnings, which presents a compelling investment opportunity given its ties to both AI and energy sectors [10][11] Industry Trends - The article discusses the broader trends of onshoring driven by tariffs and the surge in U.S. LNG exports, indicating a favorable environment for the company's growth [6][14] - The influx of talent into the AI sector is expected to drive continuous innovation and advancements, further solidifying the importance of investing in AI-related companies [12]