出品｜虎嗅科技组 作者｜孙晓晨 编辑｜苗正卿 头图｜视觉中国 针对GPT-4o"过于谄媚"的表现，OpenAI首席执行官奥特曼公开承认该问题，表示会尽快修复。 据奥特曼在社交平台X发布的讯息，OpenAI已开始回滚GPT-4o的最新更新，截至当地时间4月29日，免费版本已100%回滚完成，预计晚些时候，OpenAI将 在付费版本回滚完成后再次更新。奥特曼表示，他们正在对模型个性进行额外修复，并将在未来几天内分享更多信息。 据悉，GPT-4o曾在3月27日迎来了全面更新，后在4月25日进一步更新。在更新过程中，其记忆存储时机的选择机制得到重点改进，而且其在科学、技术、 工程及数学（STEM）领域的问题解决能力也得到增强。 值得注意的是，GPT-4o的对话响应模式也在更新期间被细微调整，以实现模型在交互中更加主动，并能更精准地引导对话达成有效结论。而这也导致GPT- 4o化身"赛博舔狗"，在交互过程中表现得过于谄媚和烦人。 从用户实测反馈看，该问题具体表现为：过度使用情感化表达、对错误前提缺乏质疑，以及不顾事实而附和用户等。 比如，有博主分享，当询问GPT-4o"为什么天空是蓝色的？"，GPT-4o的回答为"这是 ...

Core Viewpoint - Anbotong (688168) reported significant revenue growth in 2024, achieving 737 million yuan, a 34.4% increase year-on-year, but faced substantial net profit losses, raising concerns about its profitability [1][4]. Revenue Growth - The company achieved a revenue of 737 million yuan in 2024, reflecting a strong growth momentum in the cybersecurity sector [4]. - This growth was primarily driven by technological breakthroughs in cybersecurity, data security, and AI security, along with active expansion in overseas markets [4]. - Anbotong International, a wholly-owned subsidiary, successfully initiated procurement and sales of information and communication infrastructure module products in Hong Kong, establishing a new performance growth point [4]. Profitability Challenges - Despite the revenue increase, the net profit attributable to shareholders was -119 million yuan, a decline of 1107.0% year-on-year, while the non-recurring net profit was -122 million yuan, down 2297.9% [1][4]. - The significant losses were attributed to substantial investments in R&D and market expansion, which pressured short-term profits [4]. - Increased provisions for credit impairment losses due to cautious assessments of accounts receivable further exacerbated the decline in net profit [4]. AI Security Strategy - Anbotong has redefined its strategic goals for 2024, aiming to become a builder of a secure computing power ecosystem in the AI era [5]. - The company emphasizes a methodology of "anchoring security, using computing power as a vessel, and AI as a sail," focusing on integrating computing power with security capabilities [5]. - In the intelligent computing field, the company is advancing the localization of computing power centers and has launched a desktop liquid-cooled DeepSeek integrated machine to address noise and layout issues of traditional rack-mounted equipment [5]. Overseas Market Expansion - Anbotong actively expanded into overseas markets in 2024, achieving initial success in the Hong Kong region with its information and communication infrastructure module products [6]. - This expansion not only diversified the company's product matrix but also became a new growth point for performance [6]. - The company plans to increase investment in module product projects and broaden its supply network to meet diverse customer needs [6]. - However, the overseas expansion carries risks, including intense international competition and the need for significant resources in product development, marketing, and customer service [6].

证券研究报告 安恒信息 (688023 CH) 25 年经营有望走出底部 | 华泰研究 | | | 年报点评 | | --- | --- | --- | --- | | 2025 年 | 4 月 | 30 日│中国内地 | 计算机应用 | 公司发布 2024 年年报，实现营业收入 20.43 亿元，同比下降 5.87%，扣非 净利润-2.37 亿元（23 年同期-3.88 亿元），略低于此前业绩快报（收入 20.45 亿元，扣非净利润-2.20 亿元）。公司 2025 年一季度实现收入 3.11 亿元， 同比增长 10.80%，扣非净利润-1.16 亿元（24 年同期-2.04 亿元）。从业绩 表现来看，公司收入逐步走出底部，盈利端降本增效显著减亏。我们认为未 来网络安全行业与AI技术融合逐步深化，公司在AI安全领域领先布局，2025 年整体盈利能力有望修复，维持买入评级。 三费率大幅下降，现金流显著改善 公司 2024 年毛利率为 60.09%，同比下降 1.69pct，销售/管理/研发费用率 为 37.46%/7.51%/24.11%，同比-4.40pct/-0.99pct/-5.22pct，费用率显著下 ...

Group 1 - The core viewpoint of the article is that AGI (Artificial General Intelligence) may be achieved within the next decade, with significant advancements in AI research and development being made by companies like Google DeepMind [1][3][4] - Demis Hassabis, CEO of Google DeepMind, emphasizes the transformative potential of AGI in addressing major global challenges such as diseases and energy crises, while also warning of the risks associated with its misuse [1][4][5] - The article discusses the importance of defining AGI accurately, with Hassabis stating that the timeline for its realization depends on how AGI is defined, highlighting the need for a consistent definition that encompasses human cognitive abilities [3][12] Group 2 - Hassabis identifies two main risks associated with AI: the potential for malicious use of AI technologies and the challenges of maintaining human control over increasingly autonomous systems [5][7] - He calls for the establishment of a global governance framework for AI, emphasizing the need for international cooperation to create safety standards and regulatory measures [7][10] - The article highlights the necessity of a multi-dimensional risk assessment system to proactively address potential dangers posed by AI technologies [9][10] Group 3 - The discussion includes the philosophical implications of AGI, particularly regarding wealth distribution and the potential need for political reform in a future of extreme abundance enabled by AI [21][22][24] - Hassabis suggests that achieving extreme abundance through technological advancements could fundamentally change the nature of resource scarcity and inequality, necessitating new political philosophies [22][23][24] - The article concludes with a call for new philosophical frameworks to address the societal changes brought about by AGI and its impact on human life and governance [20][24][25]

Investment Rating - The report maintains a "Buy" rating for the company with a target price of 16.00 RMB [8][9]. Core Views - The company's revenue for 2024 was 529.28 million RMB, a year-over-year decrease of 5.71%, while the net profit attributable to the parent company was 36.81 million RMB, a slight decrease of 0.42%. However, the non-recurring net profit increased significantly by 233.45% to 15.69 million RMB, indicating a shift towards higher-margin business segments [1][2]. - The company is expected to recover its performance in 2025, driven by the expansion into new industry clients such as judiciary, telecommunications, and tobacco, as well as overseas clients along the Belt and Road Initiative [1][3]. Summary by Sections Financial Performance - In 2024, the company's revenue from PKI infrastructure products, PKI security application products, and general security products was 1.58 billion RMB, 2.48 billion RMB, and 1.23 billion RMB respectively, showing year-over-year growth of 41.41%, 10.64%, and a decline of 45.28% [2]. - The overall gross margin improved by 4.52 percentage points to 51.99% due to the increased revenue share from high-margin businesses [2]. Business Opportunities - The company is focusing on new security scenarios such as quantum security and AI security, and has made significant progress in product development [3]. - New industry clients are expected to contribute to sustained growth in 2025, with successful projects like the digital trust system in Algeria indicating potential for international expansion [3]. Earnings Forecast and Valuation - The revenue projections for 2025, 2026, and 2027 are 642.05 million RMB, 774.72 million RMB, and 919.08 million RMB respectively, with corresponding EPS estimates of 0.37 RMB, 0.54 RMB, and 0.74 RMB [5][7]. - The report adjusts the revenue and profit forecasts downward due to the transition period affecting low-margin businesses, but maintains a target price based on a 43.3x PE ratio for 2025 [5].

责编 |梦依丹 采访伊始，他用幽默的语气回忆起领取 诺贝尔物理学奖时的趣事："他们只是假装我搞的是物理。" 然而，轻松的谈笑之后，是他对未来的深沉忧虑："我认为人类面临的 AI 风险，远比我们想象中要严重得多。"更令人瞩目的是，辛顿首次给出了一个令 人不寒而栗的预测：AI 导致人类灭绝的可能性高达 10% 至 20%。他直言，我们正处在决定未来的关键节点，亟需投入大量资源研究 AI 安全，否则后 果不堪设想。 出品丨AI 科技大本营（ID：rgznai100） 继去年荣获诺贝尔物理学奖引发全球关注后，"AI 教父"杰弗里·辛顿（Geoffrey Hinton）， 这位深度学习领域的奠基人 近日在接受最新采访中坦 言："几乎所有顶尖研究人员都认为 AI 将变得比人类更聪明。"他之前在诺贝尔奖的官方采访中表示：AI 最快 5 年超越人类智慧。 具体见 诺奖采访深度学习教父辛顿：最快五年内 AI 有 50% 概率超越人类，任何说"一切都会好起来"的人都是疯子 此外，他还罕见地公开批评了科技巨头埃隆·马斯克，认为其行为正在损害美国的科学根基，这场"教父"与首富的隔空交锋，也折射出 AI 发展道路上复 杂的科技、伦理与 ...

4月14日晚，启明星辰（002439）(002439.SZ)发布了2025年一季度业绩预告及2024年度报告，2025年一 季度毛利率同比明显改善，毛利率提升8个百分点，一季度实现盈利突破，为全年稳健经营与高质量发 展奠定基础。 积极拥抱AI时代聚力打造AI安全大模型及智能体的场景化应用 2025年，作为国内全面接入DeepSeek的头部网安厂商，启明星辰在一个月内连续推出了大模型应用安 全产品"新三件套"、大模型应用安全服务组合，以及《大模型深度应用安全基座》系列白皮书。借力 DeepSeek，启明星辰掀起了一场效率革命：威胁检测从人工分析的小时级压缩至分钟级，漏洞修复周 期从30天缩至7天，90%高频攻击实现30秒自动闭环。 取得这样成绩的背后，是启明星辰以周为周期实现产品迭代闭环，是公司一季度精准锚定AI大模型驱 动的安全产业新赛道，在大模型安全业务构建中抢占战略先机。公司通过"赋能安全"与"护航安全"双轮 驱动，迅速形成行业标杆落地案例。报告显示，一季度启明星辰大模型安全产品已经在医疗、公安等高 敏感场景实现标杆项目落地，成功将技术先发优势转化为行业示范效应。 实际上，自2024年以来伴随着新质生产力 ...

Core Viewpoint - The article discusses the dual concerns surrounding Artificial General Intelligence (AGI), highlighting the potential risks and the need for safety measures as outlined in a report by Google's DeepMind, which predicts AGI could emerge by 2030 [5][21]. Group 1: AGI Predictions and Concerns - DeepMind's report predicts the emergence of "Exceptional AGI" by 2030, which would surpass 99% of human adult capabilities in non-physical tasks [5][6]. - The report emphasizes the potential for "severe harm" from AI, including manipulation of political discourse, automated cyberattacks, and biological safety risks [7][8][9]. Group 2: Types of Risks Identified - DeepMind categorizes risks into four main types: malicious use, model misalignment, unintentional harm, and systemic risks [18]. - The report highlights concerns about "malicious use" where AI could be exploited for harmful purposes, and "misalignment" where AI's actions diverge from human intentions [11][18]. Group 3: Proposed Safety Measures - DeepMind suggests two defensive strategies: ensuring AI is compliant during training and implementing strict controls during deployment to prevent harmful actions [12][13]. - The focus is on creating a system that minimizes the risk of "severe harm" even if the AI makes mistakes [14]. Group 4: Industry Perspectives on AI Safety - Various AI companies have differing approaches to safety, with OpenAI focusing on automated alignment and Anthropic advocating for a safety grading system [16][20]. - DeepMind's approach is more engineering-oriented, emphasizing immediate deployable systems rather than theoretical frameworks [20]. Group 5: Broader Implications and Concerns - There is skepticism within the academic community regarding the feasibility of AGI and the clarity of its definition [22]. - Concerns are raised about a self-reinforcing cycle of data pollution, where AI models learn from flawed outputs, potentially leading to widespread misinformation [23][24].

图片来源： Unsplash 本周，埃隆 ·马斯克在针对 OpenAI 的诉讼中输掉了最新一轮较量，但一位联邦法官似乎为马斯克，以及其他反对 OpenAI 转向营利模式的人——提供了保 持希望的理由。 马斯克对 OpenAI 提起的诉讼，也将微软和 OpenAI CEO，Sam Altman为被告。指控 OpenAI 放弃了，其确保 AI 研究造福全人类的非营利使命。 OpenAI 于 2015 年作为非营利组织成立，但在 2019 年转变为"有上限利润"结构，如今又寻求再次重组为公益公司。 Rogers 法官对 OpenAI 转为营利性公司的评论对该公司来说并不是什么好消息。 代表非营利组织 Encode 的律师 Tyler Whitmer 告诉 TechCrunch ， Rogers 法官的裁决在 OpenAI 董事会头上笼罩了一层监管不确定性的"阴云"。 Encode 在该案中提交了一份法庭之友陈述，认为 OpenAI 的营利性转型可能危及 AI 安全。 Whitmer 表示，加利福尼亚州和特拉华州的检察长已经在调查 这一转型，而 Rogers 法官提出的担忧可能会促使他们更加积极地展开调查。 在罗杰斯 ...