IT之家 10 月 31 日消息，微软联合创始人、前 CEO 比尔・盖茨本周（10 月 28 日）出席 CNBC 电视台《Squawk Box》节目，与主持人对谈 AI 行业目前的 现状，并预测未来 AI 行业将出现"泡沫破裂"。 IT之家注：互联网泡沫指的是 1995 年-2002 年前后，全球互联网行业发展初期出现的一场投机潮，当时人们对"互联网改变世界"展现巨大期待，大量资金疯 狂涌入网络公司，甚至许多资金流向了没有盈利能力的骗子企业。 最终这场泡沫在 2000 年 3 月 10 日达到顶峰，纳斯达克指数达到 5048 点，但从 4 月开始科技股票就大幅下跌，逐步导致市场信心崩溃，引发大量互联网公 司破产，最终纳指在 2002 年 10 月跌至 1100 点左右，不过这场泡沫也淘汰掉了许多不良商业模式，留下了亚马逊、谷歌、eBay 等真正能赚钱的公司。 他对"互联网泡沫"破裂时的景象回忆道："最终整个行业发生了具有深远意义的事件（指 2002 年纳指跌至 1100 点左右），世界变得完全不同。一些公司最 终在泡沫破裂时站稳了脚跟，但大多数公司只是跟风投机，消耗资本最终倒闭"。 盖茨预测道，一些投资数十亿 ...

Core Insights - The Guangdong Province held a release event for generative AI models, where 105 models received "Generative AI Service Filing Certificates," marking a significant milestone in the region's AI development [1][4][3] Group 1: Event Overview - The event was themed "Empowerment through Filing, Creating the Future of Guangdong" and took place at the Guangdong-Hong Kong-Macao Greater Bay Area Generative AI Safety Development Joint Laboratory [1] - Over 100 representatives from leading AI companies and relevant government units attended the event [1] Group 2: Model Filing Achievements - Guangdong ranks second nationally in the number of filed generative AI models, with 105 models filed as of October 24, 2025, covering various sectors such as government, education, transportation, and agriculture [4] - The surge in model filings reflects Guangdong's enhanced R&D capabilities in AI, laying a solid foundation for the integration of data elements and industry scenarios [4] Group 3: Importance of Compliance - The filing of generative AI models is crucial for the compliant development of AI enterprises, aimed at promoting healthy development and safeguarding national security and public interests [4] - Experts at the event provided insights into the core review points and common misconceptions regarding model filing, emphasizing that compliance is not only a policy requirement but also a market expansion opportunity for businesses [4] Group 4: Model Innovations and Applications - Several companies shared their innovations, highlighting the unique features and pain points addressed by their models, such as personalized learning paths and advanced reasoning capabilities [7][8] - Notable models include: - "Xingjie" by Guangzhou Fangzhou Information Technology, offering various user assistance functions [7] - "Smart Shadow Quick Language" by Guangzhou Softbank, integrating image and text data for enhanced content generation [7] - "Deep Report Intelligence" by Shenzhen Creative Wisdom Port, providing innovative solutions for media professionals [7] - "Smart Business ChatMall" by Zhuhai Wanda Smart Business, optimizing resource utilization in commercial real estate [7] Group 5: Industry Impact - The event served as a platform for showcasing filing achievements and promoting collaboration, aiming to boost industry confidence and facilitate the matching of supply and demand in AI application scenarios [8]

Core Insights - The market for large model integrated machines is expected to experience explosive growth from 2024 to 2025, with rapid market expansion anticipated [1][3] - The demand for large model integrated machines is projected to reach 150,000 units in 2025, 390,000 units in 2026, and 720,000 units in 2027, with the market size expected to exceed 500 billion RMB by 2027 [3][4] - The current industry landscape shows that 34% of companies have launched only inference integrated machines, while 17% have launched only training integrated machines, and 48.9% have launched both types [3][4] Industry Trends - The primary focus of the industry is on inference integrated machines, as many companies prefer to utilize existing models for application development rather than training their own [3][4] - There is a growing demand for specialized devices that integrate industry knowledge and optimize workflows, moving away from generic solutions [3][4] - The market is trending towards industry-specific integrated machines, with 21.3% of companies offering general-purpose machines and 31.9% offering industry-specific machines [4] Challenges and Opportunities - The industry faces challenges such as weak independent innovation capabilities, difficulties in adapting to application scenarios, and the need for improved security and privacy mechanisms [4][5] - Large model integrated machines are seen as a crucial breakthrough for democratizing large model technology and supporting the "Artificial Intelligence +" initiative [5]

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 ...

来源：环球网 【环球网科技综合报道】10月31日，路透社报道称，据三位知情人士透露，人工智能领域的绝对领军者 OpenAI已正式启动其首次公开募股的筹备工作，此次IPO对公司估值可能高达1万亿美元，有望成为全 球有史以来规模最大的上市案例之一，标志着AI产业资本化进程迈入全新阶段。 据悉，OpenAI内部正积极考虑最早于2026年下半年向监管机构秘密递交上市申请。初步讨论的融资规 模底线为600亿美元，且最终金额很可能远超于此。不过，知情人士强调所有计划仍处早期阶段，具体 时间表与融资额将视公司业务增速及市场环境动态调整。 公司首席财务官莎拉·弗里尔已向部分关联方透露，目标是在2027年完成挂牌上市。然而，部分参与顾 问的预测更为激进，认为上市窗口可能提前至2026年底。对此，OpenAI官方发言人保持了一贯的审慎 态度，回应称："IPO并非我们当前的焦点，因此不可能设定具体日期。我们正致力于构建一项持久的 事业并推进使命。" 此次IPO筹备紧随着本周刚刚完成的公司重大重组。新架构下，控股的非营利组织"OpenAI基金会"持有 集团26%股权及未来增持期权，既保留了其核心治理角色，又能从公司的商业成功中直接获 ...

Core Insights - The evolution of AI products is transitioning from Chatbots to AI Agents, which are expected to become the mainstream form of AI applications in the next phase [1][8] - The adoption rate of AI Agents in enterprises is projected to increase significantly, driven by commercial value [1][8] - Major tech companies like OpenAI, Google, and Lenovo are actively developing and upgrading their AI Agent products, leading to commercial success and enhanced corporate value [1][2][8] Industry Trends - AI Agents are facilitating a shift in enterprise budgets from "buying tools" to "buying results," creating favorable conditions for the growth of the Agent ecosystem [1][7] - The global AI Agent market is expected to see substantial growth, with revenues projected to exceed $5 billion in 2024 and reach $10 billion by 2025, with a CAGR of 44.9% from 2024 to 2032 [8][10] - The demand for AI Agents is driven by their ability to enhance productivity and deliver measurable ROI, particularly for B2B clients [4][6][11] Company Developments - Lenovo has launched a comprehensive upgrade of its AI Agents, introducing three major types of super Agents aimed at personal, enterprise, and urban applications [2][5] - Lenovo's enterprise AI Agent has generated 1.89 billion yuan in revenue within six months, demonstrating significant user engagement and improved conversion rates [10] - OpenAI's investment in the startup Cursor has led to a remarkable achievement of $100 million ARR within 12 months, highlighting the rapid growth potential in the AI Agent space [10] Competitive Landscape - The competition among AI Agent providers is intensifying, with companies needing to lower deployment costs and enhance efficiency to maintain market position [12][15] - A shift towards a results-based payment model (RaaS) is emerging, where companies pay based on the actual business outcomes delivered by AI Agents [11][12] - IDC reports indicate that 70% of enterprises are considering changing or adding AI platform suppliers, reflecting a dynamic market environment [10][11]

格隆汇10月31日丨云从科技(688327.SH)在投资者互动平台表示，公司是一家专注于提供高效人机协同 操作系统及行业解决方案的人工智能企业，目前公司的业务和技术布局主要聚焦于人工智能领域，不直 接涉及量子科技相关业务。 ...

Core Insights - SuperX AI Technology Limited has formed a joint venture with Huasheng Tiancheng to establish SuperX Global Service Pte.Ltd. in Singapore, aimed at providing end-to-end professional services for SuperX's global AI factory project [1][2][3] Industry Overview - The demand for computing power has surged exponentially due to the rise of generative AI like ChatGPT, leading to the emergence of modular and integrated AI data centers (AIDC) as essential engines for AI model training and inference [2][3] - The competition in the AI infrastructure sector has shifted from merely providing advanced hardware to offering comprehensive service capabilities, including deployment, management, and support [2][3][6] Company Strategy - The partnership between SuperX and Huasheng Tiancheng represents a strategic win-win, combining SuperX's innovative products with Huasheng's extensive global service network [3][4] - The joint venture will offer four core services: Global Contact Center, Deployment Service, Maintenance Service, and Managed Service, enhancing customer support and operational efficiency [4][5] Business Model Transformation - The establishment of SuperX Global Service marks a strategic elevation for SuperX, transitioning from a product supplier to a full lifecycle partner, thereby enhancing customer loyalty and long-term revenue certainty [5][6] - The joint venture aims to create a comprehensive AI infrastructure ecosystem focused on customer value, positioning SuperX as a leading AI infrastructure service provider in the Asia-Pacific region [6]

AI Industry Insights - Industry leaders, including NVIDIA's founder Jensen Huang and 360 Group's founder Zhou Hongyi, suggest that AI has transcended its role as a mere tool and should be viewed as a "virtual human" with labor value, indicating a fundamental shift in human-machine collaboration [2] - The global AI regulatory landscape is evolving, with China emphasizing its "Artificial Intelligence +" strategy and building a data market, the EU advancing the implementation of the AI Act, and the UK launching innovation support projects [2] - The digital human industry is undergoing a transformation driven by the explosion of large model technologies, with companies lacking AI R&D capabilities facing elimination, and the trend shifting towards platform development [2] - The AI wearable device market is transitioning from simple monitoring to proactive health management, with an expected market size of $304.8 billion by 2033, integrating AI analysis with digital healthcare services [2] - Chinese large models are gaining traction in Silicon Valley, with several prominent AI companies publicly acknowledging the cost-effectiveness of Alibaba's Qwen and Zhiyuan's Glm, marking China's AI technology system as a significant force in global AI development [2] AI Startup Valuations - AI companies are reportedly caught in a "burn rate cycle," raising concerns about potential over-investment, with OpenAI leading at an estimated valuation of $500 billion, primarily due to substantial expenditures aimed at maintaining its technological moat [3] Cloud Services Updates - Alibaba Cloud has officially launched services in Malaysia, becoming the first international cloud provider to offer cloud computing and AI services in the region, aimed at supporting digital transformation for SMEs and Chinese enterprises [4] - AI is reshaping cloud infrastructure in four key areas: heterogeneous computing demands, AI-enabled operations, enhanced security, and optimized resource allocation [4] - Amazon Web Services (AWS) has been recognized as the leader in the global public cloud IaaS market, with advantages in infrastructure coverage, self-developed chips, network innovation, and high security standards [4] - Amazon plans to invest $100 billion in AI infrastructure by 2025, focusing on data center innovations and self-developed chips [4] - Competition in the cloud market is shifting towards ecosystem building, with leading cloud providers moving beyond mere technology or pricing competition to enhance customer loyalty through strategic partnerships [4] - The Chinese AI cloud market is characterized by differentiated competition, with major cloud providers pursuing various strategies, such as Alibaba Cloud emphasizing overall scale and Volcano Engine leading in large model services [4] Cybersecurity Developments - Over 60 countries signed the first global convention on combating cybercrime in Vietnam, aimed at establishing an international framework for collecting and sharing electronic evidence to combat phishing and ransomware [5] - State-sponsored hackers infiltrated the internal network of U.S. telecom supplier Ribbon Communications, remaining undetected for over a year [6] - OpenAI launched the AI security analysis agent Aardvark, powered by GPT-5, which can autonomously analyze codebases, identify vulnerabilities, verify exploitability, and generate patches, currently in private testing [6] - Social engineering has emerged as the primary threat in the cryptocurrency sector, accounting for 40.8% of all security incidents projected for 2025 [6] - Australian Federal Police successfully decrypted a cryptocurrency wallet containing $6.4 million during operations against criminal groups using encrypted communication networks [6] - Singapore's revised Cybersecurity Act has come into effect, introducing regulations for third-party critical infrastructure and establishing a temporary regulatory framework for critical systems [6]

Core Insights - The article provides a comprehensive overview of the emerging field of Vision-Language-Action (VLA), highlighting its rapid growth and significance in AI and robotics [1][5]. Summary by Sections VLA Overview - VLA models have seen a dramatic increase in submissions, rising from single digits to 164, marking an 18-fold growth [5]. - A model qualifies as VLA if it uses a pre-trained backbone on large-scale visual-language data, emphasizing capabilities in language understanding, visual generalization, and task transfer [5][6]. Key Trends in VLA - **Trend 1: Efficient Architecture Paradigm** Discrete diffusion models are emerging as a new paradigm, allowing for parallel generation of action sequences, enhancing efficiency and integrating reasoning with actions [7][10]. - **Trend 2: Embodied Chain-of-Thought (ECoT)** ECoT emphasizes generating intermediate reasoning steps before actions, improving planning and interpretability, although it relies heavily on high-quality annotated data [11][12]. - **Trend 3: Action Tokenizer** The action tokenizer converts continuous robot actions into discrete tokens that VLMs can understand, bridging the gap between the robot's actions and the VLM's processing [14][16]. - **Trend 4: Reinforcement Learning (RL)** RL is reintroduced to fine-tune VLA strategies, addressing limitations of imitation learning in extreme scenarios, with notable successes in recent studies [17][18]. - **Trend 5: Efficiency Optimization** Efforts are being made to reduce the hardware requirements for VLA models, making the field more accessible to smaller research labs [19]. - **Trend 6: Video Prediction for Physical Intuition** Video generation models provide inherent understanding of temporal dynamics and physical laws, enhancing robot control capabilities [20][23]. - **Trend 7: Realistic Evaluation Benchmarks** New evaluation frameworks are being developed to overcome the limitations of existing benchmarks, focusing on meaningful generalization capabilities [24][26]. Challenges and Future Directions - The article highlights the "performance ceiling" issue in mainstream simulation evaluations, where high scores do not necessarily translate to real-world capabilities [30]. - Two critical areas needing more attention are data quality and in-context learning, which could be pivotal for advancing VLA research [31].