24-Hour AI News: Microsoft Breaks Through in "Super Intelligence in Healthcare," China Releases Two Major Marine Models

In the past 24 hours, many places in China have released AI achievements and policy interpretations in conjunction with the World Internet Conference and Ocean Week; Europe is pushing forward the implementation of AI law, and the UK financial regulators have warned of the risk of an "AI bubble"; on the industry side, there have been infrastructure cooperation projects worth over $1 billion and a new round of entrepreneurial financing.

I. Microsoft establishes "Human-Centered Super Intelligence" team to take the lead in overcoming medical diagnostic challenges.

1. On November 6, it was announced that Mustafa Suleyman would lead the effort, with Karen Simonyan as chief scientist, focusing on achieving AI capabilities that "surpass humans" in specific fields.
2. The first key area of focus is early disease screening and diagnosis, with a plan to develop verifiable "medical super-intelligence" within two to three years.
3. The strategic focus has shifted from general big models to "domain super intelligence", emphasizing controllability, security and verifiable social value.

II. The World Internet Conference's "Leading Technology Award" was announced, with AI achievements making the list in large numbers.

1. Awards will be presented in Wuzhen on November 6th, including "Script-Driven Digital Human Technology" (Baidu), "Nanhu Computing Framework" (Zhejiang Lab), and "ROCm Open Source AI Platform" (AMD).
2. The engineering R&D team has included industrialization achievements such as "GitHub Copilot: A New Generation of Programming Intelligent Agents".
3. The awards focus on open source ecosystems, computing power frameworks, and embodied intelligence, presenting a dual-track approach of "application-driven + underlying breakthroughs".

III. Two AI-powered marine models were released at the Xiamen International Ocean Week.

1. On November 6, the “Deep-Sea Habitat Intelligent Cognition and Exploration Multimodal Large Model (DePTH-GPT)” and the “Kanhai” full-link marine AI large model were released.
2. To enhance the ability to fuse and extrapolate multi-source data for key tasks such as deep-sea habitat cognition, remote sensing-reconstruction-prediction, etc.
3. Focus on the public scientific and technological achievements of the UN "Decade of Oceans" project to address the scarcity of deep-sea data and the challenges of prediction.

IV. Google is reportedly in talks to increase its investment in Anthropic, potentially raising its valuation to around $350 billion.

1. On November 6, multiple media outlets reported that Google was in "early-stage talks" and might deepen cooperation through capital increases or convertible bonds.
2. Rumors suggest a significant increase in the valuation range for the new round, which may reshape the "cloud + model" binding pattern.
3. The "triangular competition" between Amazon and Microsoft surrounding computing power and model ecosystems has intensified further.

V. Vast Data and CoreWeave enter into a $1.17 billion long-term partnership.

1. On November 6, CoreWeave announced a multi-year, $1.17 billion deep partnership, with Vast as its main data platform.
2. Optimization of training and inference data paths for GPU cloud, billed based on capacity and functionality.
3. Signaling a large-scale demand for the integration of "data, computing power, and network" on the AI infrastructure side.

VI. Bank of England Governor Warns of "AI Bubble" Risk

1. On November 6, it was stated that while output efficiency is improving, the capital market may be overly optimistic in pricing AI returns.
2. It is advisable to be wary of the chain reaction of "high valuation - high expectations - high volatility".
3. This echoes recent market discussions regarding the pullback in the AI sector.

VII. Hungary establishes a national-level enforcement framework for the EU AI Act.

1. On November 6, it was announced that two laws and supporting decrees were passed, clarifying the division of labor among national-level institutions and the enforcement path.
2. To convey the message of "proactive implementation and broad applicability," enterprises need to accelerate model classification, risk management, and compliance documentation development.
3. Provide a reference template for the practical implementation in member states.

8. The China "Light of the Internet" Expo opens, showcasing embodied robots and intelligent agents.

1. The event opened in Wuzhen on November 6th, and for the first time, humanoid robots were introduced to participate in the opening ceremony and interact with the smart bracelet.
2. Build immersive spaces such as the BOT SHOW and Phantom Cube to promote AI science to teenagers.
3. Showcase cross-industry applications such as "AI + Culture" and "AI + Rehabilitation" to enhance public awareness and connect with the industrial ecosystem.

IX. The China Academy of Information and Communications Technology (CAICT) and industry alliances released the "Research Report on the Application of Large-Scale All-in-One Machines (2025)".

1. The report estimates that the market size of large-scale all-in-one model machines in China is expected to increase from approximately RMB 123.6 billion to over RMB 500 billion between 2025 and 2027.
2. The expected demand has increased from 150,000 units to 720,000 units, and a multi-layered supply system of "hardware-model-toolchain-scenario" has been formed.
3. Mainstream scenarios cover intelligent customer service, search, data analysis and content generation, emphasizing the requirements of "low latency, privacy protection and localization".

10. New Trends in Entrepreneurship and Modeling: Funding for Code/Text Diffusion Models, Release of Open-Source Speech Models

1. Inception completes $50 million in funding, betting on the "code and text" cross-domain diffusion model to accelerate product rollout.
2. Israel's aiOla released the open-source speech model Drax, which uses a flow-matching generation paradigm and claims to have leading accuracy in real-world speech and inference speed up to 50% faster.

Frequently Asked Questions (Q&A)

Q: What is the difference between Microsoft's "Human-based Superintelligence" and "Artificial General Intelligence"?

A: The former is "domain-specific superintelligence," focusing on achieving verifiable capabilities surpassing human abilities in specific scenarios such as medical diagnosis; the latter pursues cross-task versatility. Microsoft's timeline is to achieve groundbreaking results in the medical field within 2-3 years.

Q: What is the significance of a large-scale multimodal model of deep-sea habitats?

A: It integrates multi-source data such as deep-sea images, topography, and sensors into the "cognition-deduction-prediction" chain, serving the discovery and protection of typical habitats such as seamounts and hydrothermal vents, and is a public science and technology output of the United Nations "Decade of Oceans".

Q: Why is the large-scale all-in-one PC growing so rapidly in China?

A: Government affairs, finance, manufacturing and other sectors require AI capabilities that are "low latency, localized and have strong privacy"; the report predicts that the market size is expected to exceed 500 billion yuan in 2027, with demand reaching 720,000 units.

Q: What do the latest developments in the EU AI law mean for businesses?

A: Hungary has established a national implementation framework, meaning that member states have entered the "real implementation" phase. Companies need to complete compliance actions such as systematic risk classification, data governance, and record keeping as soon as possible.

Q: What new trends have emerged in AI infrastructure?

A: The signing of multi-year, billion-dollar partnerships between cloud service providers and data platform vendors (such as Vast-CoreWeave) highlights the importance of integrated collaboration between "data, computing power, network, and storage".