May 04 AI Daily — Google AI Healthcare Expansion, MS In-House AI Chips

1. Google Announces Massive Expansion in AI Healthcare and New Partnerships

Google recently announced a large-scale expansion plan in the AI healthcare sector, unveiling new partnerships with several medical institutions and pharmaceutical companies. This expansion focuses on utilizing AI technology to improve disease diagnosis accuracy, accelerate the drug discovery process, and provide personalized treatment solutions. While this is expected to intensify competition in the medical AI market and stimulate investment from other Big Tech companies, data privacy and ethical considerations will become increasingly important.

Source: Daily AI Top3 News

---

2. Microsoft Materializes In-House AI Chip Production Plans and Strengthens Foundry Collaborations

Microsoft announced that it is materializing its plans to produce in-house AI chips and strengthening cooperation with major foundry companies. This is part of a strategy to maximize the AI workload processing efficiency of its cloud service, Azure, and reduce reliance on external suppliers like NVIDIA. By designing its own chips, Microsoft aims to secure AI computing power specialized for the Azure cloud environment, targeting cost reduction and performance enhancement, which could lead to shifts in the competitive landscape of the AI chip market.

Source: Daily AI Top3 News

---

3. NVIDIA Unveils Next-Gen GPU Architecture 'Blackwell Ultra' and Presents AI Supercomputing Vision

NVIDIA officially announced its next-generation GPU architecture, 'Blackwell Ultra', presenting a new vision for AI supercomputing. Blackwell Ultra further elevates the performance of the existing Blackwell architecture, focusing on dramatically improving the computational power required for training and inferencing large language models (LLMs). This holds the potential to accelerate the pace of AI research and development, enable the emergence of new AI applications, and further solidify NVIDIA's dominance in the AI infrastructure market.

Source: Daily AI Top3 News

---

4. Anthropic's Aggressive Expansion: Pursuing a $90 Billion Valuation Funding Round and Wall Street JV

Anthropic is in the process of raising a $5 billion funding round at an $85 billion to $90 billion valuation, potentially surpassing OpenAI's valuation. Additionally, Anthropic plans to sign a $1.5 billion joint venture agreement with major Wall Street firms, including Blackstone and Goldman Sachs. This joint venture aims to sell AI tools to private equity-backed companies and will establish a company to handle Anthropic's consulting arm.

Sources: 지피터스 뉴스레터 · Global AI News Summary  |  🔗 View Original  |  🔗 View Original

---

5. Meta Acquires Humanoid Robot Startup ARI: Strengthening Embodied AI

Meta has acquired Assured Robot Intelligence (ARI), an AI humanoid robot startup. ARI's founders will join Meta's Superintelligence Labs to integrate their robot control and autonomous learning technologies, aiming to strengthen "physical world AI models." This acquisition clearly demonstrates an expansion beyond LLMs into the field of 'Embodied AI' and is evaluated as a strategic move by Meta to explore the applicability of AI in the real world, beyond virtual reality (VR) and augmented reality (AR).

Sources: 지피터스 뉴스레터 · TLDR  |  🔗 View Original

---

6. Google Gemini Broadly Integrated into Cars and Chrome, Accelerating AI Penetration in Daily Life

Google Gemini has begun rolling out to approximately 4 million GM vehicles equipped with 'Cars with Google built-in' and has also officially launched on the desktop and iOS versions of Chrome. While driving, users can conduct natural voice conversations to search for restaurants or summarize messages, and in the web browser, they can perform various tasks such as webpage summarization, translation, and email drafting. Google is also testing 'Gemini Flash' and 'Omni' models to dominate the AI voice assistant market.

Sources: 지피터스 뉴스레터 · Global AI News Summary  |  🔗 View Original  |  🔗 View Original  |  🔗 View Original

---

7. The Rise of Open Source LLMs: DeepSeek V4 Pro Performance Proven and Xiaomi MiMo-V2.5-Pro Released

The DeepClaude project demonstrated the technical maturity of open-source LLMs by proving that DeepSeek V4 Pro can replace core functionalities of Anthropic's Claude Code model at a 17x lower cost. According to benchmarks by the US Center for AI Standards and Innovation (CAISI), DeepSeek V4 Pro is about 8 months behind leading US models but showed nearly equal performance in mathematics. Furthermore, Xiaomi released MiMo-V2.5-Pro, a powerful new open-source model, contributing to the expansion of the open-source LLM ecosystem.

Sources: AI Breakfast · 7min.ai · The Rundown AI  |  🔗 View Original  |  🔗 View Original  |  🔗 View Original

---

8. Evolution of AI Agent Technology: Open Source Platforms Launched and the Importance of 'Skill' Encoding

AI agent technology is gaining attention as the next stage of LLMs. Open-source agent engineering platforms like PandaProbe have been launched, helping developers easily build and deploy agents. Additionally, to prevent AI coding agents from missing nuanced aspects like a senior engineer's intuition, it is becoming crucial to make them autonomous and reliable in professional environments through 'agent skills' or 'discipline encoding'. This expands the frontier of AI engineering into intelligent and context-aware problem-solving.

Sources: AI Breakfast · TLDR  |  🔗 View Original  |  🔗 View Original

---

9. Social Impact of AI: Chinese Court Halts AI Firings and Controversy Over Healthcare Premiums in Kenya

A court in Hangzhou, China, set a precedent against mass layoffs caused by AI by ruling it illegal and ordering compensation for a company that fired quality assurance engineers, claiming an LLM had replaced their jobs. Meanwhile, in Kenya, an issue arose where an AI-based predictive machine learning algorithm overestimated the income of the poor during the process of setting health insurance premiums, leading to higher charges. This exposed the risks of AI algorithm bias and the exacerbation of social inequality.

Sources: 7min.ai  |  🔗 View Original  |  🔗 View Original

---

10. Open Source AI Infrastructure Innovation: Tenstorrent-based LLM Inference and Alibaba FlashQLA Released

Moreh, a company specializing in AI infrastructure, proved that the inference performance of the latest open-source LLMs, such as DeepSeek, is equal to or greater than NVIDIA's 'DGX A100' class by applying its in-house developed 'MoAI inference framework' to Tenstorrent's 'Galaxy Wormhole' system. Along with this, Alibaba open-sourced 'FlashQLA', a high-performance linear attention kernel library that accelerates LLM inference speed by up to 3 times on NVIDIA 'Hopper' GPUs, contributing to increasing the efficiency and diversity of AI infrastructure.

Sources: Global AI News Summary  |  🔗 View Original  |  🔗 View Original

---