A court document reveals Meta's revenue-sharing agreement with Llama AI model hosts, marking a significant shift in AI collaboration. This model promotes innovation, fairer value distribution, and accelerated adoption of AI technologies, potentially reshaping the industry by fostering a more collaborative and equitable ecosystem for all stakeholders involved in the development and deployment.
NVIDIA's GTC conference showcased groundbreaking AI advancements, including humanoid robots in manufacturing and massive AI data centers. The expanded Microsoft partnership integrates NVIDIA's generative AI and Omniverse' technologies across Azure, Microsoft 365, and healthcare, accelerating AI deployments and solidifying Microsoft's position at the forefront of the AI revolution, boosting productivity and security.
At GTC 2025, Nvidia's CEO Jensen Huang unveiled the company's next-generation AI hardware, Blackwell Ultra and Vera Rubin, and outlined a vision of 'AI factories' powering a future driven by agentic AI and robotics. The keynote detailed advancements in computing, shifting from data centers to specialized AI infrastructure, and highlighted the transformative potential of AI across industries.
Oracle's surprising purchase of 30,000 AMD Instinct MI355X AI accelerators signals a potential shift in the AI chip market. This multi-billion dollar deal challenges Nvidia's dominance and highlights Oracle's focus on efficiency. Despite this, Oracle's commitment to Nvidia's Project Stargate, a 64,000-GPU cluster, remains, showcasing a dual-vendor strategy in the evolving AI landscape.
Nvidia's CEO, Jensen Huang, expressed surprise at publicly traded quantum computing companies, causing a stir in the market. His remarks highlighted the industry's early stage and the long road ahead for practical quantum computers, emphasizing the contrast between long-term potential and short-term investor expectations in this volatile, high-stakes field.
Tencent's Hunyuan-T1, a new AI reasoning model, rivals top systems like OpenAI's. Built with reinforcement learning and human alignment, it excels in benchmarks like MMLU-PRO, GPQA-diamond, and MATH-500. Using the Transformer Mamba architecture, it boasts fast text processing. Its release intensifies the global AI competition, particularly with other Chinese tech giants.
Accenture unveils a no-code AI agent builder on its AI Refinery platform, powered by NVIDIA AI Enterprise. This empowers businesses to rapidly create, customize, and deploy AI agents, accelerating AI adoption and driving value. It features industry-specific solutions, continuous learning, and proactive action, transforming business processes with generative AI and agentic technology.
I tested Gemini, Copilot, and ChatGPT to see if they could find new and interesting Android apps, beyond the usual top-chart suspects. I challenged them with weather, note-taking, and gaming app requests, revealing their strengths and weaknesses in app discovery, and the importance of crafting precise prompts for optimal results.
This week's AI roundup explores Apple's delayed 'Apple Intelligence' rollout and its privacy-centric approach, contrasting it with Cohere's readily available Command R model and its competitive advantages. We also delve into the rise of 'Sovereign AI', the data center question, and the perils of 'vibe coding', urging caution in AI-powered code generation.
China deploys 'Futang·Baichuan,' a pioneering pediatric AI model, to enhance healthcare in grassroots and county-level hospitals. This AI pediatrician, developed by Beijing Children's Hospital with Baichuan AI and Xiaoerfang, offers Basic and Expert versions, improving diagnostic accuracy, addressing pediatrician shortages, and promoting equitable access to quality pediatric care across China, especially in underserved regions.
