Anthropic and Menlo Ventures launch $100M Anthology Fund to support AI startups in healthcare, education, and more. Fund focuses on AI infrastructure, cross-industry applications, security, and social impact. Startups get funding, access to AI models, and networking opportunities.

HP reboots the classic OmniBook series with the new OmniBook Ultra featuring AMD Ryzen AI 300 processor. It offers improved performance, 32GB RAM, Thunderbolt 4, and at $1,449.99. OmniBook Ultra aims for compatibility with Windows apps and boasts 55 TOPS NPU performance compared to Snapdragon X Elite-powered OmniBook X AI. The Ultra model has a 2240x1400 touchscreen, upgraded webcam, 4-speaker system, and 68Wh battery for 13 hours of usage. HP plans to provide Copilot Plus features through free

Microsoft researchers introduced SpreadsheetLLM, enhancing AI to process spreadsheets. Bridging the gap, SpreadsheetLLM enables AI-assisted data analysis and decision-making, democratizing data insights with natural language processing. Microsoft's investment in enterprise AI includes SpreadsheetLLM for improved productivity and Copilot for Finance. Despite AI challenges, tools like SpreadsheetLLM drive innovation and create more value for enterprises. Excel's future with SpreadsheetLLM looks pr

The European Union's "AI Act" sets deadlines for comprehensive AI regulations aiming to prevent high-risk AI threats to democracy, human rights, and the environment. Developers must comply by 2026, with penalties up to €35M or 7% annual revenue for non-compliance.

Google is in talks to acquire cloud security startup Wiz for $23 billion, aiming to boost its position in the cloud computing market. The deal, led by Google Cloud CEO Thomas Kurian, could enhance Google's cloud security offerings. Wiz, founded by former Microsoft employees, provides a platform for securing applications in various cloud environments. This move reflects Google's commitment to strengthening its cloud security capabilities, especially as businesses prioritize cybersecurity in AI ap

Japan's Ministry released the "White Paper on Generative AI Applications in 2024", comparing top countries. China dominates in public and enterprise usage, Japan lags due to lack of education. Generative AI finds diverse applications but poses risks. Ministry of Internal Affairs and Communications plays a vital role in promoting innovative technology development.

NVIDIA will disclose latest research in rendering, simulation, and generative AI at Siggraph 2024. Highlights include faster story image generation, 2D generative diffusion models for texture painting, physics-based simulation advancements, new renderer for engineering design, and improved rendering technology like ReSTIR for autonomous driving radar simulation. AI tools for 3D representation and design, such as fVDB framework, will also be featured.

The 8th International Conference on Smart Agriculture showcased China Agricultural University's "Shennong Model 2.0," a cutting-edge innovation enhancing breeding, planting, remote sensing, and meteorological forecasting. It provides crucial technical support for modern agriculture's transformation and upgrade.

Hugging Face introduces SmolLM, compact language models outperforming Microsoft, Meta, and Alibaba's Qwen. With models ranging from 135M to 1.7B parameters, SmolLM excels in benchmark tests for common-sense reasoning and world knowledge mastery. Its high-quality training data and focus on privacy make AI more accessible while reducing costs and environmental impact.

AI researcher Andrej Karpathy launches Eureka Labs, an AI-driven education company transforming learning experiences with personalized, scalable education. Flagship AI course LLM101n integrates AI in storytelling. Eureka Labs combines high-quality materials with AI tutors to enhance learning.

Apple, Nvidia, and Anthropic used YouTube subtitles to train AI models despite policy violations, as revealed by Proof News and Wired. This dataset includes subtitles from 173,536 YouTube videos from 48,000 channels, raising concerns about privacy and copyright. OpenAI's GPT-4 model also utilized YouTube videos for training, leading to legal disputes in the AI industry.

Microsoft is promoting Copilot in OneNote, offering intelligent handwriting processing capabilities. Users can enjoy automatic summarization, intelligent questioning, and to-do list generation. Copilot accurately converts handwritten notes to editable text, showing high accuracy and flexibility. It retains original intentions and offers instant text conversion for to-do lists. Despite handling complex handwriting well, it may face challenges with extremely difficult fonts. To experience Copilot,

Large language models (LLMs) excel at simple questions but struggle with complex tasks. "System 2" prompts enhance reasoning but slow processing. Meta FAIR's "System 2 distillation" bypasses limitations, improving performance in complex tasks.

Meta is set to launch Llama 3-405B, a massive open-source AI model with 4.05 trillion parameters on July 23. The model, designed for multimodal processing, shows promise in various domains like language processing and image recognition. Early user feedback confirms its potential in real-world applications.

OpenAI faces continued safety concerns amid claims of rushed security testing and prioritizing product launches over safety. Despite safety being central to its charter, recent controversies highlight recurring security issues. Public relations efforts may not be enough to address growing criticism of OpenAI's safety practices.

Amazon's AI shopping assistant Rufus now fully launched on the Amazon app in the US. Users can ask questions, compare products, get order updates, and receive recommendations. Rufus uses Amazon's data to provide accurate responses and track user behavior.

At VB Transform 2024, OpenAI's product manager discussed cost reductions in GPT-4o, highlighting affordability and industry trends. Optimizing costs enables new use cases and strengthens customer relationships..BackgroundImageLayout`

Hugging Face, a leading AI platform, announces profitability despite offering free and open-source services. CEO Clement Delangue credits the 220-member team for this milestone. The company's model combines free services with monetization for enterprise clients. Additionally, Hugging Face raised $235 million in Series D funding and launched an open-source Large Language Model leaderboard. They also pledge $10 million in shared GPU resources for developers.

OpenAI's secretive "Strawberry" project aims to develop AI with advanced reasoning abilities, signaling a potential breakthrough. The project builds on the Q* project, showcasing OpenAI's commitment to enhancing AI reasoning. This initiative, along with industry giants' efforts, highlights the importance of improving model reasoning capabilities amidst ongoing debates about AI's human-level reasoning. Experts foresee a transformative impact if successful, emphasizing the need for ethical conside

In recent years, MoE technology boosts Large Language Models (LLMs) efficiency without added computational burden. Google DeepMind's PEER architecture addresses limitations, supporting millions of experts for optimal performance and parameter efficiency. PEER's innovative expert retrieval system enhances MoE potential and improves model adaptability and optimization. PEER's integration with PEFT technology allows dynamic knowledge addition to LLMs, leading to significant performance gains and re