Everything Announced at Microsoft Build 2024..
++EU Approves AI Law, NVIDIA gives strong forecast..
🚀 AI Breakthroughs
Microsoft Build 2024
1. Three New Phi-3 AI Models
Microsoft has announced new additions to the Phi-3 family of AI models at the Build 2024 conference. The highlights include:
Introducing Phi-3-vision, a new multimodal model in the Phi-3 family.
Phi-3-small and Phi-3-medium models are now available on Microsoft Azure.
Phi-3 models optimize performance across various hardware platforms.
2. Azure AI Studio
Microsoft has launched the Azure AI Studio, aimed at accelerating the development of generative AI applications. Key features include:
Integrated tooling for seamless development.
Azure-grade security and a catalog of over 1600 models.
Support for deploying custom AI copilots with comprehensive monitoring and debugging tools.
3. New Microsoft Copilot Features
The latest update to Microsoft Copilot introduces new features to enhance business productivity and creativity:
Team Copilot evolves to improve productivity and collaboration.
Agents allow customization of copilots in Copilot Studio to automate business processes.
New Copilot extensions and connectors enable easy integration with business systems.
4. Microsoft Copilot Studio Got New Agent Capabilities
At the Build 2024 conference, Microsoft unveiled the Copilot Studio, which features:
Tools for creating custom AI copilots.
Capabilities for handling complex, long-running tasks across various industries.
Advanced analytics and enterprise data connectivity enhancements.
5. GitHub Copilot Extensions
GitHub has introduced Copilot Extensions, available in limited beta from May 21, 2024:
Partnerships with Microsoft Azure, Docker, and MongoDB to enhance developer workflows.
Seamless integration in IDEs like VS Code and GitHub.com.
6. Real-Time Intelligence with Microsoft Fabric Now
Microsoft has launched Real-Time Intelligence in Microsoft Fabric, offering:
Combination of Synapse Analytics, Data Activator, and new AI features in preview.
Capabilities for real-time event insight, analysis, and action across industries.
Adobe Launches Firefly-Powered Generative Remove in Lightroom for Enhanced AI Photo Editing
Adobe has unveiled the Generative Remove feature in Lightroom, powered by AI:
Effortless removal of unwanted objects across all devices.
New Lens Blur feature and mobile UI enhancements for faster, easier editing.
NVIDIA Reports Staggering 262% Annual Revenue Growth, Announces 10-for-1 Stock Split
NVIDIA has reported impressive financial results:
First quarter revenue of $26.0 billion, an 18% quarterly and 262% annual increase.
Data Center revenue hit a record $22.6 billion, up 427% from the previous year.
Announcement of a 10-for-1 stock split effective June 7, 2024, and a 150% dividend raise.
⚖️ AI Ethics:
Council Approves Pioneering AI Act Establishing First Global Standards
The Council has approved the first global AI regulations, focusing on:
A risk-based legislative approach to AI development.
Ensuring safe AI development, respecting EU citizens' rights, and stimulating investment.
Prohibitions on high-risk and unacceptable AI practices with stringent enforcement mechanisms.
Global AI Firms Commit to Unified Safety Standards at AI Seoul Summit
AI tech firms globally have committed to unified safety standards:
New 'Frontier AI Safety Commitments' signed by firms from North America, Asia, Europe, and the Middle East.
Focus on governance and public transparency.
Agreements include halting AI development if risks are unmitigable.
🎓AI Academia:
Exploring AI Interpretability Through Claude 3 Sonnet Feature Extraction
Research on Claude 3 Sonnet's feature extraction has revealed:
Scaling sparse autoencoders to decompose activations.
Features include multilingual, multimodal, and general data representations.
Identification of safety-relevant features such as bias and deception.
Exploring the Enhanced Capabilities of the Phi-3-Mini Language Models
The new Phi-3-mini language model includes:
A transformer decoder with a 4K default context, expandable to 128K.
Utilization of the tiktoken tokenizer, excelling in multilingual scenarios, and optimizing KV cache savings.
Native operation on iPhone 14, achieving over 12 tokens per second with a 4-bit quantized model.
P.S.: We curate this AI newsletter daily for free. Your support keeps us motivated. If you find it valuable, please share it with your friends using the button below!