Researchers developed PhysDreamer, a technology that uses AI models to optimize material properties for realistic object interactions in virtual environments.

Leading IT service provider Cognizant and Microsoft collaborate to advance generative AI in enterprises, integrating Copilot technology with Cognizant's expertise to deliver tailored solutions across industries. CEO Ravi Kumar S. invests $1 billion in generative AI, aiming to revolutionize sectors like healthcare and retail. The partnership underscores Microsoft's push for AI adoption, emphasizing the pivotal role of technology in shaping future work models.

Meta recently released Llama 3, an advanced language model with 8B and 70B parameters. It outperforms competitors like Gemma and Gemini. Llama 3 has increased tokens to 128,000 for efficient sequence compression. Future models with over 400B parameters are in development. NVIDIA anticipates Llama 3 400B+ will revolutionize research and startups. Get ready for a breakthrough in the AI ecosystem!

Meta AI, an AI assistant by Meta, is integrated into multiple apps. Supported by Llama 3, it offers various functions like meal planning and exam prep. Users can access it within apps without leaving. Image generation is enhanced, now in real-time. Rolling out in more countries.

Microsoft Research team unveils VASA-1: an AI model for lifelike facial videos from images and audio. Core features include diffusion-based facial dynamics and head motion generation in a learned facial latent space. VASA-1 excels in lip sync, realism, and adaptability, offering real-time processing at 40 FPS. Researchers highlight applications in education, accessibility, and healthcare.

Blockade Labs updates Skybox AI with 8K resolution, attracting 1.5M users including NVIDIA, Accenture, and T-Mobile. CEO deCourcelle highlights collaborations with Apple Vision Pro for mixed reality art. The company envisions AI-driven creation of immersive environments with user-friendly tools for professional-grade content generation.

Groq's novel Tensor Streaming Processor achieves lightning-fast AI inference speeds of over 800 tokens per second with LLaMA 3, challenging Nvidia's dominance in the market. This efficiency is crucial for latency-sensitive applications and could significantly reduce power costs for running large neural networks.

Meta released the open source Llama 3 series of large models, prompting Alibaba Cloud to offer full support and a free training service to help users create exclusive models easily. This move reflects Alibaba Cloud's commitment to providing efficient cloud services and promoting industry innovation.

London-based ElevenLabs partners with HarperCollins to use AI voice tech for foreign language audiobooks. Their natural-sounding text-to-speech tool enhances book accessibility, offering a seamless listening experience. CEO Staniszewski expresses excitement for expanding audiobook options, breaking language barriers, and reducing production costs. HarperCollins emphasizes maintaining voice actor-produced audiobooks alongside AI-generated ones for backlist series in non-English markets.

Dell Technology launches APEX File Storage for Microsoft Azure, integrating PowerScale OneFS system. Provides high-performance file storage, scalability, and data protection. Dell Support Services ensure customer satisfaction. Collaboration with Microsoft enables innovation and transformative AI results.

Meta unveils Llama 3 models (Llama 3 8B, Llama 3 70B) "significant leap" in performance compared to Llama 2. Models excel in various benchmarks, offer high 'controllability' and accuracy, improved data filtering and safety suites, but still face scrutiny over data sources and IP rights. Models available for download, to be hosted on various cloud platforms. New, more powerful Llama 3 models with over 400 billion parameters in training to be multilingual, multimodal.

Boston Dynamics introduces a new all-electric version of Atlas, its humanoid robot, with improved power and motion capabilities. Hyundai Motor Company will be among the first to test its real-world applications.

Nothing plans to integrate ChatGPT into smartphones and earbuds for improved customer service. The rollout starts on April 18th with Phone 2. Entry points for ChatGPT will be embedded into the system for enhanced user experience. New AI devices like Humane AI Pin and Rabbit R1 are also on the horizon.

Meta introduces real-time AI image generation feature on WhatsApp for US users. Meta AI transforms text prompts into dynamic images. Meta Llama 3 model generates clearer images, can turn provided images into GIF animations. Feature also available on Meta AI web version.

Boston Dynamics' new electric Atlas robot represents a significant advancement in humanoid robotics, featuring advanced electric technology for enhanced mobility and power. In collaboration with Hyundai Motor Group, it aims to revolutionize practical applications and commercialize robot technology.

Baidu Wenku unveils cross-modal AI comic & picture book creation, revolutionizing content creation with AI. Users can create content in minutes by speaking a sentence. Strategic partnership with Xuanji Technology aims to create new comic IPs based on AI technology. Public beta testing available on Baidu Wenku platform.

Israeli startup Mentee Robotics, founded by Amnon Shashua, unveils its humanoid robot Menteebot. Integrating AI technologies like ChatGPT, it aims at home and warehouse tasks. Menteebot excels in understanding human instructions, mapping 3D environments, and executing tasks via simulation. Scheduled for deployment in Q1 2025, Mentee aims to compete with other AI-driven humanoid robot developers.

The Logi AI Prompt Builder software window streamlines AI usage for Logitech users, offering text summarization and query suggestions. An AI-equipped M750 mouse model will provide instant access to new features in the US and UK for $49.99 and £54.99. Compatibility with ChatGPT currently, but integration with other chatbots planned. Accessible for Logitech keyboard or mouse users with Logi Options+ application support from April 17.

Stability AI previews Stable Diffusion 3 for text-to-image, collaborating with Fireworks AI for dual model support. They also introduce Stable Assistant Beta for content creation. Both models are in preview stage, not fully open to public.

Recently, the Allen Institute for AI announced an update to its OLMo 1.7-7B language model, enriching it with the Dolma dataset. The model now supports longer contexts and improved performance, utilizing the extensive Dolma 1.7 dataset with quality filtering and deduplication features. It also employs a two-stage curriculum for training, surpassing previous models in various tasks.