Meta Corporation unveiled V-JEPA, a video model learning from videos to understand the physical world. It builds on I-JEPA, focusing on high-level semantics for efficient learning. V-JEPA surpasses existing models, advancing AI's ability to interpret the world, with potential applications in various fields.

Apple expands Xcode testing for generative AI features, plans third-party release. Apple explores using generative AI in consumer products like Apple Music playlists, Keynote slides, and Spotlight search. Microsoft's GitHub Copilot reaches one million paid users, adding features like real-time chat. Apple CEO Tim Cook confirms launch of more generative AI features this year.

Amazon AI researchers have developed the largest text-to-speech model, "Big Adaptive Streaming TTS," with 980M parameters. Trained on 100K hours of speech data, it aims to enhance voice quality and language attributes without public release due to ethical concerns.

The US PTO rejected OpenAI's trademark application for "GPT" as too generic, hindering competition. GPT has become a common term in generative AI, used by various companies for their AI models. OpenAI has started using unique brand names like "Sora" for new services.

NVIDIA introduces a local chatbot app "Chat with RTX" for Windows PC. Built on TensorRT-LLM, requires RTX 30 or 40 series GPU with 8GB VRAM, 16GB memory, Windows 11, GeForce driver 535.11+. Upcoming TensorRT-LLM v0.6.0 promises 5x performance boost. Assist in data access, issue resolution, entertainment interactions.

OpenAI, creator of ChatGPT, doubles value to $80B+. Deal with Thrive Capital allows favorable share sales post crisis and new AI tool "Sora" released. Microsoft's investment in AI tech intensifies.

Google launches AI cybersecurity initiative to combat cybersecurity threats, leveraging AI capabilities. The initiative aims to automate threat prevention, simplify incident response, and build an "immune system" for enterprise infrastructure. Google invests billions in AI adoption and open-sources tools like Magika for malware detection. Proposed three-pillar roadmap for policymakers to ensure AI security and drive research collaboration in AI-driven defense technologies.

Details about chips in the upcoming iPhone 16 suggest a significant AI boost. A18 and M4 chips will enhance AI computing cores. Apple to unveil generative AI focus in 2024. iOS 18 brings new features and performance improvements. Compatibility may vary.

Google rebrands AI chatbot Bard to "Gemini," introduces Gemini Advanced with Ultra 1.0 model. New mobile apps for broader access. Subscribers get advanced features for $19.99 monthly. Launching in US, then globally.

Stability AI unveils AudioSparx 1.0, the latest text-to-music AI model supporting Stable Audio. Capable of producing high-fidelity, long-form stereo music with detailed structures. Also generates 44.1kHz stereo sound effects. Innovative tool for audio creators.

NVIDIA launches new business unit for customizing processors for AI and major companies. Led by Dina McKinney, the team will develop custom chips for cloud computing, gaming, and more. Market growth expected with focus on data center custom chips. NVIDIA aims to be one-stop-shop for custom chip solutions.

Microsoft is enhancing Windows 11 with more AI features, including Copilot integration for text and image summarization. Notepad app will also feature Copilot AI for text and code explanations. Windows 2024 aims to be the "AI PC year" with dedicated Copilot key on keyboards.

Microsoft rebrands Microsoft 365 Chat in Teams to Copilot, offering advanced AI capabilities. New features include personalized Copilot text suggestions and a Copilot Lab with a rich library of suggestions. Users are advised to update to the new app for enhanced functionality. Microsoft Teams app store has over 2,000 apps with 320 million global monthly active users. Additional features like 3D avatars and virtual spaces have been introduced.

OpenAI introduced a "memory" feature in ChatGPT to enhance user communication. Users can control what ChatGPT remembers and disable or "forget" memories. Usage will improve memory capability. User data can be used to improve the model, but this can be disabled. Users can also use the "temporary chat" feature to ensure privacy.

OpenAI is secretly working on its own web search service with technical support from Microsoft's Bing. This move intensifies competition with Google, leading to a drop in Google's stock price. Microsoft aims to challenge Google's dominance in search with OpenAI's technology.

OpenAI introduced Sora, a cutting-edge generative AI model that can transform text into high-quality videos up to 1080p resolution. Sora can create vivid videos with rich characters and actions, and even "extend" existing clips intelligently. Despite its impressive capabilities, OpenAI recognizes Sora's limitations in simulating complex scenes. They are working with experts to ensure safe use and exploring positive applications for this innovative technology.

Google recently open-sourced Magika, an AI-based tool for identifying file formats with high accuracy and speed. It outperforms existing tools by 20% and achieves over 99% precision and recall. Internally, Google has enhanced user security by using Magika to improve file type identification by 50%. VirusTotal also integrates Magika to enhance its platform's efficiency. This open-sourcing allows global developers to contribute to optimizing this innovative tool.

Google announced the release of Gemini 1.5, a conversational AI model with an Expert Mixture architecture for efficient complex task execution. It sets a new milestone with 1 million-token context understanding, outperforming previous models. Early access is available on AI Studio.

YouTube CEO Neal Mohan unveiled the platform's 2024 strategy, emphasizing AI, creator empowerment, user safety, and expansion into living room entertainment. The vision leverages AI for creativity, acknowledges creators, and aims to enhance subscriptions and content quality.

Alibaba releases Qwen1.5, a large-scale model with impressive performance surpassing GPT-4. It is integrated into mainstream frameworks like Hugging Face transformers and offers various model sizes. The series excels in benchmarks and long-context support, with plans for further code execution optimization. It also enhances functionality, multilingual capabilities, and system prompts for a more efficient development experience.