Meta has launched Leffa, an open-source AI virtual fitting framework designed to enhance online shopping experiences. By allowing users to generate new outfit visuals from reference images, Leffa aims to provide a more detailed and less distorted virtual fitting option, potentially reducing return rates for e-commerce.

Meta Platforms has formally requested California Attorney General Rob Bonta to intervene against OpenAI's transition from a nonprofit to a for-profit entity, arguing it could undermine the integrity of nonprofit organizations and set a troubling precedent in the tech industry.

Oracle has announced a collaboration with Meta to leverage its AI cloud infrastructure and develop AI agents based on the Meta Llama model. The partnership comes as Oracle's cloud revenue surges, driven by rising demand for AI services. Despite competition from Amazon and Microsoft, Oracle aims to expand its position in the multi-cloud market.

Meta has launched Llama3.370B, the latest in its AI model series, boasting performance improvements over GPT-4 and Gemini1.5Pro. The model is designed to be more cost-effective and is available for download, reflecting Meta's strategic push in the generative AI space. The company also revealed plans for a major investment in AI infrastructure to support future developments.

Meta has announced plans to require 1 to 4 gigawatts of nuclear power by 2030 to support its artificial intelligence initiatives. The company will issue a request for proposals to find suitable developers in the U.S. This strategy aims to enhance its energy sustainability while addressing the significant power demands of AI technologies.

Meta is establishing an AI-optimized data center in Richland Parish, Louisiana, projected to create over 6,500 jobs, including direct and indirect employment opportunities. The project aims to transform the local economy and support workforce development initiatives, with completion expected by 2030.

Meta, the parent company of Facebook, is set to invest billions in an AI data center in northern Louisiana, projected to create hundreds of jobs and enhance the state's technological infrastructure.

Carnegie Mellon University and Meta have developed VQAScore, a novel evaluation method for text-to-image models that significantly improves accuracy over traditional evaluation techniques. By utilizing Visual Question Answering (VQA), VQAScore assesses the performance of generative AI models more effectively, paving the way for better AI evaluation standards.

**Summary** 1. Zuckerberg announced that Meta is promoting the use of Llama AI within the U.S. government, with cooperation already underway with the State Department. 2. These AI technologies will be used to improve areas such as safe water, reliable electricity, and student grant applications, with no financial transactions involved. 3. Meta is also preparing to release Llama 4, which promises to be faster and more powerful than anything competitors have. 4. Meta continues to grow, with a 19% revenue increase in the third quarter and a net profit of $17.3 billion.

**Summary** 1. MarDini is Meta’s latest video generation model, developed in collaboration with KAUST. 2. It excels at tasks like video interpolation, image-to-video generation, and video expansion. 3. MarDini uses innovative planning and generation models to create high-quality videos faster and more efficiently. 4. It can be trained from scratch using unlabeled video data, making it extremely flexible. 5. MarDini is more cost-effective and time-efficient than many other video generation models on the market.