Meta launches Llama 3

Also: Stability AI Lays Off 10% Of Staff

Author

AI KATANA
April 18, 2024

Hi!

Today’s roundup of significant AI developments starts with Meta's introduction of Llama 3, an advanced AI model offering notable improvements in coding, reasoning, and instruction-following capabilities, soon to be available on major cloud platforms. Google has streamlined its AI operations under DeepMind’s Demis Hassabis to quicken its AI deployment and better compete in the generative AI market. In other news, Stability AI has announced a 10% workforce reduction amid internal challenges and a recent leadership shake-up, reflecting ongoing pressures in the AI startup ecosystem. Meanwhile, Oracle has committed to a substantial $8 billion investment in Japan to boost its cloud computing and AI services, aiming to meet increasing demand and accelerate digital transformation in the region.

Sliced:

  • 🆕 Meta launches Llama 3

  • 👩🏻 Microsoft shows off VASA-1, an AI framework that makes human headshots talk, sing

  • ⏩  Google streamlines structure to speed up AI efforts

  • 😕 Stability AI Lays Off 10% Of Staff

  • 💵 Oracle to invest over $8 billion in Japan in cloud computing, AI

Meta Llama 3 introduces a groundbreaking leap in AI capabilities, designed for broad utility and powered by advanced features that promote both performance and responsible usage. The latest model comes in 8B and 70B parameter versions, optimized for diverse applications such as coding, reasoning, and instruction-following, showing marked improvements over previous iterations. Meta Llama 3 will soon be accessible across major cloud platforms like AWS, Google Cloud, and Microsoft Azure, and is backed by support from key hardware providers to ensure a wide reach and high efficiency. A key focus has been placed on ethical AI development, featuring new safety tools like Llama Guard 2 and Code Shield to mitigate potential risks associated with AI deployment. With its open-source nature, Llama 3 aims to foster innovation within the AI community, pushing the boundaries of what these models can achieve while maintaining a commitment to safety and accessibility.

Microsoft recently unveiled VASA-1, a groundbreaking AI framework designed to animate static human headshots into realistic talking and singing videos. This innovation requires minimal input, using just a single image and an accompanying audio file to generate content with synchronized lip movements and expressive facial dynamics. Although VASA-1 demonstrated its capability with various examples, including an animated Mona Lisa rapping, Microsoft highlighted that this technology is still a research project and not ready for commercial release due to potential misuses like deepfake creation. The technology, which was developed by Microsoft's research division, also allows users to control the animation's features, such as motion sequence and emotional expression, enhancing its potential for creating lifelike virtual characters.

Google has restructured its AI division, consolidating all teams related to AI model development under the leadership of DeepMind founder, Demis Hassabis. This restructuring aims to accelerate the development and deployment of AI technologies, including its flagship Gemini chatbot. This move is part of Google's strategy to counteract criticisms for its slower pace in commercializing generative AI, compared to competitors like Microsoft, who has significantly benefited from its collaboration with OpenAI. Additionally, Google has made organizational changes within its platform and devices segment, signaling a broader intention to enhance product innovation and efficiency across its services. This restructuring comes as Google prepares for its upcoming first-quarter earnings report and its annual I/O developer conference, amidst internal challenges including employee protests against a controversial AI and cloud contract.

Stability AI, an AI startup known for its involvement in the visual art space with projects like Stable Diffusion, has announced layoffs impacting 10% of its workforce, roughly equating to 20 employees. This decision comes amidst a challenging period marked by previous ambitious fundraising efforts that once positioned the company at a valuation around $1 billion. The layoffs follow controversies surrounding exaggerated claims by founder Emad Mostaque about his background and the company's contributions to AI. Additionally, internal turmoil led to Mostaque's departure following an investor revolt, with the company citing a need to "right-size" after a period of unsustainable growth. This restructuring is indicative of broader trends affecting the AI industry, where other startups have also faced the need to streamline operations and focus on sustainability.

💵 Oracle to invest over $8 billion in Japan in cloud computing, AI

Oracle has announced a significant investment of over $8 billion over the next ten years in Japan to expand its cloud computing and AI infrastructure. This strategic move is aimed at strengthening Oracle Cloud Infrastructure's presence across the country, meeting the growing demand for cloud services and AI capabilities. The plan includes bolstering operations and supporting engineering teams with Japan-based personnel, ensuring that Oracle continues to enhance its service offerings in the region. This investment reflects Oracle's commitment to enhancing its technological footprint and supporting the digital transformation of businesses in Japan.

🛠️ AI tools updates

Meta is innovatively integrating real-time AI image generation into WhatsApp, enhancing user experience by allowing images to evolve dynamically as users type their descriptions. This feature, currently in beta for U.S. users, exemplifies Meta's push to embed more AI functionalities across its platforms. The underlying technology, powered by the Meta Llama 3 model, not only generates sharper and higher-quality images but also introduces the capability to animate these images into GIFs. This development is part of Meta's broader strategy to deploy advanced AI features seamlessly across all its applications, including Instagram, Facebook, and Messenger, ensuring that users have a cutting-edge, interactive, and highly personalized media experience.

💵 Venture Capital updates

Mindverse, a Singapore-based AI startup founded in 2022 by former Meta executives and AI researchers, has successfully raised US$5 million in seed funding led by Square Peg. The funding will support further research and product development, particularly in advancing Mindverse's large personalization model. The startup's flagship products include MindOS Studio, which enables the creation of AI-native websites with personalized chat experiences, and Mebot, a productivity tool designed to learn and remember users' thoughts and preferences. This innovative approach aims to enhance user interaction by organizing and regenerating memories and learning individual thinking patterns for more personalized services.

🫡 Meme of the day

⭐️ Generative AI image of the day