1) Amazon Introduces Nova AI Models in AWS Bedrock
Amazon has unveiled six AI foundation models under the “Nova” brand, now part of the Amazon Bedrock library on AWS. The lineup includes Nova Micro, optimized for speed and cost; Nova Lite, a low-cost multimodal model handling images, video, and text; Nova Pro, a high-performance multimodal model; and Nova Premier, designed for complex reasoning, coming in 2025. Additionally, Nova Canvas supports image generation, and Nova Reel specializes in video creation—both with watermarking features for responsible AI use. Amazon also plans speech-to-speech and native multimodal models and is collaborating with Anthropic to build the world’s largest AI compute cluster using Trainium 2 chips.
➡️ Read more

2) Hugging Face CEO Warns on Chinese AI Models
Clément Delangue, CEO of Hugging Face, has raised concerns about Western companies leveraging high-performing open-source AI models developed in China. In a recent podcast, Delangue warned of unintended consequences, emphasizing that while these models offer advanced capabilities, their adoption could inadvertently strengthen China’s technological standing and pose challenges related to transparency and security. He urged Western enterprises to carefully assess the strategic implications of relying on foreign AI technologies, advocating for a balanced approach that considers both innovation and potential geopolitical risks.
➡️ Read more

3) Google Cloud Unveils Veo AI Video Generator
Google Cloud has introduced Veo, an AI model developed by Google DeepMind, now in private preview on Vertex AI. Veo transforms text or image prompts into cinematic, high-definition videos exceeding 60 seconds, maintaining seamless visual coherence. This positions Google Cloud as the first major provider to offer a video generation model to its customers, expanding creative possibilities in sectors like marketing and advertising. Alongside Veo, Google has launched Imagen 3, a text-to-image model producing photorealistic visuals with advanced editing features, available to all Vertex AI users. These tools aim to streamline AI workflows, enabling the creation of high-quality assets for various applications.
➡️ Read more

4) Tech Bonus - Liquid AI’s STAR Surpasses Transformers
Liquid AI, an MIT-born startup, has unveiled STAR (Synthesis of Tailored Architectures), a novel AI model architecture that automates the generation and optimization of AI models using evolutionary algorithms and a numerical encoding system. Unlike traditional methods relying on Transformer architectures, STAR explores a vast design space, enabling iterative optimization tailored to specific tasks and hardware requirements. In tests, STAR-evolved architectures outperformed optimized Transformer++ and hybrid models, achieving up to a 90% reduction in cache size while maintaining high performance. This advancement represents a significant shift in AI model design, emphasizing efficiency and adaptability.
➡️ Read more

MagicSchool AI: Revolutionizing K-12 Education with Generative AI
MagicSchool AI empowers teachers with 70+ AI tools for lesson planning, assessments, and IEPs, tackling burnout and enhancing student learning experiences.

➡️ Read our comment

Over 60% of business owners believe AI will improve customer relationships