Amazon launches new Nova foundation models
Plus OpenAI releases upgraded o1 reasoning model
Today’s Highlights:
📰 News: Amazon launches new Nova foundation models + OpenAI releases upgraded o1 reasoning model
💰 Funding: xAI raises $6Bn
⚡️ Top News Stories:
1. Amazon has launched its Nova foundation models, integrated into Amazon Bedrock, offering a diverse suite including cost-efficient options like Nova Micro and Lite, high-performance Nova Pro, and creative tools like Nova Canvas for image generation and Nova Reel for video generation, with fine-tuning, Retrieval Augmented Generation (RAG), and distillation capabilities to enable tailored AI applications for industries like advertising, sports, and entertainment.
2. Anduril Industries and OpenAI have partnered to develop advanced AI solutions for U.S. and allied national security missions, focusing on counter-unmanned aircraft systems (CUAS) by integrating OpenAI’s models with Anduril’s Lattice platform to enhance real-time threat detection, response, and situational awareness through data-driven training on Anduril’s CUAS data.
3. OpenAI has fully released its upgraded o1 reasoning model, featuring enhanced coding, math, and image reasoning capabilities, and introduced a $200/month ChatGPT Pro tier offering exclusive access to o1 Pro mode, GPT-4o, and Advanced Voice mode for users with complex needs.
4. OpenAI has partnered with Future, integrating content from over 200 brands like Marie Claire and PC Gamer into ChatGPT with attribution and links, while Future uses OpenAI’s tools to create interactive chatbots for brands such as Tom’s Hardware and Who What Wear to enhance user engagement.
5. Perplexity has expanded its revenue-sharing Publisher Program to include outlets like LA Times, Adweek, and The Independent, despite facing legal challenges from publishers like The New York Times and Dow Jones over accusations of inaccurately summarizing paywalled and copyrighted content.
6. World Labs, founded by Fei-Fei Li, has unveiled an AI system that generates interactive and modifiable 3D scenes from a single image, offering realistic features like a controllable camera and adjustable depth of field, with applications for movie studios, game developers, and engineers to reduce costs and accelerate 3D world creation.
7. DeepMind’s Genie 2 generates interactive 3D worlds from text descriptions or single images, simulating lifelike physics, lighting, and NPC behavior with features like varied perspectives, memory for off-screen elements, and user interactions, designed for research and prototyping short simulations rather than full-fledged gaming.
8. Hume AI has launched "Voice Control," a no-code tool that lets developers customize AI voices along 10 dimensions like gender, confidence, and enthusiasm with real-time, reproducible adjustments, enabling precise, unique voice creation tailored to brands and applications without relying on voice cloning.
9. ElevenLabs’ Conversational AI platform enables developers to create voice agents in minutes with low latency, scalable solutions, and tools like Speech-to-Text, Text-to-Speech, LLM integration, and advanced turn-taking, offering flexibility to use default models, integrate the latest LLMs, or host custom servers for diverse and dynamic conversational applications.
10. Luma AI’s Ray 2 Video Model generates high-quality cinematic videos from text and image prompts, now supporting clips up to 1min, with lifelike characters, smooth motion, and advanced cinematography, and will be integrated into Amazon Bedrock through a partnership enabling developers to incorporate advanced video generation into AI applications.
11. Amazon, in collaboration with Anthropic, is building "Project Rainer," the world’s largest AI supercomputer powered by Trainium 2 chips, offering 30–40% cost savings over Nvidia GPU clusters and promising fourfold performance improvements with Trainium 3 by 2025.
12. Tencent's open-source AI video model, Hunyuan, with 13Bn parameters, generates high-resolution, 5-second videos from text prompts, showcasing strong motion quality and style diversity, though initial tests reveal mixed prompt adherence and contextual accuracy compared to competitors like Runway Gen-3 and Mochi-1.
13. Google Cloud’s Vertex AI now offers Veo, a state-of-the-art video generation model, and Imagen 3, a photorealistic image generator with advanced editing features, enabling businesses like Mondelez, WPP, and Agoda to create campaign-ready visuals and videos quickly, reducing production time and cost.
14. DeepMind’s GenCast, an advanced AI ensemble weather forecasting model, delivers high-resolution (0.25°) predictions in just 8 min, outperforming traditional systems with 50+ probabilistic forecasts for extreme weather, benefiting sectors like renewable energy and disaster response.
15. Cohere’s Rerank 3.5, a multilingual AI search model now supporting 100+ languages, uses advanced cross-encoding to improve query precision by over 23% compared to traditional systems, offering significant benefits for industries like finance, healthcare, and manufacturing where data accuracy is critical.
16. Google's PaliGemma 2, an AI model that generates image captions including objects, actions, and emotions, has sparked ethical and scientific concerns over the reliability and cultural biases of emotion detection from facial features.
17. Microsoft has launched Copilot Vision, an AI tool available in the $20/month Copilot Pro plan, that reads and responds to questions about pre-approved U.S. websites, summarizing, translating, and highlighting details like product discounts and game tips.
18. At NeurIPS 2024, Google DeepMind will present over 100 papers, including Test of Time award winners, featuring breakthroughs in AI agent adaptability with tools like AndroidControl and in-context abstraction learning, alongside innovations in 3D scene creation using CAT3D, Neural Assets, and SDF-Sim for scalable simulations and enhanced object manipulation.
💰 Top Funding News:
1. Elon Musk’s xAI, which develops "Grok," raised $6Bn in new funding.
2. Cleerly, which uses AI-powered software to analyze CT scans for early detection of coronary artery disease, raised a $106M Series C extension, led by Insight Partners, w/ Battery Ventures.
3. Tractian, which leverages AI to prevent unplanned industrial downtime by integrating hardware and software solutions for asset monitoring, operations, and maintenance management, raised a $120M Series C, led by Sapphire Ventures, w/ General Catalyst, Next47, and NGP Capital.
4. AMP, which leverages AI-driven deep learning to revolutionize waste sortation and recycling, raised a $91M Series D, led by Congruent Ventures, w/ Sequoia Capital, XN, Blue Earth Capital, Liberty Mutual Investments, CalSTRS, Wellington Management, Range Ventures, and Tao Capital Partners.
5. Lawhive, an AI-based SaaS platform that helps small "Main Street" law firms automate processes, reduce costs by up to 50%, and improve efficiency, raised a $40M in a Series A co-led by GV and TQ Ventures, w/ Balderton Capital, Jigsaw, Episode 1.
6. Yurts, which develops AI integration platforms and secure AI-powered chat assistants for high-security environments, such as the U.S. Department of Defense (DoD), raised $40M Series B, led by XYZ Venture Capital.
7. Axiado, which develops AI-driven, hardware-anchored security solutions for AI data centers and accelerated computing platforms, raised $60M in a Series C, led by Maverick Silicon, w/ Samsung Catalyst Fund, Atreides Management, and Crosslink Capital.