- NATURAL 20
- Posts
- Adobe Unveils Firefly Video Model
Adobe Unveils Firefly Video Model
PLUS: Dario Amodei Advocates for AI Good, Tech Firms to Invest in Data Centers and more.
We scour 100+ sources daily
Read by CEOs, scientists, business owners and more
3.5 million subscribers
Today:
Adobe Unveils Firefly Video Model
Google, Kairos Partner for Nuclear Power
Microsoft AI VP Moves to OpenAI
Dario Amodei Advocates for AI Good
Adobe has launched a new Firefly Video Model (beta) at its Adobe MAX conference. This generative AI model, designed for video creation, is safe for commercial use and expands Adobe’s AI offerings, which already include image, vector, and design models.
The Firefly models, integrated into popular tools like Photoshop, Illustrator, and Premiere Pro, now offer faster image generation and new video capabilities like Text to Video and Image to Video. Firefly has been used to generate over 13 billion images since March 2023, and is rapidly being adopted by major brands for creative content.
Kairos Power and Google have partnered to deploy 500 MW of advanced nuclear energy by 2035. This agreement marks the first corporate deal for multiple reactor projects in the U.S. Kairos will build and operate nuclear plants, selling electricity to Google data centers under Power Purchase Agreements. The first plant is expected by 2030 to help Google meet its carbon-free energy goals.
This partnership supports Kairos Power's efforts to commercialize advanced nuclear technology and accelerate clean energy adoption. Google's commitment reinforces their leadership in decarbonization and advancing innovative energy solutions.
Sebastien Bubeck, a prominent AI researcher and Vice President at Microsoft, is leaving to join OpenAI. Bubeck played a crucial role in Microsoft's development of small language models. A Microsoft spokesperson confirmed his departure, stating that Bubeck will continue working on artificial general intelligence (AGI) at OpenAI.
The spokesperson also emphasized that Microsoft looks forward to maintaining a collaborative relationship through Bubeck’s work at OpenAI.
Dario Amodei, CEO of Anthropic, argues that while AI presents significant risks, it also holds radical potential for improving human life. He outlines five key areas where AI could have transformative positive effects: biology and health, neuroscience and mental well-being, economic development and poverty reduction, peace and governance, and the future of work and meaning.
Amodei emphasizes that achieving this positive future depends on careful management of AI risks, cooperation, and a shared commitment to harnessing AI for societal good.
🧠RESEARCH
Meissonic is a new model for text-to-image generation that enhances non-autoregressive masked image modeling. By incorporating architectural innovations, better encoding, and optimized sampling, Meissonic generates high-quality, high-resolution images more efficiently than previous models like SDXL. Extensive tests show it matches or surpasses existing text-to-image systems.
The paper presents a multi-agent collaborative data selection framework to improve the efficiency of pretraining large language models (LLMs). By treating each data selection method as an agent and integrating their information dynamically, the approach accelerates LLM training and boosts performance by 10.5% over current methods, according to extensive experiments.
StructRAG, a framework that improves large language models' performance on knowledge-intensive reasoning tasks. Unlike existing retrieval-augmented generation (RAG) methods, StructRAG structures scattered information into optimal formats for better reasoning. Experiments show StructRAG achieves state-of-the-art results, excelling in complex tasks and offering a significant advancement in handling knowledge-heavy scenarios.
PositionID is a method to improve large language models' (LLMs) control over text length and accuracy in copy-paste tasks. By introducing PositionID Prompting and Fine-Tuning, the model gains better positional awareness, enhancing length adherence and copying accuracy. Experiments show significant improvements without affecting the quality of responses.
AlphaLLM-CPL is a novel framework that enhances large language models' reasoning skills through Monte Carlo Tree Search (MCTS) behavior distillation. By leveraging stepwise trajectory pairs and curriculum preference learning, the method efficiently distills MCTS trajectories. Experimental results show that AlphaLLM-CPL significantly improves performance in mathematical reasoning tasks, surpassing previous distillation methods.
🛠️TOP TOOLS
LlamaCoder - Turn your idea into an app
Hubspot AI - AI for videos, blogs, and support, all in one place
Langtail - Debug prompts, run tests, and observe what’s happening in production
Tabnine - The AI coding assistant that you control
WPS Office - AI-Powered Office Suite for Everyone
📲SOCIAL MEDIA
Big news, photographers! We’re thrilled to unveil powerful new AI features and performance boosts for our Lightroom MAX 2024 release! 🚀
Dive into our latest release to see how we’ve tailored Lightroom just for you 📸 ✨ #MadeWithLightroom#AdobeMAX
— Adobe Lightroom (@Lightroom)
1:07 PM • Oct 14, 2024
🗞️MORE NEWS
Four U.S. tech companies, including CyrusOne and ServiceNow, pledged over $8 billion to expand data centers in the U.K., driven by increased demand for artificial intelligence. The British government sees these investments as critical infrastructure.
PlayVS partnered with Omnic.AI to provide AI-driven feedback for gamers, improving in-game performance and communication skills. The initiative targets middle and high school students, offering insights to boost gameplay, leadership, and academic success.
Nvidia CEO Jensen Huang praised Elon Musk for building the xAI supercomputer, Colossus, in just 19 days. Huang called the feat "superhuman," noting Musk's exceptional engineering skills and the rapid deployment of 100,000 Nvidia GPUs for the project.
Beyond Presence, a Munich-based startup, raised $3.1 million in pre-seed funding to develop hyper-realistic avatars for real-time conversations. Led by HV Capital, the funds will advance the technology and release a public beta later this month.
What'd you think of today's edition? |
Learn AI with us. Let’s Build the Future Together. |
Hello fellow AI-obsessed traveler, Over the past 2 years, as we’ve grown to over 250,000 subscribers between the YouTube Channel and this newsletter, we've received an overwhelming number of requests for one specific thing. While the newsletter helps keep you up to speed with AI news, many of you have asked for the next step: to learn how to actually apply AI in your work. Today we’re finally announcing the solution with NATURAL 20, the community for like-minded AI learners. As a loyal newsletter reader you are getting access at the lowest price it will ever be: JOIN NATURAL 20 AI UNIVERSITY TODAY What you get: * Tutorials by experts across various AI fields. * Daily tutorials by Wes Roth about the latest use cases. * Building Autonomous AI Agents to Automate Your Life and Business (NEW!) * A network of the top 1% of early AI adopters. * Access to community-only resources and software. * And many more features rolling out soon. |
Reply