Claude Browses, Nvidia's GR00T, LLMs Improve Themselves & OpenAI's New Pricing
March 24th 2025 | Superintelligence Newsletter
Hey Superintelligence Fam 👋
From Claude’s real-time web search to Nvidia’s open-source leap in humanoid robotics and a peer-reviewed AI scientist, this week showcased how fast agentic and generative AI are evolving. While OpenAI’s “o1 Pro” pricing—$150/million tokens input and $600/million tokens output—sparked debates, breakthroughs in LLMs, visual generation, and ethical AI took center stage.
Innovations like Stable Virtual Camera, Gemini Co-Drawing, and Adobe’s AI Orchestrator hint at a creative and intelligent future, while platforms like OpenAI.fm enable AI music creation. Meanwhile, global AI ethics discussions intensified with scrutiny on bias, regulation delays, and copyright concerns.
Reading Time: 3 Minutes
Anthropic's Claude Gains Web Search Capabilities : Anthropic's AI assistant, Claude, now features web search, allowing access to up-to-date information and providing citations for verification.
NVIDIA Introduces Isaac GR00T N1 for Humanoid Robotics : NVIDIA unveiled Isaac GR00T N1, an open-source foundation model aimed at enhancing humanoid robot development, offering generalized reasoning and skills.
AI Scientist Achieves Peer-Reviewed Publication : Sakana AI's "The AI Scientist" autonomously generated a paper that passed peer review at a leading AI conference, marking a milestone in automated research.
LADDER: Self-Improving LLMs Through Recursive Problem Decomposition : LLMs boost integration accuracy from 1% to 82% (and 90% with TTRL) by recursively decomposing complex problems into easier variants.
Agentic Reward Modeling: Integrating Human Preferences with Verifiable Correctness Signals for Reliable Reward Systems : A hybrid reward system combines human feedback with factual and constraint checks, outperforming baselines on RM-Bench and JudgeBench for reliable LLM training.
Fractal Generative Models. : Fractal modules recursively generate images, achieving state-of-the-art 3.14 bits/dim on ImageNet 64×64 and FID 6.15 on 256×256 class-conditional synthesis.
Adobe Experience Platform Agent Orchestrator enhances customer experiences by orchestrating AI agents, boosting productivity, and delivering personalized interactions at scale.
OpenAI.fm is a platform dedicated to AI music generation, offering tools and resources for creating AI-composed music tracks.
Stable Virtual Camera transforms 2D images into immersive 3D videos, providing dynamic perspectives and realistic depth without complex reconstruction.
Gemini Co-Drawing is an interactive AI drawing tool that allows users to collaboratively create and modify images using Gemini 2.0's native image generation capabilities.
Unlock the Power of Large Language Models
Delve into the world of AI with a free 7-day email course covering LLM fundamentals, architectures, training processes, applications, and ethical considerations.
Enrol Now to enhance your AI expertise.
The global discourse on AI ethics intensified. OpenAI’s video generator Sora came under scrutiny for reinforcing racial and gender biases in generated content. In the UK, delays in passing the AI safety bill drew criticism, with warnings about regulatory gaps. Meanwhile, Alden Global Capital’s newspapers opposed OpenAI and Google’s push to weaken copyright protections for creators. At GTC 2025, Nvidia unveiled ethical AI innovations, including the Blackwell Ultra chip and Groot N1 for robotics. In assistive tech, the University of Glasgow introduced RoboGuide, an AI-powered guide dog for the visually impaired.
Thank you for tuning in to this week's edition of Superintelligence Newsletter! Stay connected for more groundbreaking insights and updates on the latest in AI and superintelligence.
For more in-depth articles and expert perspectives, visit our website | Have feedback? Provide feedback.
If you wish to partner with us then Explore Here
Stay curious, stay informed, and keep pushing the boundaries of what's possible!
Until Next Time!
Superintelligence Team.