Project Rumi: Augmenting AI Understanding Through Multimodal Paralinguistic Prompting
New research from Microsoft aims to help AI understand non-verbal cues like facial expressions, tone of voice, gestures, and more
Research
A collection of our articles covering the latest in AI research. Get comprehensive analyses and insights into the cutting-edge AI models and breakthroughs from leading tech companies and research labs.
Meta Open Sources AudioCraft
The company says responsible innovation can’t happen in isolation. By open sourcing its research and resulting models, it hopes to ensure that everyone has equal access.
Seeing Through Darkness: Purdue Researchers Unveil AI Thermal Imaging Technology
Improving upon traditional machine vision and perception, this method is designed to help machines perceive and interact with their surroundings in ways that were previously unthinkable.
RLHF In the Spotlight: Problems and Limitations with Key AI Alignment Technique
A new paper takes a closer look at serious open challenges related to RLHF, and emphasizes that relying solely on it for AI alignment is profoundly risky
Med-PaLM M is a Multimodal Biomedical AI from Google Research and Google DeepMind
Across all tasks in the benchmark, Med-PaLM M reached or exceeded state-of-the-art performance, often surpassing specialized models optimized for individual tasks by a wide margin.
“Sure, I’ll Help You Build a Bomb”: New Research Reveals Alarming Gaps in AI Safety
The new class of adversarial attacks are capable of circumventing the alignment measures designed to prevent the generation of inappropriate or harmful content in multiple LLMs.
Google DeepMind Unveils RT-2, Bringing Robots Closer to General Intelligence
Essentially, the model can comprehend and 'speak robot', translating abstract concepts learned from vast web data into actionable knowledge that informs robot behavior.
Meta Releases Open Catalyst Demo to Help Discover Catalysts Faster With AI
The research represents a potential game-changer for material science, enabling catalyst simulations at unprecedented speeds and scales.
Highlights from Apple's Natural Language Understanding Workshop 2023
Topics covered included privacy preservation, knowledge grounding, multimodal understanding, and controlled synthetic data generation.
Meta Unveils CM3leon: A Breakthrough AI Model for Advanced Text-to-Image Generation and Image Understanding
With its versatile capabilities and improved performance, CM3leon represents a significant step towards higher-fidelity image generation and understanding, paving the way for enhanced creativity and applications in the metaverse.
Toyota Unveils New Generative AI Technique to Streamline Vehicle Design
The novel technique facilitates the inclusion of preliminary design blueprints as well as engineering limitations when generating design concepts.
Take a listen to Voicebox, Meta's new generative AI for speech
This new model is a significant leap forward in AI speech synthesis, demonstrating its versatility and efficiency by outperforming existing models in various tasks, even those it was not specifically trained for.
ChatGPT is About to Get a Lot Better At Math
New research from OpenAI explores how training reward models with process supervision significantly improves mathematical reasoning compared to outcome supervision.
Google Researchers Publish Barkour, A Benchmark for Quadiped Robot Agility
A new study establishes an agility course and scoring system for evaluating four-legged robot locomotion.
New Research From Microsoft Teases Next-Level Multimodal Generative AI Capabilities
The research team has developed an AI model that can process and generate any combination of text, images, video and audio.
OpenAI Releases Shap·E: A Leap Forward in Generating 3D Implicit Functions from Text
Their latest research in 3D model generation gets a major upgrade with Shap·E, addressing limitations of their previous model, Point·E.
Google DeepMind is Using AI to Teach Robots to Play Soccer
The OP3 Soccer Team at DeepMind has applied deep reinforcement learning to train low-cost humanoid robots to play 1v1 soccer. This research could open doors to more practical applications of reinforcement learning in robotics.
Meta AI Unveils DINOv2: A Game-Changer in Self-Supervised Vision Transformer Models
Meta AI follows up on its Segment Anything release with DINOv2, a self-supervised model that's poised to redefine image processing across industries
Can Generative AI Simulate Human Behavior?
Researchers from Google and Stanford put 25 ChatGPT agents in a virtual world called Smallville to find out.
Meta AI Open Sources its Segment Anything (SA) Project
The project includes a new a new task, dataset, and SAM—a fascinating new model for image segmentation