Robots learn by watching online videos • AI designs solar cells • AI creates Minecraft map • GPT-4o mini released • AI news this week
Welcome to the AI Search newsletter. Here are the top highlights in AI this week.
Researchers have developed a new framework that enables robots to learn complex manipulation skills by watching online demonstration videos. The framework, called TieBot, can track object motion in the video, learn grasping points and placing points in a simulated environment, and deploy the policy to a real robot. The approach allows robots to perform tasks such as knotting a tie, and has the potential to simplify and enhance the training of robotics algorithms via human demonstrations. Read more
DeepFaceLive, is a real-time face-swapping tool, allowing users to create convincing and hilarious videos. Not only can you swap faces, but you can also animate any photo of a face in realtime with your own expressions. Check out the full tutorial here.
A neural network has been designed to build spatial maps, allowing it to create its own map of a Minecraft environment. The network was trained on videos of a player traversing the Minecraft world and was able to learn how objects within the world are organized relative to one another, storing representations of the objects spatially with respect to each other. This ability to spatially store and organize information could ultimately help neural networks get "smarter," enabling them to solve truly complex problems like humans can. Read more
Engineers have developed OptoGPT, an AI-powered tool that can design optical multilayer film structures for various applications, including solar cells, smart windows, and telescopes, in a matter of seconds. OptoGPT uses a transformer architecture to predict the optimal material structure for a given optical property, and has been shown to outperform previous models in terms of design speed and accuracy. Read more
Turbotype is a Chrome extension that allows you to type faster by setting customizable shortcuts, designed to boost productivity and save time. Easily create, save, and use keyboard shortcuts for frequently used phrases. It’s free forever - try it out today!
OpenAI has unveiled GPT-4o mini, a smaller and cheaper AI model that outperforms industry-leading small AI models on reasoning tasks involving text and vision. The model is designed for developers and consumers, and is being released through the ChatGPT web and mobile app, with enterprise users gaining access next week. GPT-4o mini is significantly more affordable to run than its previous frontier models, and more than 60% cheaper than GPT-3.5 Turbo. Read more
AI meets cartography, enabling mapping tools to create satellite images from text prompts. This technology allows users to create maps from free-form textual descriptions, and even synthesize satellite images based on a given textual prompt or geographic location, with potential applications in urban modeling, navigation systems, and natural hazard forecasting. Read more
ChatLLM by Abacus AI is an integrated platform that allows enterprises to use multiple LLMs, deploy custom agents, and collaborate with team members. Choose from state-of-the-art LLMs such as GPT-4o, Claude 3 Opus, and their new open source Smaug. Try it for free today!
Researchers have developed a new technique that can guarantee the stability of robots controlled by neural networks. This approach uses a novel verification formulation that enables the use of a scalable neural network verifier to provide rigorous worst-case scenario guarantees, allowing for safer deployment of robots and autonomous vehicles. Read more
Researchers have developed a new system called Bunny-VisionPro, which enables intuitive teleoperation of a robotic manipulator in real-time. The system allows human operators to control dual robot arms and multi-fingered hands in real-time, receiving visual and haptic feedback, making the experience more immersive and improving the system's teleoperation success rates. Read more
AuraFlow v0.1, an open-source text-to-image generation model, has been released, capable of generating high-quality images. This model is exceptionally good at prompt following. AuraFlow is a collaboration between researchers and developers, and its release is a significant step forward in the development of open-source AI models. Read more