AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
ImageBind: A Surprisingly Simple Approach for Multimodal Instruction Following
Image Bind is a new open source tool that maps videos and images to the same latent representation, making it possible for models to follow instructions based on audio, infrared, and other modalities. Developed by the University of Cambridge and the Institute of Science and Technology, Image Bind is surprisingly simple but a big deal in AI technology. It's available for download and experimentation.