Google Muse AI Explained: How Does It Work? - Dataconomy

Google Muse AI is the latest additon from the tech giant to a swarm of AI tools we have been seeing lately. The new text-to-image transformer model claims to be quicker than competing methods, because it uses parallel decoding and a compact, discrete latent space. According to its developers, Google Muse AI can produce images at state-of-the-art image generation performance. We present Muse, a text-to-image Transformer model that achieves state-of-the-art image generation performance while being significantly more efficient than diffusion or autoregressive models. Google Muse AI is an allegedly improved version of earlier text-to-image transformer models like Imagen and DALL-E 2. Muse is trained on a masked modeling task in discrete token space using the text embedding acquired from a pre-trained large language model (LLM).