Fine-tuned LAnguage Net (FLAN) is a 137B parameter, instruction-tuned NLP model from Google AI that performs well in zero-shot learning
Translatotron 2 is a robust direct speech-to-speech translation (S2ST) model from Google AI which can directly translate speech between 2 languages & retain speaker's voice in the translated speech.
Facebook AI's Generative Spoken Language Model ( GSLM ) is the first textless NLP model that generates expressive speech just from raw audio
Google AI's SoundStream is a neural network audio codec for speech and music, producing higher quality audio
Codex is OpenAI's flagship product to generate computer code using English instructions. It is a descendant of GPT-3 model, fine-tuned on Github's public repositories
GAN consists of two neural networks Generator & Discriminator, trained alternatively to create new images
Triton is a language to write kernels for GPU to carry out effective parallel processing & optimization of tasks
Facebook AI has blended long-term memory & contextual search to build Blender Bot 2.0, a state of the art opensource chatbot