RL agents 🤖 need a lot of data, which they usually need to gather themselves. But does that data need to be real? Enter *Synthetic Experience Replay*, leveraging recent advances in #GenerativeAI in order to vastly upsample ⬆️ an agent’s training data!

Paper: arxiv.org/abs/2303.06614