How Midjourney Uses Community & RLHF to build product

Midjourney builds their product by understanding the nuance and making unique product decisions based on their company. They create the best model that produces beautiful and on-point images. To achieve this, they have a small engaged community that provides tight feedback loops. This community constantly refreshes, upvotes, and downvotes the image outputs. Additionally, Midjourney runs relatively smaller models every six to eight weeks, using RLHF (Reinforcement Learning from Human Feedback) to quickly iterate and improve their models. By leveraging the subjective nature of beauty, the community feeds data into the model and strives to level up their product regularly.