In this episode of Unsupervised Learning, Jacob interviews Nicole Brichtova and Oliver Wang from Google about Nano Banana, Google's new image model. They discuss the model's character consistency and image quality, favorite use cases such as turning photos into figurines and colorizing old photos, and feature requests like higher resolution and transparency. Oliver explains how the model can answer harder questions and provide helpful suggestions, like redecorating a room. They also touch on the role of LLMs in improving image models, the challenges of building products for different user types, and future interface designs. The conversation further explores the evolution of image models, the impact of Midjourney, and the relationship between image and video models, as well as the future direction of AI, including the importance of factuality and proactive models.
Sign in to continue reading, translating and more.
Continue