Google unveils Genie 2 AI model for creating interactive 3D worlds

Google has introduced its Genie 2 AI model, which can create interactive 3D worlds using images or text descriptions. This model builds on the original Genie, which focused on 2D games, and aims to enhance AI training in realistic environments. Genie 2 features a "long horizon memory," allowing it to remember and accurately render parts of a virtual world as they come back into view. This addresses a significant challenge faced by previous models, which struggled with maintaining coherence over time. Despite these advancements, details about Genie 2's training data and practical applications remain limited. The model's effectiveness beyond short demonstrations is still uncertain.