TL;DR: Figure’s new OpenAI-powered demo shows their humanoid holding full conversations while making coffee, and honestly, it’s wild.
Remember when ChatGPT was just a text box? Yeah, those days are over. Figure just dropped a demo of their Figure 01 robot powered directly by OpenAI’s tech, and it’s equal parts impressive and eerie.
The robot doesn’t just take commands. It listens, processes the scene visually, and talks back in real time. In the demo, a human asks for something to eat. The bot looks at a table holding an apple, a plate, and a cup, identifies the apple as the only edible item, and hands it over while explaining its reasoning out loud.
This is what embodied AI looks like. They’ve linked a massive vision-language model straight to the robot’s physical actions. It isn’t running pre-programmed scripts. It’s actually figuring out its environment on the fly. The race to build a truly general-purpose worker just hit another gear.