ChatGPT Grew Arms and Legs

Have you ever wished a robot could handle some of your daily chores? Well, you might be in luck. Figure AI, an American company founded in 2022 by Brett Adcock, is turning this vision into reality with its humanoid robots powered by advanced artificial intelligence. The company's first model, Figure 01, debuted in 2022, followed by Figure 02, which took its first steps in August 2024. In February 2025, Figure AI introduced Helix, an advanced AI system designed to further enhance its robots' capabilities.

These robots are vision-language-action (VLA) models, meaning that they combine visual perception, language understanding, and physical movement to execute tasks and interact with their surrounding environment. Their smoothly moving upper bodies enable them to grasp small household objects with precision, allowing them to work efficiently both independently and alongside other Figure robots.

Up until February 4th of 2025, Figure AI was working with OpenAI (ChatGPT’s parent company) to adapt OpenAI’s technology to suit Figure’s humanoid robots. However, CEO Brett Adcock recently announced that they had developed their own AI, Helix, to power their robots, therefore ending the partnership with OpenAI. Figure AI’s goal is to “give artificial intelligence a body” by creating humanoid robots capable of doing human tasks to “support us on a global scale”. They intend for their robots to fill crucial workforce gaps, stating that there are 10 million unfilled jobs in the United States, 7 million of which are key roles in warehouses, transportation and retail. With known investors such as Nvidia, Microsoft, and Jeff Bezos, the company is quickly becoming a major player in the tech world.

How do they work?

Unlike past robots, Figures do not require programming for hundreds of tasks. Instead, they learn through experience, interacting with their environment and storing new information in their AI system, therefore contributing to an ever growing knowledge base. They use sensors, natural language processing and machine learning to collect incredible amounts of data that is measured in terabytes (TB). To put this into perspective, an iPhone typically has 128GB to 1TB of storage, and 1TB equals 1,000GB. Figure’s robots generate significantly more data in a day than an iPhone will store in its lifetime.

Here’s why their AI is so effective: Figure robots start with an AI model that contains pre-existing data to guide their actions. As they interact with their environment, they collect additional information that enhances their AI system. This creates a continuous learning cycle, allowing the robots to become more capable and improve over time. Additionally, Figure robots share a central AI system, meaning that when one robot learns something new, that knowledge is given to all the robots. This contributes to efficient learning across the entire fleet.

Figure AI posted a YouTube video on February 20th 2025 showcasing two Figure 02 robots unloading a bag of groceries into a kitchen – despite never having seen any of the items before. The robots collaborated seamlessly, handing objects to one another and placing things where they belonged. They successfully placed an apple in a fruit bowl, and stored multiple groceries in a refrigerator.

Watch the video here: https://youtu.be/yEOjGKA6Kvo

What this means for the future

These advancements in AI-powered robotics are a first step to reshaping the workforce and everyday life. With the ability to perform human-like tasks autonomously, robots like Figure 02 could replace manual labor roles in warehouses, logistics, and even homes. This innovation could signify the beginning of a future where robots assist in daily life, reducing the burden of routine tasks.

One day we may look back on the video of the robots unloading groceries and laugh at the simplicity of our “primitive” technologies, but as it stands, this remains an astonishing milestone in AI powered robotics, paving the way for a future where humanoid robots seamlessly integrate into society.

Sources: