Figure unveils Index, a new platform for robot training data, emerging from stealth

Figure has officially launched Index, a new system designed to gather real-world physical data for training general-purpose robots, marking the company's emergence from stealth operations. The initiative aims to create the largest and most varied robot training dataset globally, addressing the scarcity of suitable data for scaling advanced robotics.

Index was initially developed over four months as a stealth app, allowing Figure to establish a dedicated pipeline for high-throughput, diverse, and quality-controlled data collection. The platform is now available on Google Play and the App Store, enabling individuals to contribute physical task videos from their homes or workplaces.

Since its stealth launch, the Index app has achieved over 264,000 downloads across more than 100 countries, engaging over 44,000 weekly active users. Contributors have uploaded more than 16 million videos to the platform, with the app processing 30 minutes of video uploads every second, equivalent to 4.9 years of human work daily.

Figure has compensated Creators with $15 million to date for their data contributions. The company has also committed to investing over $1 billion in data collection and compute resources over the next 12 months, aiming for a 100x increase in its data capabilities. The dataset itself boasts significant diversity, with every 1,000 hours collected containing 373 unique tasks, 1,146 distinct manipulated objects, and 116 different environments.

Index is an initiative by Figure, a company focused on developing general-purpose robots. The data collected through Index is specifically designed to enhance Figure's AI stack, known as Helix, which relies on extensive data to improve its capabilities and achieve greater generalisation in robotic tasks.

The launch of Index is significant because it addresses a critical challenge in robotics: the lack of diverse, real-world physical data needed to train truly general-purpose robots. Figure's approach, by crowdsourcing data from human 'Creators,' allows for the capture of a wide array of tasks, objects, and environments that are difficult to replicate or define through traditional data acquisition methods.

This strategy leverages the inherent diversity of human behaviour and environments, providing the long-tail variations essential for robust AI learning in robotics. The company's internal results already indicate that this data-centric approach is validating its thesis on generalisation.

Figure's long-term vision for Index extends beyond data collection, laying the groundwork for a future where robots can be ordered as a service. The company anticipates a shift from human assistance for daily chores to fully autonomous robotic solutions, suggesting a transformative impact on how tasks are performed in homes and businesses.

The ongoing investment in data and compute, alongside the continuous expansion of the Creator network, will be crucial in determining the pace at which Figure can realise its ambition of deploying highly capable, general-purpose robots.