Figure AI has launched Index, a data collection platform designed to train its Helix robotics models using videos of humans performing physical tasks.
In four months, Index has reached more than 264,000 downloads across 108 countries, with over 44,000 weekly active users contributing more than 16 million videos. Figure says around 30 minutes of new footage is now uploaded every second and has paid contributors $15 million to date.
The company plans to spend more than $1 billion on data and compute over the next 12 months as it looks to scale the platform by another 100x.
Why Figure Is Collecting Human Data
Figure’s bet is that the data needed to train general-purpose robots doesn’t exist online at sufficient scale.
Tasks like folding laundry, stocking shelves, cleaning a kitchen or working inside a factory involve different objects, environments and ways of completing the same job. For every 1,000 hours collected, Figure says Index contains 373 unique tasks, 1,146 manipulated objects and 116 different environments.
That diversity is the point. Rather than relying entirely on data generated by its own robot fleet, Figure can collect examples of physical work from people around the world and use them to train Helix.
Building the Data Layer for Helix
Index adds another piece to Figure’s increasingly integrated approach to robotics. The company builds its own humanoids, manufactures them through BotQ, develops Helix internally and now operates its own large-scale pipeline for collecting training data.
As robot hardware improves, data is becoming one of the industry’s biggest competitive advantages. Figure is betting that collecting more diverse examples of how humans interact with the physical world will help Helix generalize across more tasks and environments.
Index also points toward Figure’s longer-term vision. People can currently use the platform to book a Creator to help with physical tasks. Eventually, Figure wants a robot to take their place.
