Data Collection

With a fresh $15 million in funding and a massive new dataset, Build AI is doubling down on the bet that human video is the key to universal robot intelligence.

The $5.6 billion startup reveals that sufficiently large robot models naturally learn to understand human video, potentially solving the industry's data bottleneck.

In a technical addendum to its GEN-0 launch, Generalist AI reveals new details on its pretraining methodology, introducing metrics like "Reverse KL" to measure model creativity and claiming infrastructure capable of absorbing 6.85 years of robot experience per day.

Researchers from the National University of Singapore have developed a generative video pipeline that converts third-person human activity footage into synthetic training data for humanoid robots, effectively bridging the embodiment gap at scale.

The London-based startup wants to be the infrastructure layer for robot learning, offering free cloud-native data tools to researchers to solve the "plumbing" problem.