9:37 in the recordingVision-language-action and foundation models have changed robotics by enabling robots to learn from multiple modalities and shifting programming toward learning from demonstrations rather than classical programming.Fuel for Physical AI
10:17 in the recordingThe best training data is the same kind of data the model will encounter when it is deployed.Fuel for Physical AI
10:46 in the recordingA general-purpose VLA may get a robot 80–90% of the way, but effective factory deployment requires capturing, structuring, and training on the proprietary data inside each factory, including its specific products, workers' skills, and environments.Fuel for Physical AI