@DrDabbles, do you understand the difference between control (i.e. low-level actuator commands) and planning (i.e. making decisions and determining a trajectory) in an autonomous car context?
Do you understand that HW2+ Autopilot currently uses a neural network trained via imitation learning for at least one planning task, i.e., path prediction (but not for control)?
Do you understand that the deep learning technique Tesla used to train path prediction — imitation learning — is the same technique DeepMind used to train AlphaStar Supervised?
Do you understand that HW2+ Autopilot currently uses a neural network trained via imitation learning for at least one planning task, i.e., path prediction (but not for control)?
Do you understand that the deep learning technique Tesla used to train path prediction — imitation learning — is the same technique DeepMind used to train AlphaStar Supervised?
Last edited: