China's humanoid robots aren't smart enough to take your job -- yet A human trainer teaches a robot how to package noodles while holding a controller and wearing a headset fitted with motion sensors at a robot data training center in Liuzhou, Guangxi, China. LIUZHOU/BEIJING - Inside a training center in southern China, more than 100 humanoid robots stand in neat rows. The humanoids are slow and awkward. A novice trainer might get them to produce one usable movement in 300 attempts; an experienced trainer, one in 50. Robot maker UBTech won an $18 million tender from the Guangxi regional government last October to supply the training facility with humanoids and related hardware. The mission: to produce training data for so-called embodied AI -- robots able to perceive, decide and act in the physical world.

Human explanation (e.g.,interms offeature importance) hasbeenrecently used to extend the communication channel between human and agent in interactive machinelearning.

Adversarial Training (AT) has become arguably the state-of-the-art algorithm for extracting robust features. However, researchers recently notice that AT suffers from severe robust overfitting problems, particularly after learning rate (LR) decay. In this paper, we explain this phenomenon by viewing adversarial training as a dynamic minimax game between the model trainer and the attacker. Specifically, we analyze how LR decay breaks the balance between the minimax game by empowering the trainer with a stronger memorization ability, and show such imbalance induces robust overfitting as a result of memorizing non-robust features. We validate this understanding with extensive experiments, and provide a holistic view of robust overfitting from the dynamics of both the two game players. This understanding further inspires us to alleviate robust overfitting by rebalancing the two players by either regularizing the trainer's capacity or improving the attack strength. Experiments show that the proposed ReBalanced Adversarial Training (ReBAT) can attain good robustness and does not suffer from robust overfitting even after very long training. Code is available at https://github.com/PKU-ML/ReBAT.