As AI algorithms increasingly participate in daily activities that used to be the sole province of humans, we are inevitably called upon to consider how much machines are really like us. To address this question, we turn to the Turing test and systematically benchmark current AIs in their abilities to imitate humans. We establish a methodology to evaluate humans versus machines in Turing-like tests and systematically evaluate a representative set of selected domains, parameters, and variables. The experiments involved testing 769 human agents, 24 state-of-the-art AI agents, 896 human judges, and 8 AI judges, in 21,570 Turing tests across 6 tasks encompassing vision and language modalities. Surprisingly, the results reveal that current AIs are not far from being able to impersonate human judges across different ages, genders, and educational levels in complex visual and language challenges. In contrast, simple AI judges outperform human judges in distinguishing human answers versus machine answers. The curated large-scale Turing test datasets introduced here and their evaluation metrics provide valuable insights to assess whether an agent is human or not. The proposed formulation to benchmark human imitation ability in current AIs paves a way for the research community to expand Turing tests to other research areas and conditions. All of source code and data are publicly available at https://tinyurl.com/8x8nha7p
Graph representation learning is an important task with applications in various areas such as online social networks, e-commerce networks, WWW, and semantic webs. For unsupervised graph representation learning, many algorithms such as Node2Vec and Graph-SAGE make use of "negative sampling" and/or noise contrastive estimation loss. This bears similar ideas to contrastive learning, which "contrasts" the node representation similarities of semantically similar (positive) pairs against those of negative pairs. However, despite the success of contrastive learning, we found that directly applying this technique to graph representation learning models (e.g., graph convolutional networks) does not always work. We theoretically analyze the generalization performance and propose a light-weight regularization term that avoids the high scales of node representations' norms and the high variance among them to improve the generalization performance. Our experimental results further validate that this regularization term significantly improves the representation quality across different node similarity definitions and outperforms the state-of-the-art methods.