Weaknesses: Although I appreciate the benchmark for the "concept learning" aspects that set it apart from most tasks in computer vision, I am skeptical about the type of stimuli used to measure these capabilities. The images are black-and-white drawings with fine lines. This makes the "visual recognition" part require different capabilities than what the state-of-the-art models for computer vision were designed for (natural images). The authors should first assess whether CNNs are suitable at all for this kind of visual input, regardless of the concept learning/reasoning aspects. This could be done by training a recognition task with a large training set of visual stimuli of this type (i.e. in a standard supervised setup).

Heterogeneous graph neural networks have become popular in various domains. However, their generalizability and interpretability are limited due to the discrepancy between their inherent inference flows and human reasoning logic or underlying causal relationships for the learning problem. This study introduces a novel solution, HG-SCM (Heterogeneous Graph as Structural Causal Model). It can mimic the human perception and decision process through two key steps: constructing intelligible variables based on semantics derived from the graph schema and automatically learning task-level causal relationships among these variables by incorporating advanced causal discovery techniques. We compared HG-SCM to seven state-of-the-art baseline models on three real-world datasets, under three distinct and ubiquitous out-of-distribution settings. HG-SCM achieved the highest average performance rank with minimal standard deviation, substantiating its effectiveness and superiority in terms of both predictive power and generalizability. Additionally, the visualization and analysis of the auto-learned causal diagrams for the three tasks aligned well with domain knowledge and human cognition, demonstrating prominent interpretability. HG-SCM's human-like nature and its enhanced generalizability and interpretability make it a promising solution for special scenarios where transparency and trustworthiness are paramount.

Microsoft is positioning itself as the world's platform for artificial intelligence, and that's a smart move. In 2014 I wrote that Microsoft's Cortana would be the next big thing. Redmond's vision for its johnny-come-lately AI is that it, like the GUI before it, will be pivotal in the evolution of the personal computing user interface. Microsoft's ambitions for Cortana were evident in 2014. Microsoft envisions an unbounded AI that developers and partners will incorporate into a range of everyday and innovative devices enabled by the Cortana SDK.