چكيده فارسي :
Artificial intelligence (AI) refers to a vast set of computational entities that demonstrate reasonable and intelligent behavior in performing diverse tasks, ranging from simple to complex ones. AI tools can be categorized based on data we need to make them, algorithms they employ, and the way we interact with the tools. Three recent achievements in AI have disrupted many fields, including material science, and has increased hope to reduce the cost and experiment time of tedious material design process. The first achievement is significant increase in the accuracy of AI tools for modeling and behavior prediction of complex and large systems. The second one is mitigation of need for huge and integrated training data. And the last accomplishment is possibility of communication with AI tools in a natural language. The later one has paved the way to exploit documented human knowledge forlow-cost interactive self-learning in the field as well as to use AI as a cheap and 24h semi-expert assistant for analysis, synthesis, and evaluation of materials. Material science-aware prompt engineering and material-science-based large language models is essential for these purposes. AI tools for modeling and behavior prediction are trained by using heterogenous expert knowledge and cleaned big experimental data, which is sparce, commercially protected, and expensive to collect in the material science domain. Integration of new machine learning methods, such as generative and interactive ones, with martial science-based digital twins and computational AI-powered models, play a key role to compensate expertise and data sparsity as well as resolving the sever out-of-distribution experimental data problem. In addition, federated learning using distributed data of different data-owners facilitates data privacy protection while achieving shared benefits. Explainability and safety plus causal modeling are three hot research topics in AI. In material science applications, explainability and causal modeling play a central role for filtering AI results prior to experimentation. Interdisciplinary research for development of explainable and causal material science-aware AI tools is of high value.
