By Konstantinos Paraskevoudis, Data Scientist

Non experts most likely have heard the terms “Big Data” and “Data Science”. What about “Machine Learning” and “Artificial Intelligence”? From taking and processing photos on your smartphone’s camera to clicking on fine tuned recommended news on your browser, one thing is for sure: you have definitely used tools that are built upon these terms. We as users get in touch with AI -without even noticing- through popular or commercial applications which are used in a daily basis, which in turn gives a valuable training data set to feed such models! Email spam detection, news recommendation, bank fraud detection, text-based sentiment analysis are only some of them. However, what happens in the background of AI, away from the spotlight and usually not available as a commercial application for people to use, can be even more exciting.

Lights on Materials Development now; this is one of such fields, where Artificial Intelligence has considerably speeded up the applied research and has offered solutions that would have not been possible otherwise. Conducting traditional experiments (or computational modelling) can be consuming in terms of time and resources. The recent boost of Deep Learning and the development of various state-of-the-art algorithms in addition to the availability of large datasets has made it possible for researchers to overcome these challenges: using AI algorithms that iteratively learn from data, computers can find hidden insights without being explicitly programmed where to look. This enables the parsing of large volumes of existing knowledge, which is not possible for humans to do so!

AI is used nowadays for accelerating materials discovery and design. Specifically, Machine Learning algorithms are used for understanding and extracting relationships between properties of materials (macroscopic or microscopic) and related impact factor such as design parameters. Such tools have successfully replaced traditional experiments, which can have a complex setup, cost time and resources and can be often faulty. Regarding materials discovery, the experimental and computational screenings for new materials discovery involve element replacement and structure transformation. Again, these methods consume a lot of time and resources. Researchers have been using AI to parse enormously large amounts structural and component material data. Trained models on these data are fed with a desired output property and suggest recommended structures and/or compositions of materials that can achieve this.

The field of Artificial Intelligence will not stop making huge development steps within the next years!  Along with this progress, fields of application such as Materials Science will also grow. The global adoption of Deep Learning in Materials Development is just a matter of time..