Artificial intelligence (AI) can be defined as a combination of data mining, machine learning, and statistical methods. The combination of these methods in materials science is gaining traction to satisfy industry demands for more efficient materials discovery, development, and commercialization. AI in advanced materials research and development is the logical next step in information generation after experiment, theory, and simulation. However, the problem with implementing AI has many facets, as described in a recent article published in the MRS Bulletin. The biggest challenge for researchers looking to capitalize on materials informatics has been the lack of data standards and central repositories combined with limited access to published data. Gaining valuable insights from artificial intelligence requires data collection and organization on an immense scale: the world’s materials data locked in tables, figures, and images.
Despite materials informatics receiving significant attention a decade ago, only with recent advances in computing and governmental support from the White House Office of Science and Technology Policy’s Materials Genome Initiative and the National Network for Manufacturing Innovation, has there been notable progress. Large, open-source databases, such as Materials Project and the NIST (National Institute of Standards and Technology) Standard Reference Databases, are examples of the open-source movement in a traditionally guarded field of research. The most valuable data, however, often comes from aggregating and analyzing decades of published research in a subject area. Researchers at the University of California Santa Barbara and the University of Utah have demonstrated the value of discerning patterns from manually collected data on thermoelectrics and lithium-ion electrode materials, respectively. The full benefit of materials informatics on materials discovery, development, and commercialization is likely to depend in part on such manual data collection and organization efforts.
Innovation managers, researchers, and professors alike will have to wait patiently for the ultimate Advanced Materials prediction platform, but those willing to compromise on this vision of a comprehensive prediction platform may see the beginnings of data-driven platforms that support the materials innovation pipeline. Companies like IBM, Nutonian, Enterra Solutions, and Citrine Informatics all tout AI, industrialized data science, big data analytics, and machine learning as they aim to deliver materials informatics to Innovation 1000 companies. The differentiator between these players will be the development of automated data collection and organization methods, such as machine vision, which mine relevant data sets and enable consistently successful AI predictions – something clients should keep in mind as they evaluate early-stage partnerships.
Dayton Horvath is a Research Associate at Lux Research. Lux Research provides strategic advice and ongoing intelligence for emerging technologies. Leaders in business, finance and government rely on Lux to help them make informed strategic decisions. Through their unique research approach focused on primary research and their extensive global network, they deliver insight, connections and competitive advantage to their clients.