Students Need to Learn and Understand How AI Works

The intersection of chemistry and artificial intelligence (AI) is a fascinating area that attracts a lot of attention in both research and industry. We talked to people working in the field about the potential of AI to revolutionize chemical research, but also about concerns, (current) limitations, and ethical implications for chemical applications. We also asked for ideas to try or experiment with, as well as useful articles and videos for beginners and advanced users.

Associate Professor Oliver T. Hofmann, Institute of Solid State Physics, TU Graz, Austria, works in the field of computational surface science.

What fascinates you about AI?

Artificial intelligence in its many forms has the potential to greatly speed up the productivity in science and other fields. Different forms of artificial intelligence aid in many different tasks, from image analysis to text generation to simply smarter and better analysis of data. The really fascinating part is that most forms of AI are easy to use, and sometimes even easy to implement yourself. I strongly believe that this low-threshold availability, together with the often astonishingly good results, are the reason why AI is here to stay.

Is there anything we should fear?

Fearing new technology is always ill-advised. Nonetheless, we should be carefully considering what we do with this new technology.

A major risk is that AI will hardly be 100 % correct, and we have to evaluate carefully if we’re willing to accept false positives or false negatives in sensitive areas, such as law enforcement and medicine. Also, despite its name, AI is not actually intelligent. It only learns from the data that it is provided. This bears the risk that AI perpetuates implicit or explicit biases in the data that are already available.

In my opinon, the only sensible way to deal with these issues is education—students need to learn and understand how AI works, in order to correctly assess its risks.

Do you have something for our readers to try out or experiment with?

For people interested in structure search of molecules on surface, we have developed the SAMPLE code, available at www.map-design.tugraz.at/

SAMPLE, short for Surface Adsorbate Polymorph Prediction with Little Effort, is a software package that facilitates surface structure search for commensurate organic monolayers on inorganic substrates, by using coarse-grained modeling and machine learning.