Monopolizing AI Technology Limits Innovation

Monopolizing AI Technology Limits Innovation

Author: Aron Walsh, ChemistryViews

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 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.

Professor Aron Walsh, Imperial College London, UK, works in the fields of materials design and discovery.


What fascinates you about AI?

We have entered a new era for computational chemistry where AI is helping us to tackle more complex problems. I am fascinated by the ability of these new techniques to discern patterns and relationships in vast datasets beyond the capability of previous methods. In solid-state chemistry, AI models are being used for tasks as diverse as the prediction of synthesis routes for novel crystals and the training of universal force fields that are both accurate and scalable. These developments open new avenues for materials discovery. The field is still evolving, with novel techniques and applications appearing weekly, so there is no chance of getting bored.


Is there anything we should fear?

I do fear any one company having a monopoly over any AI technology, especially in the realm of chemistry. This could restrict access to essential tools and slow down innovation.

The rapid advance in AI has been supported by open-source projects with accessible datasets and benchmarks. A collaborative ecosystem promotes transparency and benefits the entire field. There is also an ethical concern of ownership when proprietary models, based on unspecified training data, form part of a discovery process.

I am confident that we can overcome these issues as a community.


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

Practically, the best way to get started is to use Python in your research and get comfortable reading and processing data (using the pandas library) and making plots (using the matplotlib library). The next step of exploring machine learning models (using scikit-learn) then comes easily.

There are many online databases to play with. For those interested in crystals, I suggest the package matminer, which comes loaded with 45 datasets covering many types of materials and properties and examples of how to build your own predictive machine learning models.


Can you recommend a good article/video/website for beginners and one you enjoyed recently?

For a true beginner: “Applications of Artificial Intelligence in Chemistry” by Hugh Cartwright. It was published in 1994 but still does a wonderful job of getting across the central concepts of AI and its potential for chemistry. I am sure you can find a secondhand copy for a bargain price.

We tried to get some of the same excitement across in our review paper published in Nature:

For those who want to dig deeper, I recommend “Understanding Deep Learning” by Simon J. D. Prince. It has an approachable style, great figures, and practical exercises. It is also extremely up-to-date with coverage of cutting-edge topics such as diffusion models and deep reinforcement learning.  


Thank you very much for the insights.

back to the overview “Opinions on AI & Chemistry



Leave a Reply

Kindly review our community guidelines before leaving a comment.

Your email address will not be published. Required fields are marked *