A glimpse of AI in chemistry: innovation, human capital, and partnerships

March 14, 2024

The Dutch chemical industry grapples with pressing challenges, such as green energy supply stability, the need for enhanced facilities and infrastructures to facilitate efficient scaling up of chemical processes, and the complexities surrounding material transition. These challenges underscore the imperative for increased innovation and robust public-private cooperation. Marco Tibaldi, from the Amsterdam Chemistry Network, explains what role Artificial Intelligence (AI) can play in surmounting these challenges.

How can AI be part of the solution for innovations in chemistry?

“When it comes to AI and innovation, we should keep in mind that the problem definition is crucial. It is the human teaching the AI how to find the solution, especially with supervised learning models.”

“Take the AI tool GNoME (Google DeepMind) for example. GNoME is capable of generating new, stable materials. The next step will be identifying certain properties needed for an application (which might be batteries, or catalysts, or light-to-electricity conversion) and stirring the algorithm towards generating solutions with a high chance of actually being useful.”

What are the steps on the road ahead?

“Concerning AI and chemistry, the public interest has been triggered, but it will take some years for the best ideas to become adopted. There will be products designed to be commoditized, easy to integrate into the daily activities of any R&D department. This will allow the non-technical managers to embrace the innovation (and the investment needed). From design-of-experiments to retrosynthesis, from literature screening to digital lab assistants, I picture AI as a decision support system for the chemists of the future.”

[Check the outstanding demo of Google Gemini, and its potential in Unlocking insights in scientific literature here.]

Are public-private partnerships important for AI in chemistry?

“Of course: big companies can afford sophisticated hardware and computing power, while academia generates the talent needed to solve complex and interdisciplinary challenges. In the Netherlands we have an outstanding human capital when it comes to AI and other ‘beta’ sciences. Academia has the role to offer more interdisciplinary education, making a natural choice for a chemist to follow courses in Machine Learning and AI.”

“However, companies should remember that the already existing expertise in AI applied to science is not readily available everywhere in the world. They should play their role in nurturing this talent.”

“When it comes to new molecules in pharma, we see that new players like Google, Microsoft, NVIDIA, and even the mother company of TikTok (Bytedance) are hiring teams of molecular experts to work with AI researchers. I wish to see more experts in digitalization entering the chemical industry as well.”

Does that mean there is a need for investments?

Exactly, I suggest readers have a look at this letter published in de Volkskrant on the problem of the AI infrastructure in the Netherlands. To give a quote (in Dutch): ‘Uit onderzoek van Brookings blijkt dat commerciële platformen 29 maal meer rekenkracht hebben dan universitaire platformen. Zonder eigen AI-infrastructuur moeten wetenschappers hun toevlucht nemen tot buitenlandse platformen en raken Nederlandse bedrijven achterop als alle data, getrainde algoritmen, en het talent in het buitenland verblijven.’ Read the full letter here.

For more thoughts on the latest developments in chemistry, please check out Marco Tibaldi’s blogs on Chemistrynl