Innovation That Matters

| Photo source Argonne National Laboratory

Using ChatGPT-style AI to develop battery electrolytes


One company is using GPT language models to help discover and synthesise new and better battery electrolytes


Become a member today and get early access to the ideas transforming our world from just £39 per month*

Exclusive member benefits:

  • Access to over 13,000 innovations
  • Monthly horizon scanning reports
  • Exclusive feature articles

Already a member? Sign in here

Spotted: Creating new molecules in chemistry is a difficult process of trial and error. Adding one atom to see how it changes a molecule’s properties may not be too challenging, but combine several atoms in multiple places and the process becomes difficult and slow. 

Now, researchers at the University of Michigan may have a solution. They believe that the utilisation of GPT language models can be applied to more than just predictive language, they could also be instrumental in predicting and creating new battery electrolytes – the solution inside batteries through which electrons travel from one electrode to another during charge and discharge cycles.  

Perhaps the most familiar language models to us are GPT models, like those used in the Chat-GPT application. This artificial intelligence (AI) technology relies on crunching incredibly large data sets to essentially predict the most likely word to come next in a sentence. On a large enough scale, this means the AI can form coherent and insightful sentences. This concept can be re-specialised to a chemical model that helps predict better electrolytes.

Much like Chat-GPT is fed raw text without annotation, the chemical model will be fed raw data in the form of text-based atomic structures without additional information on their known chemical properties. Once the model has ingested enough of this data to be able to predict the missing atoms in certain small molecules, the researchers will move on to ‘fine-tuning’, where the model will be fed the chemical properties of some compounds before being asked to predict the properties of others based on what it has learned.

Once this process of fine-tuning is complete, the researchers hope they will be able to feed in requests for molecules with certain properties (in this case, an electrolyte that’s compatible with both electrodes) and have the AI tell them the composition of the molecule. Hopefully, this will enable the creation of batteries with superior energy densities compared with existing lithium-ion batteries.

The work is being made possible by a one-year grant from the Department of Energy, which is providing access to Polaris – a supercomputer at Argonne National Laboratory to run the model.

AI technology has come on leaps and bounds, and is now being used across a variety of industries. Springwise has also spotted AI that helps clinicians navigate research overload as well as an AI that helps speed up home building

Written By: Archie Cox



Download PDF

Springwise Services:
Our expertise in spotting the latest innovations is the best resource to empower your team’s future planning.

Find out More