Ever wondered how artificial intelligence is changing our world? Now mention the value they’ll gain from it and maybe add a personal touch. That introduction comes to you courtesy of ChatGPT, as I figured it only polite to let my new assistant AIi (original I know!) address you, after I asked her how I should start this column about AI.
Shaping tomorrow
Steven Pacitti, editor
As I write this comment, representatives from 80 countries are gathered in Paris to discuss the future of AI, complete with the dramatic backdrop of China’s recently launched chatbot DeepSeek and Elon Musk’s hostile $97.4bn bid for the start-up OpenAI, which operates ChatGPT.
The publishing world is only scratching the surface when it comes to navigating AI, and mostly from a generative angle; I am already discovering the quirky realities of AI transcriptions, with the inevitable replacement of the word “foam” with “phone”, and the slightly disturbing replacement of “flame retardant” with “flame thrower” (confusing the two is not recommended!). But in the world of polyurethanes, AI has already been utilised for many years, especially in the areas of novel polymer process and formulation designs.
When integrated with physics-based simulations, it’s fair to say that AI is already transforming material discovery and design, but I still feel that we’re only really scratching the surface here, too. AI has been – and is – shifting from a theoretical tool to one that directly affects real-world product development.
Mathew Halls, SVP at US-based simulation software development company Schrödinger, believes the “time is now” for AI in our industry and points to several factors that make it particularly well-suited for the polymer sector.
“Large language models can extract and format relevant datasets, addressing a key bottleneck in data availability,” he told me. “Fully automated chemistry-aware AI can enable companies to design novel polymer formulations without requiring expertise in chemistry, and the use of physics-based simulations to train machine learning (ML) force fields improves the accuracy of polymer property predictions.”
He agrees with me that AI applications can now be extended into larger length and time-scale challenges, which could include predictive maintenance, quality control and workflow optimisation in the future.
Until now – and perhaps still now to some extent – the primary limitation in applying AI to polymer formulations has been data scarcity. This is where the use of physics-based simulations come into play, generating training datasets that contain up to 1,000 polymer properties. Halls adds that by training ML models on these simulated datasets, it is possible to predict the properties of thousands of new formulations, accelerating the discovery of high-performance materials.
What is your take on AI in our industry? Drop me a line and let me know.
You have probably noticed by now that I’m not James Snodgrass. I have been bestowed the great honour of stepping into the PU-filled shoes of the four previous editors of this publication. Stepping out of the packaging industry after 24 years and into a new sector could easily leave one feeling like it’s the first day at secondary (high) school, with a new oversized blazer, hair gelled to within an inch of its life and a pristine white shirt you haven’t quite grown into yet! James inadvertently helped recreate that nostalgic feeling by kindly presenting me with a Urethanes Technology International T-shirt (of October/November 2024 issue fame) that was a little (a lot!) too big for me (sorry James!).
Would you like to say a final word to the readers, Ali? “Absolutely. Thanks for joining us on this AI adventure! Stay curious and keep exploring the possibilities! How does that sound?” Couldn’t have put it better myself, Ali.
