Earlier this week, Google revealed how its research team used machine learning to train an artificial intelligence (AI) to recognise different smells based on molecule shapes.
This new AI, which is still in the early stages of development, could revolutionize the fragrance sector as a whole and have considerable impact on the sector’s marketing and sales strategies in the long run.
A scientific paper published on October 25th details how this new AI predicts “the relationship between a molecule’s structure and its odor.” It describes how the Google Research team, in close collaboration with the University of Toronto, has assembled a novel and large dataset of expertly-labeled single-molecule odorants, and trained a graph neural network to predict the relationship between a molecule’s structure and its smell.
This research made considerable advancements on the scientific issue of Quantitative Structure-Odor Relationship (QSOR). The latter is a significant and on-going challenge in chemistry – and developing an efficient AI for QSOR would have a tremendous impact on the manufacture of synthetic fragrances, and sensory neuroscience as a whole.
In this context, the researchers have demonstrated state-of-the-art results on this QSOR task, and further stated that their model and its learned embeddings might be generally useful in the rational design of new odorants. However, predicting the relationship between a molecule’s structure and its odor remains a difficult, decades-old task, as the researchers admitted.
“Machine learning has already had a large impact on the senses of sight and sound. Based on these early results with graph neural networks for molecular properties, we hope machine learning can eventually do for olfaction what it has already done for vision and hearing,” the team concluded.