Google Trains AI To Smell
Google’s algorithms can identify odors based on their molecular structures.
For years now, perfumers and researchers have struggled to predict the relationship between a molecule’s structure and its scent. Scientists can look at a wavelength of light & identify what color it is when it comes to fragrances, and researchers can’t only look at a molecule and detect its odor. A research team from the Google Brain Team are hoping AI might change the confusion. In a research study published on Arxiv, they explain how they’re training AI to recognize smells.
Google scientists created a data set of nearly 5,000 molecules identified by perfumers, who labeled the molecules with descriptions ranging from “buttery” to “weedy” and “tropical.”
Google Trains AI To Smell- How Did Google Achieve This?
The team used about two-thirds of the data set to train its AI to connect molecules with the descriptors they often receive. Google researchers then used the remaining scents to test the AI. Surprisingly, the AI passed. The AI’s algorithms were able to predict molecules’ smells based on their structures.
For the novice, any two people might describe the same scent differently; for example, “woody” or “earthy.” Sometimes molecules have the same atoms and bonds, but they’re arranged as mirror images and have an entirely non-identical smell. Things get even more complex when you start combining various scents.
Google Trains AI To Smell- More On AI Research
Still, Google researchers believe that training AI to associate specific molecules with their scents is an essential first step. It could have an impact on chemistry, the understanding of human nutrition, sensory neuroscience, and how we manufacture synthetic fragrance.
Google isn’t alone. In Russia, Artificial Intelligence is being used to sniff out potentially deadly gas mixtures, and IBM is experimenting with AI-generated perfumes.