New Research: Artificial Intelligence Can Describe Scents 'As Well as Humans'

The result of a new study shows that artificial intelligence is capable of identifying the scent of various compounds by analyzing their molecular structure. Scents are the only type of sensory information that go directly from the relevant sensory organ - in this case, the nose - to memory and emotional centers of the brain. Due to this direct pathway, aromas can evoke specific memories and trigger emotions. Alexander Wiltschko, a neuroscientist at the startup 'Osmo' in Cambridge, Massachusetts, along with his team, has designed a type of AI system that can describe the scent of nearly 5,000 aromatic substances using 55 words by examining the relationship between the structure of a chemical and its scent. This AI system has also analyzed the chemical structure of each scent to determine the relationship between structure and aroma. The results of this research were published in the scientific journal 'Science'. Predicting the scent of perfumes based solely on chemical structure is difficult because molecules with similar chemical structures may have completely different scents, while the scent of two molecules with very different chemical structures can be the same. This system identified about 250 relationships between the structure of a chemical and a specific scent. Researchers mapped these relationships in a 'main scent map' so that the AI could use it when predicting the aroma of a new molecule. To test this map against the human nose, researchers trained fifteen volunteers to associate specific scents with the same set of descriptive words used by the AI. Next, they collected hundreds of aromatic substances that are not found in nature but are familiar enough for people to describe. They asked volunteers to describe 323 of them and also asked the AI to predict the scent of each new molecule based on its chemical structure. The result was that the AI's guesses were very close to and even better than human responses. Alexander Wiltschko states that the next step is to 'understand how scents combine' and create something that the human brain perceives as a completely different scent from any of the individual fragrances. Researchers say this will be very challenging, as combining just 100 molecules in different combinations of ten creates 17 trillion new combinations, and the number is so large that analyzing them becomes difficult for computers, while humans perceive scents in this way. This report was compiled using sources such as 'Science' and 'Nature'.