Scientists have long argued about how exactly the receptors of the human body allow us to perceive a wide range of odors and give them one or another description. In an effort to solve this problem, teams of engineers around the world were asked to create AI that could perceive odors as well as humans.
Predicting color is not that difficult: for example, if a light wave reaches 510 nm, then most people will say that it is green. But figuring out what a particular molecule smells like is much more difficult. 22 teams of scientists have created a set of algorithms that can predict the odors of various molecules based on their chemical structure. The full range of practical uses of the program remains to be determined, but the developers hope that, first of all, it will help perfumers, pharmacists and food workers develop new, unique combinations of smells.
The work began with a recent study by Leslie Vosshall and colleagues at Rockefeller University in New York, in which 49 volunteers were asked to guess 467 odors. For each of them, a comparison system was developed consisting of 19 basic patterns: the subjects said whether the smell was similar to fish or garlic, assessed the intensity and individual pleasantness of the aroma. As a result, a catalog was created, numbering more than a million cells, which characterize certain odorous molecules.
When computational biologist Pablo Meyer found out about this, he immediately saw the study as an opportunity to test whether a computer system could predict how people would judge odors. Despite the fact that researchers have discovered about 400 odor receptors in the human body, it remains a mystery to scientists how exactly they work together so that a person can distinguish even subtle shades of odors. In 2015, Meyer and his colleagues launched the DREAM Olfaction Prediction Challenge. Participants of the competition received at their disposal the same rating tables of volunteers describing smells, along with the chemical structure of the molecules that produce them. In addition, the participant provided a database of 4800 descriptions for each individual molecule - its atoms, their mutual arrangement, general geometry,which in the end amounted to about 2 million data points. Eventually, the data should be used to train computer programs to recognize odors based on structural information.
The competition was attended by 22 teams from all over the world, and although many have done well, two teams are worth highlighting. The Michigan team, led by Ian Phan Guang, was the best at predicting odors for individual items. Another team from the University of Arizona, led by Richard Guerkin, was the best at training the program for the average odor rating across the sample. Meyer reports this in an article published in the journal Science.
Of course, many scientists are skeptical about developments, saying that the work done, although it makes a significant contribution to science, is still a rather primitive selection, and 19 descriptive elements for the entire spectrum of odors in nature are clearly very, very few. Alternative studies with volunteers used 80 or more of these criteria to verbally evaluate different odors. It is unclear whether the existing algorithm will be able to correctly predict the odor assessment if it has to deal with such an array of information. So, today, the perception of odors remains a mystery for both physicians and engineers.