The digital sense of smell is coming from zero to one, and AI "perfumer" predicts smells comparable to humans

To help test the new artificial intelligence "nose," Drexel University's Jonathan Deutsch spent several hours sniffing and describing the smell of unknown chemicals. Image source: "Science" magazine

According to a report in Science magazine on the 1st, a newly designed machine learning model by British and American scientists has reached a level comparable to the human sense of smell and can describe the smell of chemical substances in words. The researchers used it to "paint" odor maps that corresponded to hundreds of chemical structures, such as "fruity" or "grassy." This guide map could help researchers design new synthetic scents and may provide new insights into how the human brain interprets smells, marking another step closer to digitizing smells.

To explore the connection between the structure of chemicals and their smell, Osmo, a Google Research spin-out startup, collaborated with the Monell Chemical Senses Research Center in the United States, the University of Reading in the United Kingdom, and Arizona State University in the United States to design a neural network system that matches one or more of 55 descriptive words to a description of a smell. The team trained the AI ​​using an industry dataset, which includes smells from around 5,000 known odorants. AI also analyzed the chemical structure of each scent to determine the relationship between structure and aroma.

The system identified correlations between specific patterns in the structure of approximately 250 chemicals and specific odors. The researchers combined this relevant information into a master odor map (POM). This figure can be used as a reference when AI predicts the smell of new molecules.

To compare the smell levels of the POM with those of the human nose, 15 human volunteers matched specific odors to the same set of descriptive words used by the AI. Next, the researchers collected hundreds of odorants that don't exist in nature but are familiar enough to be described. They asked volunteers to describe 323 of these molecules and asked the AI ​​to predict the smell of each new molecule based on its chemical structure. As a result, the AI's guesses are very close to the average human response and closer to the correct answer. Specifically, the model performed better than the panel average in 53% of the molecules tested.

Editor-in-Chief’s Circle

The "ethereal" smell is both invisible and intangible. How can artificial intelligence predict it? Don’t forget that artificial intelligence can often show powerful advantages for things that are regular, logical, and computable. There is a special connection between the smell and structure of chemical substances. This is where AI can predict scents, and how it can compete with human perfumers.

(Source: Science and Technology Daily)

The above is the detailed content of The digital sense of smell is coming from zero to one, and AI "perfumer" predicts smells comparable to humans. For more information, please follow other related articles on the PHP Chinese website!