Google AI has developed an artificial intelligence model that compares the structure of molecules to their odour. It is reported that it will help develop certain food tastes or detect compounds that repel disease organisms.
Google AI also spoke about the project on social media.
Today we interduce an ML-generated sensory map that related issues of molecules and their operational products, incorporating the adoption of codes from unseine moecules and providing a voluntary tool to add global health issues. https://t.co/wmiq6wPKv5
The most famous examples of these maps relate to color vision, including the color circle, and the more complex options used to record color in the video are reported by Google AI in a press release. There are more than 300 smellful receptors, unlike the three color sensors of the human eye. It is therefore very difficult to create a map of odour.
In 2019, Google's artificial intelligence developed a model of a graphal neural network. It's a type of neural network that works directly with the graph structure. The typical application of GNN is the classification of nodes.
Now the Google AI network has started to explore thousands of examples of different molecules in combination with names of smells, such as "metal", "color" or "mint." The goal is to learn how to connect the structure of the molecule with the probability that it will have a special odour mark.
In the end, engineers create a map that links the structure of the molecule to the odour it emits, and the neural network measures how close molecules are in terms of odour. In total, scientists used more than 5,000 molecules from two different sets of tastes and odour data.
In addition, in a separate study by Mainland's colleagues, a neural network was used to create a map linking the molecular structures of mosquitoes to the extent to which insect smells were perceived as intimidating.