AI rivals the human nose when it comes to naming smells

3kCS...AYb8
4 Sept 2023
430

Have you ever wondered how your nose can identify different smells, such as coffee, chocolate, or flowers? How does it translate the chemical signals from the molecules in the air into meaningful names? And can artificial intelligence (AI) do the same thing?

📌In this article, we will explore how a team of researchers from the University of Reading and Osmo Systems have developed an AI system that can predict the names of smells from the molecular structures of odorants. We will also discuss the implications and applications of this breakthrough for various fields and industries.



How does the human nose work? 🧠

The human nose is a complex organ that can detect and distinguish millions of different odors. It consists of two main parts: the olfactory epithelium and the olfactory bulb.
The olfactory epithelium is a thin layer of tissue that covers the roof of the nasal cavity. It contains about 10 million olfactory receptor neurons, each expressing one type of olfactory receptor protein. These proteins bind to specific odorant molecules and trigger electrical signals that are transmitted to the olfactory bulb.

The olfactory bulb is a part of the brain that processes and integrates the signals from the olfactory receptor neurons. It contains about 2,000 types of glomeruli, which are spherical structures that receive input from a specific subset of olfactory receptor neurons. Each glomerulus corresponds to a distinct odor quality, such as fruity, floral, or spicy.

The pattern of activation of the glomeruli forms an olfactory code that represents the identity and intensity of the odor. This code is then sent to higher brain regions, such as the piriform cortex and the orbitofrontal cortex, where it is associated with semantic labels, such as apple, rose, or cinnamon.


How does the AI system work? 💻

The AI system developed by the researchers is called POM (Predictor Of Molecular Properties). It is based on a deep neural network that learns to map molecular structures to olfactory properties.

The researchers trained POM on a dataset of 5030 odorants with 21 perceptual descriptors, such as sweet, sour, woody, or minty. The molecular structures were represented by SMILES (Simplified Molecular-Input Line-Entry System), which are strings of symbols that encode the atoms and bonds in a molecule. The perceptual descriptors were represented by binary vectors, where each element indicates whether the descriptor applies to the odorant or not.
POM consists of two main components: an encoder and a decoder. The encoder converts the SMILES strings into fixed-length numerical vectors, called latent representations. The decoder converts the latent representations into binary vectors, corresponding to the perceptual descriptors.

The researchers evaluated POM on several tasks, such as predicting perceptual descriptors from molecular structures, predicting molecular structures from perceptual descriptors, and generating novel odorants with desired properties. They compared POM with several baselines, such as random guessing, linear regression, and random forest.
They found that POM outperformed all the baselines on all the tasks, achieving an accuracy of 90% in predicting perceptual descriptors from molecular structures. They also found that POM could generate realistic and diverse novel odorants with specified properties.


What are the implications and applications? 🌎

The ability to predict and generate smells from molecular structures has many potential implications and applications for various fields and industries.
📌For example, it could help:

  • Perfumers design new fragrances with desired characteristics and effects.
  • Chemists discover new molecules with specific functions and properties.
  • Biologists understand how different animals perceive and communicate with smells.
  • Food scientists enhance the flavor and quality of food products.
  • Medical professionals diagnose and treat diseases based on smell signatures.
  • Environmentalists monitor and control air pollution and waste management.
  • Artists create new forms of sensory expression and experience.


Conclusion 🙌

In this article, we have learned how an AI system can rival the human nose when it comes to naming smells. We have seen how it uses a deep neural network to map molecular structures to olfactory properties. We have also discussed some of the implications and applications of this breakthrough for various fields and industries.

😒I hope you have enjoyed reading this article and learned something new. If you have any questions or comments, please feel free to share them below.💬👇


And if you are curious about how POM works in more detail, you can check out their research paper.

I hope you like this article, and learn something new. If you have any question or comments, please feel free to share them with me. I would love to hear from you. And if you want to learn more exoplanets, astronomy, A.I and other science topics, please check out our other articles and resources.



📚 Sources:


  • (1) Science | AAAS | Science.
  • (2) AI rivals the human nose when it comes to naming smells | LifeScience.
  • (3) AI 'nose' predicts smells from molecular structures - Phys.org | Phys.org.



📖Read My Latest Posts :


Mythology

  • Kali Yuga: The Age of Darkness and Chaos | Link


Astronomy


Science and Health


Related to Time 🕰️


Science and Biology



Space Exploration


Artificial Intelligence


Technology


-----


👋 If you enjoyed this topic, we'd love to hear from you! Show your support by reacting and leaving a comment below. Let us know your thoughts, or any additional ideas related to this discussion. Your feedback is greatly appreciated and will help us improve and bring you more exciting content in the future. Thank you for being a part of our community! 👍💬😊

Write & Read to Earn with BULB

Learn More
Enjoy this blog? Subscribe to Saitama

14 Comments

B
No comments yet.
Most relevant comments are displayed, so some may have been filtered out.