How AI Learns to Smell with Alex Wiltschko - #771

2026-07-08 · Source: The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Life Sciences & Biology, Data Science & Analytics · Depth: Intermediate, quick

Summary

Alex Wiltschko, founder and CEO of Osmo, is spearheading the development of "olfactory intelligence" to enable computers to perceive smell. This ambitious goal involves deciphering the intricate science behind smell, from human olfactory receptors to mapping the relationship between molecular structure and odor, while adhering to safety regulations. Osmo utilizes graph neural networks and advanced embedding spaces to capture the multi-dimensional structure of scents, organizing them into perceptual neighborhoods and creating machine learning representations that predict how molecules smell. The company built the largest proprietary olfactory dataset from scratch to train its predictive models, envisioning applications far beyond fragrance, including disease detection, emotion sensing, and consumer devices.

Key takeaway

For AI Scientists and Machine Learning Engineers developing novel sensory intelligence, this work highlights the potential of graph neural networks and advanced embedding spaces. You should consider these techniques for mapping complex molecular structures to perceptual outcomes, especially when building large, proprietary datasets. This approach extends AI capabilities beyond traditional senses, opening avenues for applications in areas like disease detection and consumer devices.

Key insights

AI can learn to "smell" by mapping molecular structures to odor perceptions using graph neural networks and proprietary datasets.

Principles

• Mapping molecular structure to odor is fundamental.
• Olfactory intelligence must meet safety regulations.
• AI groups scents into perceptual neighborhoods.

Method

Osmo uses graph neural networks and advanced embedding spaces to capture multi-dimensional scent structures, creating machine learning representations from the largest proprietary olfactory dataset to predict molecular smells.

In practice

• Develop AI for fragrance applications.
• Explore disease detection via scent analysis.
• Integrate scent AI into consumer devices.

Topics

• Olfactory Intelligence
• Graph Neural Networks
• Molecular Structure
• Machine Learning
• Sensory AI
• Proprietary Datasets

Best for: AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• Structured Representation Learning with Locally Linear Embeddings and Adaptive Feature Fusion — A neuroscientifically inspired RL framework disentangles dynamics and reward features using LLEs and adaptive fusion for improved learning.
• Batch-Invariant Spectral Intelligence for Robust and Explainable Insect Authentication — BISN uses learnable preprocessing and adversarial training to achieve robust, explainable insect authentication despite batch variations.
• Building the blocks of life — Integrating physics and geometric deep learning enhances protein modeling, improving generalizability and reducing data needs.
• EvoForest: A Novel Machine-Learning Paradigm via Open-Ended Evolution of Computational Graphs — EvoForest evolves computational graphs and functions, guided by diagnostics, to discover optimal computations for structured prediction.
• Is Fixing Schema Graphs Necessary? Full-Resolution Graph Structure Learning for Relational Deep Learning — FROG enables learnable graph structure for Relational Deep Learning by modeling table roles and enforcing semantic consistency.
• IEEE Transactions on Neural Networks and Learning Systems, Volume 37, Issue 4, April 2026 — The volume showcases diverse neural network and learning system advancements across various applications and theoretical foundations.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence).