How AI Learns to Smell with Alex Wiltschko
July 8, 2026
AI Summary
5 min read“I want to give computers a sense of smell.” That is the mission Alex Wiltschko, founder and CEO of Osmo and a former Google DeepMind researcher, described on the TWIML AI Podcast. Unlike vision or hearing, smell has largely resisted digitization because there has never been a practical map for it—no equivalent of RGB for color or frequency for sound. Wiltschko’s team is building that map, using graph neural networks and massive new datasets to model, predict, and design scents. The conversation covers how they cracked a century-old problem, why the embedding space they discovered is suspiciously similar to biology, and what it will take to bring smell into the age of foundation models.
The Missing Map for Scent
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 (01:56) **The Core Problem: Digitizing Smell** - Alex defines the three-step process (read, map, write) needed for computers to handle scent, and explains why the "map" is the missing piece.
- 2 (03:38) **Why Smell is a Harder AI Problem Than Vision** - Alex contrasts the human eye’s ~3 color channels with the nose’s 300+ olfactory receptor channels, explaining the biological basis for the problem's complexity.
- 3 (07:08) **The "Trillion Smells" Myth and Human Capability** - Alex debunks a famous study and asserts that humans are actually "freaking amazing" at smelling, capable of detecting molecules at parts-per-billion levels.
- 4 (09:30) **The First Breakthrough: The Structure-Odor Relation** - Alex describes the initial AI approach: using a graph neural network to predict the smell of a single molecule from its chemical structure.
- 5 (12:06) **Discovering the Principal Odor Map** - Alex explains how "cracking open" the neural network revealed a 300-dimensional embedding space that acts as a continuous, predictive map of smell.
- 6 (16:41) **From Map to Business: Designing New Fragrances** - The conversation shifts to application: using the map to design novel, safe, and manufacturable scent molecules for the fragrance industry.
- 7 (22:53) **The Data Moat and the Business Loop** - Alex details the self-perpetuating cycle where the fragrance business funds data collection, which in turn improves the AI models for future scent design.
+ Full timestamped outline available in the app
Guests on this episode
Show Notes
In this episode, Alex Wiltschko, founder and CEO of Osmo, joins the show to discuss his goal of giving computers a sense of smell and what it takes to build olfactory intelligence.
We explore the science behind smell, from the hundreds of olfactory receptors in the human nose to the challenge of mapping the relationship between molecular structure and odor, ensuring safety regulations are met, and building foundation models for smell. Alex explains how graph neural networks and advanced embedding spaces allow AI to capture the multi-dimensional structure of scents, grouping them into perceptual neighborhoods, and creating a machine learning representation that predicts how molecules smell.
We also cover how Osmo built the largest proprietary olfactory dataset from scratch to train a fleet of predictive models, and how olfactory intelligence could eventually power applications far beyond fragrance, including disease detection, emotion sensing, and consumer devices.
🗒️ Full show notes: https://twimlai.com/go/771.
More from this podcast
The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) →