When you buy through affiliate links in our content, we may earn a commission at no extra cost to you. Learn 
A new startup is taking charge of giving artificial intelligence (AI) the ability to understand and generate smell.
Led by CEO and co-founder Alex Wiltschko, Osmo is dedicated to developing AI that can comprehend and replicate scents, an ambitious goal that builds upon the advancements of companies like OpenAI, which have successfully taught AI to generate text and sound.
The Nose Knows: Teaching Computers to Identify Smell
Osmo started as Wiltschko’s personal project while he was a research scientist at Google before he launched it as an independent startup two years ago.
With a neuroscience and olfactory research background, Wiltschko brings expertise and credibility to Osmo’s mission. The company aims to enhance human health and wellness by developing AI to process and understand scents.
https://twitter.com/philipvollet/status/1468830487892398080
“We’ve known that smell contains information we can use to detect disease. But computers can’t speak that language and can’t interpret that data yet,” Wiltschko explained to CNBC in an interview, explaining how AI scent detection can aid medical sciences.
So far, many LLM solutions have tried diagnosing medical conditions. Scientists have asked ChatGPT to assess over 100 case studies from the medical publication Medscape, but the chatbot could only achieve a correct diagnosis 49% of the time.
This highlights the need for more accurate and reliable solutions. Osmo is taking a step in this direction, with long-term goals including disease detection.
However, in the short term, the company is focusing on a more immediate impact: making consumer products like perfumes, shampoos, and laundry detergents safer for consumers.
To develop their model, Osmo needed a vast dataset of scents. However, Wiltschko explained that such a dataset didn’t exist, so they had to create one from scratch.
Osmo worked with a number of reputable companies in the fragrance sector, where they obtained thousands of molecules and scent descriptions.
This data is then fed into graph neural networks (GNNs), which helps their AI understand the atoms, bonds, and how the structure determines its odor.
GNNs are a type of deep learning that utilizes neural networks to analyze or learn from data with relational or interconnected structures, such as molecules. Using GNNs allows them to capture complex relationships, handle variability in molecules, and learn hierarchical features easily. While scents are a practical use case for Osmo now, the company still wants to ensure its goal of achieving disease detection a reality in the future.
AI Rapid Growth in the Health Sector
The use of AI has continued to increase. The global market is estimated at around $23 billion. Still, Polaris Market Research predicts it could be worth $431 billion by 2032 as more industries and governments invest in leveraging machine learning with medicine.
One area that has witnessed growth is the use of generative AI in healthcare.
For instance, Pfizer and AWS collaborated on a solution called Vox, which helps medical professionals summarize medical resources, freeing up time to focus more on the patient’s complex needs. Amazon Pharmacy also uses generative AI in its prescription process and offers transparent pricing.
