AI Startup Osmo Advances Scent Recognition for Health and Wellness

Key Takeaways

A new AI startup, created by former Google researchers, is making strides in teaching machines to understand and recreate smells.

Osmo's innovative technology could revolutionize industries like perfumery and virtual reality by building a digital scent library.

This development is part of a larger trend where AI is being used to enhance and replicate human sensory experiences.

A fresh startup is taking on the challenge of enabling artificial intelligence (AI) to recognize and generate scents.

Osmo, led by CEO and co-founder Alex Wiltschko, is on a mission to develop AI capable of comprehending and mimicking smells. This ambitious goal follows in the footsteps of AI pioneers like OpenAI, which has successfully trained AI to create text and audio.

Teaching Computers to Smell

Osmo began as a personal project for Wiltschko, a former research scientist at Google, and has now evolved into an independent startup over the past two years.

With a background in neuroscience and olfactory research, Wiltschko brings deep knowledge and credibility to Osmo's vision. The company is focused on improving human health and wellness by creating AI that can interpret and replicate scents.

https://twitter.com/philipvollet/status/1468830487892398080 

"We know that smell holds valuable information that can help us detect diseases. But currently, computers don't understand that language," Wiltschko shared in an interview with CNBC, emphasizing how AI scent recognition could benefit medical science.

Although many large language models (LLMs) have been used to diagnose medical conditions, the results have been mixed. For example, when scientists tested ChatGPT with over 100 medical case studies from Medscape, the chatbot accurately diagnosed only 49% of the time.

This underscores the need for more reliable solutions. Osmo is working toward this goal, with a long-term vision of using scent detection for disease diagnosis.

In the meantime, the company is concentrating on more immediate applications, such as improving the safety of consumer products like perfumes, shampoos, and laundry detergents.

To build their AI model, Osmo needed a vast dataset of scents. However, such a dataset didn’t exist, so the company created one from scratch.

Osmo collaborated with several reputable companies in the fragrance industry to gather thousands of molecules and scent descriptions.

This data is processed using graph neural networks (GNNs), which help the AI understand the molecular structures that determine a scent's characteristics.

GNNs, a type of deep learning model, are particularly effective at analyzing complex relationships and learning from interconnected data structures like molecules. While Osmo’s current focus is on practical applications like scent recognition, the company remains committed to its ultimate goal of using this technology for disease detection.

Artificial Intelligence

Anticipating Apple's Intelligence Features

AI’s Rapid Growth in Healthcare

The use of AI in healthcare continues to expand. The global AI market, valued at around $23 billion, is projected to grow to $431 billion by 2032, according to Polaris Market Research.

One area experiencing significant growth is the application of generative AI in healthcare.

For example, Pfizer and AWS have developed a solution called Vox, which assists medical professionals by summarizing medical resources, allowing them to dedicate more time to complex patient needs. Amazon Pharmacy also uses generative AI to streamline its prescription process and provide transparent pricing.