Ever pondered over Alexander Graham Bell's question, "Did you ever try to measure a smell?"
Well, the world of artificial intelligence has taken a whiff at this challenge! The quest to measure odours, much like we do with images and sounds, is no longer a pipe dream but a fascinating reality being pursued by scientists and tech start-ups alike.
The human nose is essentially a chemical detector, sniffing up volatile organic compounds (VOCs) that give off smells. But what if we could replicate this process with machines? Enter Osmo, an olfactory AI start-up aiming to "give computers a sense of smell". The goal? To digitally encode scents, much like we do with sights and sounds.
Why, you might ask? Well, some VOCs can be harmful and worth detecting before they reach our noses. Others might signal dangers like gas leaks or concealed explosives. Some might even indicate health issues. A breath that smells like freshly mown clover? It could be a sign of liver failure.
The challenge lies in mapping molecules' structures to their perceived smells. This is where artificial intelligence comes into play, specifically a form of AI called a graph neural network (GNN). By training the GNN using large datasets linking molecular structure to odour, researchers can now predict odours for molecules without needing to synthesise them first.
The potential applications are endless. From revolutionising the flavour and fragrance industries to aiding in disease prevention and pandemic tracking, the digitisation of smell could transform numerous sectors. Imagine being able to recreate any scent as a printer recreates a picture. Or perhaps, bringing back to life novelties like the 1950s cinema format Smell-O-Vision.
While we're still at the early stages of this olfactory journey, one thing is certain: we're inching closer to answering Bell's question with a resounding "yes". So, next time you take a whiff of your morning coffee or the freshly cut grass, remember, there's a whole new world of scent exploration unfolding right under our noses!
Made with TRUST_AI - see the Charter: https://www.modelprop.co.uk/trust-ai
Comments