Can AI Smell? Olfactory Intelligence Could Help Detect The Invisible

A smell can usually tell us when something is burning, that food has gone off or that the rain is on its way. What we notice as a scent is actually the result of molecules moving through the air and reaching our noses.

These molecules can carry a lot more information than our senses can actually pick up. Volatile organic compounds (VOCs) are released by everything from food and industrial processes to the human body, and changes in their concentration can reveal that something has changed.

Researchers are building electronic noses that can detect these chemical patterns and using machine learning to interpret the data they produce. Rather than teaching a machine to experience smell like a person does, we can turn the chemistry behind a smell into information that technology can recognise and analyse.

Technology Behind Artificial Smell

Electronic noses, or e-noses, use a variety of gas sensors that respond differently when exposed to particular chemicals. Instead of producing a simple answer about what something smells like, the sensors generate a pattern of signals based on the compounds they encounter.

Machine learning systems can analyse those patterns and compare them with previous measurements. With enough data, an algorithm can learn that certain combinations of signals are associated with specific substances, materials or conditions.

According to research published by Nature, e-noses combine multiple gas sensors with algorithms that can analyse these patterns and help identify different gases and chemical signatures. The technology is being explored for uses including environmental monitoring, food spoilage detection and healthcare.

The sensors can also pick up information that our noses might miss. Humans can struggle to detect very small amounts of VOCs, while certain hazardous gases have no smell at all, which makes chemical sensors useful for detecting changes that could go unnoticed.

What Can A Chemical Pattern Tell Us?

Food is only one example of what these chemical patterns could reveal. However, in industrial settings, changes in the chemicals in the air could help identify a leak or show that something in a process has changed, while environmental monitoring could use the technology to detect pollutants and other changes that aren’t immediately obvious.

An e-nose doesn’t need a chemical to have a recognisable smell for it to be detected. It can also collect information from the air and use machine learning to work out whether that pattern matches something it has seen before.

Could A Machine Detect What We Can’t?

Our bodies release VOCs through things like breath, skin and urine, and changes in these compounds can sometimes reflect changes happening inside the body. That has researchers looking at whether the chemical information we leave behind could reveal something about our health.

Cancer is one area being explored, with researchers investigating whether e-noses can recognise chemical patterns in a person’s breath that could be linked to the disease. Research found on PubMed has shown results for using e-noses to detect cancer from exhaled breath, although the studies also highlighted issues such as small sample sizes and differences in how the technology was tested.

The research isn’t suggesting that a machine can “smell” cancer; it could potentially be another way to see if these chemical signatures can provide more information when looking for signs of disease.

What Are The Limits?

There is a difference between detecting a chemical pattern and knowing what caused it. An algorithm can learn that certain combinations of signals usually appear alongside a particular condition, but the pattern itself doesn’t necessarily explain why those chemicals have changed.

The sensors also need to work consistently outside a controlled laboratory because temperature and humidity can also affect readings, while the air around us can contain many different chemicals at the same time, making it harder to separate one signal from another.

Spotting a pattern once isn’t enough, as the system would need to recognise the same changes consistently across different places, conditions and samples for its results to be reliable.

A New Way To Read The Air

An e-nose could give technology another way to understand what is happening around us, picking up chemical changes that our own senses aren’t always able to notice.

These systems can turn changes in the air into data that can be measured and analysed, which gives us another way to notice when something has changed even when there is no obvious sign of it.

We already use technology to see and hear things that we can’t detect on our own. If artificial olfaction becomes accurate and reliable enough, it could add another sense to that list, giving us access to chemical information that our own noses just can’t pick up.