Artificial intelligence is sniffing out what makes some drinks special

Artificial intelligence (AI) is already being deployed in many settings, from driverless cars to smart home devices, and researchers around the world are racing to expand its uses.

One German scientist has built AI into a giant plastic nose that gives you a big smile when it works out what it has found.

Researcher Horst Hellbrück holds the nose over a glass of Irish whiskey to demonstrate its skills. Seconds later, a green smiley face lights up on the back. The AI in the computer has identified the smell as whiskey, saying it is 99 per cent certain this is what’s in the glass.

“We are not yet as good as a sniffer dog,” says Hellbrück, pointing out that his device cannot yet distinguish between Scotch and bourbon, let alone individual types of whiskey.

Beyond whiskey, the small computer can smell out dangers, too, thanks to four sensors able to measure carbon monoxide concentration and distinguish substances in ambient air.

“We want to use it to show how AI can be applied. You no longer need a mainframe computer for this,” says Hellbrück. The individual components for his device cost less than US$110.

So far, the nose can easily distinguish between coffee, whiskey and the air in the room. Even more helpfully, the AI can warn people if the concentration levels of a hazardous substance reach an excessively high level – at an industrial workplace, for example, before it causes problems for people, says Hellbrück.

Scientists around the world are working on teaching AI to smell and while the computers cannot yet match an animal’s ability to identify odours, some take an approach that is close to nature.

At Massachusetts Institute of Technology (MIT) in the US, researchers were impressed to find that a computer model managed to teach itself to smell in minutes, building a neural network that resembles olfactory circuits animal brains use to process odours.

“Animals from fruit flies to humans all use essentially the same strategy to process olfactory information in the brain. But neuroscientists who trained an artificial neural network to take on a simple odour classification task were surprised to see it replicate biology’s strategy so faithfully,” according to an MIT publication.

Researchers at the University of Loughborough in Britain have been working for some time on using AI’s ability to smell as a diagnostic tool. “My colleagues and I are developing an AI system that can smell human breath and learn how to identify a range of illness-revealing substances that we might breathe out,” Andrea Soltoggio, a senior lecturer at Loughborough University, wrote in a blog post.

AI is already being used in medical settings, and the German institute’s staff are working on technology to carry out eye examinations through a smartphone to enable family doctors to make initial diagnoses.

The project aims to ensure people in rural areas get access to better levels of healthcare, without necessarily having to travel long distances to see specialists.

“But we don’t want to replace the ophthalmologist with this,” Hellbrück says. These efforts should help diabetics, for example, enabling them to attend check-ups at their GP instead of having to see a specialist.

While some doctors might not want machines interfering with their area of expertise, others might be more open to having support for routine tasks, he says.

“Artificial intelligence can be a help in medicine,” says a German health insurance specialist, adding that practical tests need to demonstrate whether the technology is able to support doctors in making diagnoses or treatment.

AI has to make doctors’ everyday lives easier and bring additional benefits for patients, though technology is likely to be used as an add-on, with the doctor-patient relationship central to the healing process even in the digital age.

“I am an engineer who wants to improve people’s lives,” says Hellbrück. “I want to be able to detect diseases better.”

The government of the northern state of Schleswig-Holstein, where he works, values the research and is stumping up ¤9 million (HK$74.3 million) for 12 new AI professors at the universities of Flensburg, Kiel, Heide and Lübeck.

They are to start teaching in the winter term. “Data is the raw material for the use of artificial intelligence,” says Dirk Schrödter, head of the state chancellery.

However, not everyone welcomes the use of AI and many scientists are understanding about people who are fearful that use of artificial intelligence could get carried away.

“You have to explain AI,” says Hellbrück. “And science fiction films in particular don’t help.”

He points out that people are already living with AI in settings from image processing to computer hotlines. A decade from now, we will not be able to imagine life without it, he says.

“Often people don’t even realise that they are dealing with AI. When you feel the benefit, you also lose your fear of it.”