AI now has a nose too

This is according to a customer story that was presented to journalists who recently visited the IBM research center in Zurich in which the heads of Symrise, a manufacturer of fragrances for perfumes, cosmetics and food for humans and animals, participated

We have heard a lot about the use of artificial intelligence for applications such as artificial vision (eyes) and transferring content from voice to text (ears), but it turns out that a computer based on artificial intelligence can also be used as a nose for perfume developers.

So far machines can't smell, at least not like humans. But as a result of joint research by fragrance maker Symrise AG with IBM, perfume creation can now be added to the list of things AI-based robots will be able to do in a human-like manner. This is according to a customer story that was presented to journalists who recently visited the IBM research center in Zurich in which the heads of Symrise participated, headed by Achim Daov, the company's CEO, and David Appel. Vice President of the company's fragrances.

According to Daub, "Symrise is a German company traded on the Frankfurt Stock Exchange with sales of 3 billion euros. We are also a global company. Our products are sold in over 200 markets. We work with the leading brands in the field of perfumery, skin care and more - we are a B2B company. We employ 70 fragrance experts, and each of our products is unique. Because we are driven by creativity and innovation, we are always pushing the boundaries of innovation, especially in adopting new technologies and new opportunities that technology creates."

"Perfume development is an art, it cannot be replaced with artificial intelligence, but artificial intelligence is used by us to support the decision-making process for developing new fragrances for food for humans and animals, for the cosmetics industry and other industries," he added.

A system based on artificial intelligence, similar in some ways to IBM's Chef Watson
Therefore, Symrise and IBM collaborated in order to develop Philira - an artificial intelligence-based system, similar in some ways to Chef Watson, IBM's artificial intelligence system that surprised the restaurant world by recommending interesting recipes using unexpected combinations of foods and flavors. The companies took Chef Watson as a starting point and adapted it to deal with fragrances and chemical formulas.

Richard Goodwin, an IBM research scientist, said at the event that Philira has been trained to identify similarities between the formulas and 1.5 million existing fragrances that include strong fragrances (such as women's perfume and men's aftershave) as well as more "modest" scents used to make shampoo and candles.

The company's people worked according to formulas of perfumes that were commercially successful or that customers liked them, and then ran a series of algorithms including deep neural networks and supporting machines. or dissatisfaction with a new smell. Then a series of algorithms – including deep neural networks, vector systems as well as a random component – ​​all looked for hidden correlations in the data.

According to Goodwin, "Analysis begins with measuring the "distance" between fragrances. We prepare the contract model how close two fragrances are going to be, or find among three fragrances the two most similar. This allows us to identify white spaces - those areas where there are no fragrances similar to those already in our catalog."

Fragrance manufacturers use approximately 3,000 ingredients (flowers, fruits, and countless other scents) that serve as raw materials to work with to create their blends, and each fragrance formula can contain anywhere from seven to 100 ingredients. Together, an almost infinite number of potential combinations can be created.

According to Apple, "Artificial intelligence offers new scents from a huge variety of recipes. After analyzing the thousands of molecules of existing perfumes, she creates scents that stimulate the imagination - which is what every perfume developer wants."

However, you still need a chemist to produce the compounds, and a scent expert, like Apple, to try the scent and improve the formulas, even if artificial intelligence has made his life easier, because he doesn't have to plan combinations himself, and sometimes he is even surprised by the scent combinations offered by the software. Except that it does the planning operation in minutes - an operation that previously took months of human trial and error, he added.