AI that mimics aspects of the behaviour of flies seeking food can be used to design new beer recipes. It can also accurately recreate an existing brew using alternative ingredients when supplies are unstable.
at the University of Greenwich, UK, and his colleagues say that brewers traditionally follow existing recipes or create new ones using the time-consuming process of trial and error. He says that the AI tool is intended to 鈥渇lip this whole equation around, so instead of making brewers dependent on recipes, they will be able to pick the characteristic of the beer first鈥�.
A user can input the required alcohol content as a percentage, the level of bitterness and desired colour, and the AI will find recipes that match the criteria as closely as possible based on the brewer鈥檚 stock of ingredients.
Advertisement
鈥淭hey can list the ingredients they have into the system and then specify the beer that they鈥檇 like to have. So they would say, I would like a beer with this alcohol content, with this bitterness and with this colour, and then the AI model will try to get them as close as possible to what they want,鈥� says al-Rifaie.
The AI works by generating a large number of possible recipes at random and then assessing how close each is to producing a beer with the required characteristics, giving each a score. The researchers compare their approach to the way that real social insects alert others when they find food.
These recipes each represent a fly searching for food, and they give out a signal to other 鈥渇lies鈥�, or recipes, to announce the quality of their solution. This encourages 鈥渇lies鈥� with recipes that fit the desired criteria less well to reject their previous solutions. They instead move towards the quality 鈥渇lies鈥� and devise a new recipe with similar features, with the aim of finding a solution that is even better.
Al-Rifaie says that the tool would be useful for large brewers that want to create consistent products while changing ingredients, perhaps due to rising costs or limited availability. But it could also be used by smaller breweries to create new beer recipes.
He says that future versions may also include the option to tweak other desirable variables, including the foamability of the beer, as well as flavour and aroma qualities, although the latter two are more subjective than the criteria the AI currently factors in.
Sam Dickison at in London says that the AI could be useful to him as a kind of 鈥渋nspiration generator鈥�, allowing him to play with different ingredients and get a sense for what they would produce, but that this was a process that he had always enjoyed doing manually.
鈥淚n my head I know what all the malts and hops taste like and how they react, and that has been from trial and error, putting ingredients into beers, seeing how they act and then tweaking them,鈥� he says. 鈥淚n craft beer it鈥檚 so much about experimenting with the flavours of different ingredients, which is a little bit more esoteric than just kind of numbers and data.鈥�
Reference:
