Not that I’ve given it all that much thought to it, but if I’d been asked, I don’t think that I’d have put brewing beer very high on the list of candidates for AI involvement.
Not that the beer industry is any stranger to using up-to-date technology. It’s widely used in brewing’s production and logistics processes. But making beer that tastes better? I would have said that this is more art than science. Sure, the major market share industry players with the more widely-known and consumed brands are more focused on the “science” parts of production and logistics – after all, a Corona’s a Corona and a Bud’s a Bud. And the microbrewers (not to mention the homebrewers) would come down more on the “arts” side, using trial and error to come up with the ideal mix.
Of course, even Corona and Budweiser are always introducing new products, and whether you’re one of the big guys or one of the little guys, creating a beer that tastes good isn’t easy. Figuring out whether – to borrow from an ancient Miller ad – a beer tastes great and/or is less filling can involve drafting (and educating) employees and “civilian” beer drinkers to act as taste testers for their products. But, as a recent MIT Technology Review article said, “running such sensory tasting panels is expensive, and perceptions of what tastes good can be highly subjective.”
Enter AI.
Research published in Nature Communications described how AI models are being used to find not only how consumers will rate a beer, but also how to make a beer that’s better tasting.
This wasn’t an overnight process. Over a five-year period, researchers analyzed the chemical properties and flavor compounds in 250 commercial beers.
The researchers then combined these detailed analyses with a trained tasting panel’s assessments of the beers—including hop, yeast, and malt flavors—and 180,000 reviews of the same beers taken from the popular online platform RateBeer, sampling scores for the beers’ taste, appearance, aroma, and overall quality.
This large data set, which links chemical data with sensory features, was used to train 10 machine-learning models to accurately predict a beer’s taste, smell, and mouthfeel and how likely a consumer was to rate it highly.
The result? When it came to predicting how the RateBeer reviewers had rated a beer, the AI models actually worked better than trained tasting experts. Further, the models enable the researchers “to pinpoint specific compounds that contribute to consumer appreciation of a beer: people were more likely to rate a beer highly if it contained these specific compounds. For example, the models predicted that adding lactic acid, which is present in tart-tasting sour beers, could improve other kinds of beers by making them taste fresher.”
Admittedly, having lactic acid in a beer doesn’t sound all that appealing. But if the beer tastes fresher, well, just don’t read the fine print on the ingredients list.
One area where they anticipate the AI approach will prove particularly effective is in the development of non-alcoholic beers that taste as good as the real thing. This will be great news for those who want to enjoy a beer without having to consume any alcohol.
There are other instances of AI being used in brewing. Way back in 2016, a UK AI software startup IntelligentX, came out with four beers based on their Automated Brewing Intelligence algorithm. The release of Amber AI, Black AI, Golden AI, and Pale AI caused a brief flurry of excitement as the first AI developed beer. Unfortunately, it looks like none of them made much of an impact in the beer market. When I searched for them, I couldn’t find any references beyond 2019.
Maybe the models that the Belgian researchers produced will have more luck creating a successful AI beer.
——————————————————————————————
The full research report from Nature Communications can be found here.