The next pint you pour may tickle your tastebuds thanks to technical tinkering by artificial intelligence. It may even be the next big thing in the food and beverage space—certainly an easier sell than vaginal beer and artificial steak.
Already an acquired taste for particular palates, the taste of beer could be made better with the help of AI. A group of scientists went well beyond asking a chatbot for some random soda flavor in a fancy presentation. By combining extensive chemical analyses with AI models trained on sensory data, researchers have identified specific compounds that can boost the taste of your favorite brews.
"Our study reveals how big data and machine learning can uncover complex links between food chemistry, flavor, and consumer perception," the research paper published today in the journal Nature Communications says. "This paves the way for developing novel tailored food products with superior, crowd-pleasing flavors using a data-driven approach rather than just trial-and-error."
The research, conducted by an international team of over a dozen experts, analyzed over 250 commercial Belgian beers spanning diverse styles. Each beer underwent rigorous chemical profiling to measure over 200 different properties like esters, alcohols, acids and aroma compounds.
This chemical fingerprint was then matched against quantitative sensory evaluations from a trained panel of tasters scoring 50 distinct flavor attributes as well as over 180,000 online reviews from beer enthusiasts rating aroma, taste and overall appreciation.
Leveraging this massive dataset, the researchers next trained and tested various machine learning algorithms to parse the complex relationships between a beer's chemical makeup and its perceived flavors and consumer appeal. One common algorithm emerged as the top performer, significantly outperforming conventional statistical methods.
"The best-performing algorithm, gradient boosting, yields models that significantly outperform predictions based on conventional statistics and accurately predict complex food features and consumer appreciation from chemical profiles," the research reads. This method, researchers say, “allows identifying specific and unexpected compounds as drivers of beer flavor and appreciation."
Gradient boosting is a machine learning technique that combines multiple small, inaccurate models into a larger one that is capable of making good predictions. It’s widely used in banking and healthcare as well as for marketing campaigns that try to predict the outcomes of specific endeavors.
Crafting good beer was not likely its main purpose, but gradient boosting outperformed other prediction models tested by the research team.
The researchers found that “flavor compound concentration does not always correlate with perception,” meaning that consumers tend to appreciate more nuances beyond just flavor when enjoying a good beer. This is hard to predict with conventional models and simple analyses, and it is one of the reasons why the team considered such a large number of variables.
Through interpretable machine learning techniques, the scientists uncovered some unexpected compounds that strongly influence the quality of a beer. For example, methanethiol and ethyl phenyl acetate are typically associated with staleness, but can make a beer taste good in small doses.
More familiar flavor drivers like ethyl acetate and lactic acid were also examined.
To validate their models, the researchers conducted tasting experiments in which they spiked poorly rated beers with the machine learning-identified compounds that increased appreciation. This simple adjustment led to significant increases in trained panelists' flavor scores and overall preference ratings for the modified beers compared to the originals.
The brewers and beer lovers of the world may soon have to thank artificial intelligence for enhancing our enjoyment of one of humanity's oldest and most beloved beverages. But the applications extend far beyond better beer. Similar AI-powered chemical profiling could optimize the flavors of everything from plant-based meat alternatives to low-sugar sodas and snacks.
"The present work demonstrates that despite some important remaining hurdles, combining the latest developments in chemical analyses, sensory analysis, and modern machine learning methods offers exciting avenues for food chemistry and engineering," the researchers argue. “Soon, these tools may provide solutions in quality control and recipe development, as well as new approaches to sensory science and flavor research."
The next time you crack open a perfectly balanced, delectable beer, it may disprove the old adage that 'artificial' is inferior and give pause to raise a glass to AI.
Edited by Ryan Ozawa.