Artificial intelligence has emerged as one of the most relevant technological, social, and economic phenomena of the decade. It’s also become one of the most-feared advancements—viewed as a potential threat to humans’ job security. It would seem as though it has taken over just about everything—from kids’ homework and customer service, to self-driving cars and military intelligence-gathering. No industry has been too far from its reach, including craft beer.
It shouldn’t come as a surprise that, like many companies and entrepreneurs across the country, craft brewers are looking at how AI might improve efficiency and the overall brewing experience. One of the prime examples is Sugar Creek Brewing Company in Charlotte, N.C., which uses AI and a network of devices to monitor the quality of its packaged beer, including temperature, fill level, and foam level.
However, beyond the standpoint of business and operations, brewers and other industry professionals are also experimenting with the tech to test its creativity. They’re seeing if AI can push the envelope on recipe development and flavor.
Craft breweries in different parts of the country have already concocted AI-generated beers. This includes Asbury Park Brewery in Asbury Park, N.J. and Atwater Brewing in Detroit, Mich., each with their own AI-IPAs, and Urban Forest Craft Brewing in Rockford, Ill. which released a brew called Put the AI in sAIson.
Michael Oxton, co-founder of Night Shift Brewing in Boston, used OpenAI’s industry-leading large language model ChatGPT to make a hazy IPA. “It was fully capable of putting together a recipe with instructions and a water profile and suggested hops,” says Oxton. He and his team were particularly impressed by AI’s ability to come up with a target water profile within minutes, saying, “It might take us a few weeks or even a month to come up with something like that based on a lot of research and testing.”
Night Shift’s head brewer drew up a batch resulting in what they called AI-PA—a beer that first hits you with a burst of mango and watermelon, and ends with citrusy lemon-lime notes. Oxton summed up, “It was delicious.”
Brewing Outside the Box
Instead of using ChatGPT or other commercially available models, Species X Beer Project took things a step further. For Species X, founder Beau Warren had two different “species” of beer: the AI-created Silicon Species, and his own human-made recipes for the Carbon Species. Warren—who has studied both beer microbiology and data analytics—programmed seven different AI models himself for the Silicon Species.
He wrote code to build each model using a combination of two types of machine-learning techniques: regression models and neural networks. Each of the models he coded was then trained using proprietary data and information on different styles of beer (including water chemistry, yeast, hops, flavor profiles, ABV, human-made recipes that have worked in the past, etc.). He also used data from his Carbon Species, including customer and internal reviews, to teach his AI models what the parameters for a highly rated beer are. And Warren created models to focus on specific styles of beer—one for IPAs and others for lagers, fruited sours, smoothie beers, low-alcohol beers, and even hard seltzers.
“Some of these recipes are so out of left field, I cannot imagine myself sitting down and creating one of these recipes by myself,” says Warren. “It throws tradition out the window entirely.” Of course, the models have guardrails coded into them (except the Beh3moth model, which is trained on all types of beers with minimal parameters), mainly to make sure they can brew the beers within the laws of physics and not break any of their equipment.
One of the wildest brews Warren has ever worked on was an amber lager made with 100% Maris Otter malt, which is typically reserved for bitters, porters, and stouts for a rich, malty flavor. In the end, AI didn’t steer him wrong and he came out with a well-balanced lager that turned out to be a hit.
Beh3moth also created an imperial pastry Baltic porter using a plethora of ingredients. The recipe involved three different types of fruit, coffee, vanilla, 12 different types of malts, lactose, and marshmallows. Then the batch was dry-hopped like a New England IPA. “Absolutely insane, but tasted amazing and balanced,” summed up Warren. “Literally unlike anything I have ever had in my life.”
Each of Warren’s AI-generated creations is unique, clearly built from the ground up, down to the water profile. He says, “It’s not like we’re using the same base recipes, swapping out hops, or we’re doing a fruited sour and swapping out fruit. These things are completely brand new, beer to beer, and we can do it quick.”
Moreover, it’s not just craft breweries that are expanding the creative possibilities of craft beer using AI. Colorado-based yeast propagation facility Propagate Lab has given brewers the opportunity to come up with creations that go against the grain with their AI chatbot “Yeast Buddy.”
Yeast Buddy helps clients filter through a database of more than 120 yeast strains and narrow it down to the one that best suits the beer style, flavor profile, and other parameters for their specific brew. It’s kind of like an automated library catalog, making sure brewers don’t need to go through the tedious Dewey Decimal System.
As a result, brewers have easier access and are able to discover the diversity of yeasts at Propagate Lab, allowing them to break away from defaulting to the “standard.” Instead of going for a strain that a friend used or choosing something familiar, they have a chance to be more adventurous.
“With beer, there are so many different ingredients between hundreds of hop varietals, hundreds of malt choices, and several hundred yeast choices,” explains Matthew Peetz, founder of Propagate Lab. “[Yeast Buddy’s] giving them more options in terms of putting together a recipe, rather than being pigeonholed with what you used before in your batches of beer.”
Is This a Taproom Takeover?
With the conversation on machine learning comes the looming thought that AI might take over industries and jobs. If robots can be baristas, what’s stopping them from infiltrating the taproom? After all, AI only gets smarter by the minute. However, the truth of the matter is that it’s not yet poised to replace humans.
“It was like having a brewing assistant right by your side that could come up with ideas,” describes Oxton of his experience using AI. Admittedly, while the original AI-PA recipe fresh from ChatGPT was a hit, Night Shift had to make a few human adjustments in order to work with ingredients and materials they had available, for practicality and preference. And while creating the AI-generated art for AI-PA’s label, Night Shift’s designer had to make some modifications to fix some kinks and get it ready to display on a can.
“I think, if anything, it was like a creativity enhancement more than a replacement,” says Oxton. He adds that an AI assistant might be more useful to a newbie still trying to learn the ropes, as opposed to a head brewer with over 10 years of experience.
There is no denying that AI models are wildly impressive, especially with Species X’s being particularly powerful and accurate in terms of hitting the mark on flavor profiles. But Warren emphasizes that a human still needs to be in the loop. “AI needs humans to, first of all, put guardrails on it,” he says. “And then second of all, to introduce new training data. Otherwise, you’re going to be stuck with a non-updated model forever, essentially.”
AI still needs someone to pull all the strings, feed it the correct information, and vet the results it yields in order to make sure they are sound. Peetz echoes this sentiment, saying, “It’s as useful as a calculator. If you are punching in the wrong numbers, you’re going to get the wrong result.”
Taking a Step Back
For now, though AI is readily available, the craft beer industry has nothing to fear. The full picture of this technology’s future as it applies to beer is still uncertain. (At the time of publishing, Species X had permanently closed its doors after two years of operation, due to economical constraints.) However, what is clear is AI’s potential not just to streamline business systems, but to help brewers knock down creative barriers.
Warren says, “It makes humans superhumans, essentially, because of how powerful it is, as long as it’s used in the right way.”
CraftBeer.com is fully dedicated to small and independent U.S. breweries. We are published by the Brewers Association, the not-for-profit trade group dedicated to promoting and protecting America’s small and independent craft brewers. Stories and opinions shared on CraftBeer.com do not imply endorsement by or positions taken by the Brewers Association or its members.
