Hit enter to search or ESC to close
Entertainment

How Artificial Intelligence Learned to Brew Beer

A can of AI-generated IPA beer pictured in a black can against a background of 1s and 0s.

Brewing beer is one of humanity’s oldest arts. So how would a new-fangled technology like artificial intelligence do at the task?

That was the question facing Denham D’Silva, founder and owner of Barossa Valley Brewing in Adelaide, Australia, after a chance encounter with Simon Lucey, a professor in the School of Computer Science at the University of Adelaide.

The two men met at a tech incubator called Stone and Chalk and hit it off immediately, thinking, “Wouldn’t it be cool to take one of the oldest industries, brewing, and one of the newest industries, AI, and find some synergy there?,” Lucey said.

After all, D’Silva is an expert at brewing and Lucey is an expert in AI, serving as the Director of the Australian Institute of Machine Learning – at the center of that “newest industry” – and principal scientist at Argo AI, a Pittsburgh-based autonomous vehicle products and services company.

“We wanted to see how far we could go [with AI.] It was a very exciting result. To use AI to go through the recipes and apply it to this space, we were in rarified air,” Lucey said.

About those recipes: to build the substantial dataset needed to make the craft beer, interns from the machine learning institute used a staggering 260,000 craft beer recipes. From these, they built a neural network that learned from the other recipes how to make the new brew.

“It’s unique and a great example of how technology can allow people to do more”

And it worked! Using the existing pale ale recipes and D’Silva’s and his team’s feedback, the neural network yielded a beer that tastes like any of the brewery’s libations created the old-school way.

“We didn’t say, ‘This needs to taste like our beer.’ But in the process of creating the neural network, there are decisions we make, so in doing that, it has your character in that beer, “ D’Silva said.

Still, he had his doubts about the project. He started thinking about how data-driven and emotionless the AI is, at odds with his sense that art goes into making a beer.

“I wasn’t sure whether the technology was going to spit out something soulless and heartless, that didn’t taste like something I was happy with,” D’Silva said.

D’Silva was concerned that if a machine tried to make what it perceived as the “best” beer, it would think best-selling, not best-tasting. He worried that what should taste like a brew with character could taste like… a popular American-style pale lager, not to name names.

D’Silva had reason to be worried that the beer wouldn’t be well-received. Years ago, he was among the first craft brewers in Australia and experienced the difficulty of educating the market. But he was pleasantly surprised at sales, seeing that the market had progressed more than he anticipated.

D’Silva named the ale Rodney AI2PA, in honor of Australian robotics pioneer Rodney Brooks, whose company iRobot developed the Roomba vacuum. The beer found an audience quickly on tap and in cans in dozens of venues across Australia. It was so popular that most places sold out faster than anticipated.

“There are a few that say it’s heartless and soulless. But 95 percent of what I’m hearing is positive,” D’Silva said.

The Human Touch

In fact, the AI kept humans at the center of the process.

“You need the human element. ‘This is too hoppy. How can I move it this way?’” Lucey said. “The master brewer needs to go through it. It’s a unique and great example of how technology can enhance productivity and products.”

Typically, international conglomerate brewers use this kind of technology — not smaller breweries like D’Silva’s. Still, Lucey believes the success of their collaboration is an example of how AI can enhance efficiency at small businesses while creating jobs.

“Small business owners are the lifeblood,” said Lucey. “This showcases how AI can help their businesses, and I think that’s a really pleasing thing. It’s unique and a great example of how technology can allow people to do more, and that’s really important to creating jobs and raising standards of living.”

Lucey and D’Silva said AI innovation in the food and beverage industry is obvious to them, and that producers in nearly any industry that relies on a recipe with various combinations could benefit. Think wine, cheese, whiskey and even medicine and vaccines.

Asked whether this is the tip of the iceberg of AI applications, D’Silva responded, “Before, I thought that. Now, I know that.”

As a craft beer producer, D’Silva knows his industry is powerfully consolidated internationally, and is using machine learning for all kinds of applications in marketing, process and logistics. But he sees AI as a way for small businesses to get a competitive edge, letting them make new products in just weeks, instead of the months or years it takes less-nimble conglomerates.

“The big guys use that to take the information to try to convince you that this is the product you want to buy. Smaller producers have the ability to take that and change their product. So, it makes smaller producers much more competitive,” D’Silva said.

Does AI-Assisted Brewing Have a Place in the U.S.?

Barossa Valley may be in rare air, but it’s not flying solo. Deschutes Brewery in Bend, Oregon, and Kensington Brewing Company, a Toronto brewery that debuted its A Little Robotic IPA in 2021, are other houses that have used AI to brew beer.

Derek Lintern, brewer at NOLA Brewing Company in New Orleans, is about to join that club.

D’Silva reached out to Lintern for assistance in getting his AI beer brewed and made available at the conference on Computer Vision and Pattern Recognition (CVPR) in New Orleans this summer. Lintern, who studied computer science in college and was a professional programmer before getting into brewing, obliged.

Lintern is doing a 20-barrel batch that will be distributed at the huge AI conference. It will be D’Silva’s new recipe for a saison-style beer.

“We’re doing something that’s been done the same way since the beginning of civilization. Why not try something new?” Lintern said. He referenced his brewery’s very traditional set-up. “Here at NOLA, the system is about as manual as it can be. We have a digital thermometer but that’s it. It will be really interesting to take something that’s state-of-the-art and bring it to a 30-year-old manual brewing system.”

Lintern sees AI-assisted brewing as a fun phase that could have longevity in the right applications, like facilitating initial recipe development and tweaking with human feedback.

“I can see that as being a positive,” Lintern said.

Paul Gatza, Senior Vice President of the Professional Brewing Division at the Brewers Association, sees pros and cons. In his current position, Gatza assists craft brewers with their recipes and troubleshoots them.

For Gatza, the thought of using AI to craft beer isn’t as interesting as using it in other areas, such as figuring out how to optimize the malting process or, on the distribution side, making ordering and routing less problematic. There’s also a one-size-fits-all conundrum.

“Beers can be made technically great. But what works in one brewery may not work in another,” Gatza said.

But he also liked the idea of incorporating AI into the evaluation process to determine differences from batch to batch. In essence, a more efficient brewer’s log. Gatza sees this catching fire not within the next five years, but perhaps after that as more American breweries learn about it.

“Brewers have the DNA to be open minded enough to say, ‘Hey, let’s give it a try,’” Gatza said. “It only adds to the interest and fun.”

Must Reads