ALGOMA – AI has slightly oaky, spicy tones, dried fruit flavors and a silky mouthfeel. AI is light-bodied with floral aromas and fruit and citrus flavors.
That’s possible at Von Stuhl Winery, which is using AI to determine the taste and aroma of groundbreaking new grape wines.
Algoma Winery’s Artificial Intelligence White V2.0 and Artificial Intelligence Red V2.0 are blended wines that use AI to shape the recipe of the blend based on consumer feedback.
AI is increasingly being used in growing and harvesting wine grapes, but to Von Stiehl’s general manager Anthony Bilwen, no one else has used it to essentially create wine with flavors crowdsourced directly from customers. It’s a big, innovative step for Wisconsin’s oldest licensed winery, but Von Stiehl only does it when the opportunity arises, Bilwen said.
“At von Stiehl, we’ve always strived to be on the cutting edge of technology,” says Birwen. “Trying new things is in our DNA.”
Here’s how it works:
AI is already being used in vineyards, but it’s new to winemaking
Artificial intelligence is rapidly making its way into the wine industry, as numerous articles in wine trade publications over the past few years have attested.
But AI is mostly used in vineyards, where weather and growing data can help determine the need for irrigation, detect disease and predict the best time to harvest the grapes, and in winery cellars to help winemakers monitor the fermentation process.
But AI also directly influences wine drinkers’ palates.
According to an article in Decanter magazine, French winemakers Aubert & Mathieu entered information about their location and the grape varieties they had available into ChatGPT, a popular AI chatbot system, and asked it to create a red wine based on that information. ChatGPT responded with basic winemaking information and suggestions for the bottle, label, marketing, and name (“The End”), as well as a general wine recipe.
And according to an April article on the BBC.com website, California technology company Tastry is developing an app that can analyze a wine’s sugar, acidity, tannins and other chemical components and compare them to the tastes of some 248 million American wine drinkers. The idea is to guide winemakers, especially those producing for the masses and aiming for greater consistency, to more widely acceptable wine blends.
But as Birwenn says, von Stiel’s AI V2.0 wines are the first to use direct consumer feedback to determine the wine recipe, and an online search didn’t reveal any other wineries using von Stiel’s method.
It all started with taste and QR codes
Von Stier produced and sold two AI V1.0 wines last summer, a red wine blend and a white wine blend, and has begun the process of AI V2.0 wines.
Each V1.0 wine bottle has a QR code on its label that can be scanned by anyone who has the bottle and tastes the wine. Once scanned, the wine drinker is directed to a survey where they can fill out a tasting review of what they liked and didn’t like about the wine’s taste, aroma, mouthfeel, and even the bottle’s label.
These responses were sent to the popular AI chatbot systems ChatGPT and Google Gemini, which analyzed all the data. Von Stier’s winemakers then asked the chatbot for a paragraph summarizing the critiques of the V1.0 blend, as well as suggestions on how to best modify the recipe to meet those critiques.
In other words, based on the V1.0 wines, consumers primarily determined the taste, aroma and feel of this year’s AI white and red wines V2.0, which is what von Stier was aiming for.
“We collected all the data on customer feedback: is the wine too dry, too fruity, too floral? What should we do to change that,” Birwen said. “We’ve always valued customer feedback, but with the introduction of AI-driven wine, we now have the ability for our customers to directly influence the winemaking process like never before.”
But AI won’t take away the craftsmanship of winemaking.
According to Birwen, winemakers weren’t given specific recipes by the ChatGPT and Gemini report. Instead, the report offered suggestions like “reduce the sweetness” and “balance the acidity” and offered ideas on how to execute them, including specific grape varieties that could be added to the blend to achieve that goal.
It was then up to the winemakers to figure out how to solve the problems created by those suggestions, which means that even though it’s an AI-driven wine made from a crowdsourced recipe, it still relies on at least a little bit of old-fashioned winemaking techniques to come to fruition.
“The technology isn’t yet smart enough to come up with a foolproof recipe for blending wines,” Birwen says. “It would be different if the recipes were out there, but most winemakers don’t and won’t publish their recipes.”
This winemaking method may one day be the future of the business, but as of 2024, of the 40-plus wines and ciders von Stiel produces, the only one made this way is AI Red and White V2.0.
“We don’t want to do this with all our wines,” says Birwen, “because if we did, we would lose the artisanal aspect.”
So what changes have been made from AI V1.0 to V2.0? And how does V2.0 taste?
According to von Stier’s website, ChatGPT and Gemini sent similar summaries back to von Stier about AI Red V1.0. Both noted that consumer opinions varied considerably regarding taste, flavor, mouthfeel, aftertaste, and even the label. Consumers clearly preferred the deep red color and quickly noticed the dark fruit flavors and oak tannins in the aftertaste, but seemed divided on whether the last two were positive or negative.
In general, the chatbot recommended creating a semi-dried wine by reducing sweetness while still maintaining its fruity character, increasing body and depth, and improving the acidity balance, and it also suggested several grape varieties that could achieve these goals.
AI White V1.0 has also received mixed reviews, as noted in the chatbot summary, with consumers generally favoring the citrus and pear aromas, but split on whether the taste and aftertaste is too sweet. As with red wine, white wines were recommended to be less sweet and have a balanced acidity, with ChatGPT again recommending the addition of certain grapes to the blend.
ChatGPT also proposed adding more information about the AI wine initiative to the bottle labels of both wines, perhaps using another QR code.
As a result, AI Red V2.0 is “less sweet and has more oaky, slightly spicy characteristics” than V1.0, while AI White V2.0 retains the lightness of V1.0 but is slightly drier and more acidic, according to the von Stier website.
Where can I find AI V2.0 wines?
Of course, samples of the AI wines will be available in Von Stier’s tasting room, but wine drinkers will have to purchase a bottle, scan a QR code, and submit a review for consideration in future wine recipes.
Bottles are available for purchase through the winery and its website, and Bilwen said they will be available in liquor and grocery stores that carry Von Stiel, such as Festival Foods, by the end of June.
Hold on, did you say “considering future wine recipes”?
Yes, this will become an annual event.
According to Vilwen, the winery plans to create a new AI-generated blend every year, and people who buy the V2.0 wine can scan a QR code on the label to give their opinion about the wine, which will be collected in 2024 and compiled by the AI to be used by the winemaker for the V3.0 wine to be released next year.
For more information
The von Stiehl winery, along with its cider mill and distillery, is located at 115 Navarino St. in Algoma. For more information, call 920-487-5208 or visit vonstiehl.com.
CContact Christopher Clough at 920-562-8900 or cclough@gannett.com.
more: Door County marina that has been in business for nearly 60 years is being sold to a Texas-based company
more:Door County marine construction company gets new owners and moves to new, larger location
For more Door County news: Visit our website
