Beer brewers across the globe are turning to AI to create new ales
来源：Josh Sims SCMP | 发布时间：2025-07-31
摘要：啤酒厂正借助人工智能来提升啤酒的口感。但像 ChatGPT 这类人工智能，真能酿出值得一喝的一品脱好酒吗？ Breweries are tapping artificial intelligence to make beer taste better. But can the likes of ChatGPT create a pint worth drinking?
Back in 2023, just after ChatGPT was launched to the public, Raoul Masangcay, co-founder of Elias Wicked Ales & Spirits in Manila, the Philippines, thought an experiment was in order. He asked an artificial intelligence (AI) chatbot to devise a hazy pale ale, which his brewery then made.
“People thought the idea was cool even if the beer was fairly standard and not so distinctive,” says Masangcay. “We even used AI to name the beer [Foggy Daze] and generate the [promotional] artwork. Clearly, AI can be a tool in the brewing industry as it is in others.”
Masangcay, who also works as an engineer for Intel, isn’t the only craft brewer to identify AI’s potential. Last year, Britain’s St Austell Brewery launched an IPA wittily named Hand Brewed by Robots, while in the United States, Asbury Park Brewery used AI to help create a new India pale ale, which it called AI-IPA.
Japan’s Coedo Brewery teamed up with IT company NEC this year to use AI to develop four craft beers based on the characteristics and preferences of people in their 20s, 30s, 40s and 50s.
AI is appealing to the big guns of brewing, too. While Beck’s has its Autonomous beer, which the beer giant says was created by AI, Heineken has recently opened its global GenAI lab in Singapore. It might not yet be involved in recipe creation, but the plan is for the lab to boost productivity by looking into how a particular segment of the population might feel about a particular kind of beer.
Indeed, AI’s potential to make better beer is only just starting to be explored. At KU Leuven, a university in Belgium, Kevin Verstrepen is conducting research into the flavour compounds found in 500 Belgian beers, scoring them against a trained tasting panel and 180,000 online reviews, allowing him to use the data to build a machine-learning model that predicts how a beer would taste based just on its composition. AI could then be used to make precise but critical improvements to recipes. He is now extending that research to wine and spirits.
“From the chemistry alone we can’t know how a beer will taste precisely because the interaction of aroma compounds is so complex, but machine learning can help us connect the dots better and make predictions about the taste, and whether someone will like it more or less,” says Verstrepen, who has worked with a number of Chinese breweries, as well as baijiu distillers.
“The model will never be perfect because taste is individual. There’s no ideal beer for everyone. But the model could be used to tweak beer recipes to make them appeal to more people, or certain people, and breweries have asked us to do that.”
But, according to Verstrepen, since alcohol compounds are so crucial to the overall quality of beer, the big breakthrough in AI would be to more closely mimic the aroma of alcohol to improve the experience of drinking non-alcoholic beers. His research on this subject is due to be published later this year.
In South Australia, Denham D’Silva, founder of Barossa Valley Brewing, has AI to thank in part for his company’s success, having joined forces with the University of Adelaide’s Australian Institute of Machine Learning to help found Deep Liquid, a consultancy that uses generative AI to provide custom-made beers and, more lately, health drinks.
AI’s real power is “in its ability to efficiently and quickly process inputs and feedback from everyone, especially now when people love jumping on social media to express their thoughts on products”, says D’Silva.
“It’s not AI taking over the development of the beer but listening to customers and distilling the insights into that beer.” D’Silva also believes that since smaller brewers tend to be more nimble, responding to those insights could give them an edge over the global giants.
“It’s still the brewer that has to decide on the right recipe in the end”
Denham D’Silva
While there has been much talk about AI replacing certain jobs, it is unlikely to spell the end of the brewer’s skill, prized as that has been for millennia. “The brewer can only consider so many variables, unlike AI,” says D’Silva. “But it’s still the brewer that has to decide on the right recipe in the end.”
Bob McLynn, co-founder of Asbury Park Brewery, agrees. “Our brewmaster was pleasantly surprised by AI’s recipe suggestions, but the process was always a collaboration. Ultimately you’re working with real ingredients, so the flavour of the beer, and the human intervention to refine that taste, will always be critical.”
Yet Verstrepen stresses that however much the beer marketers like to push a more traditional image, “more horse-drawn carts, not so many white laboratories”, the truth is that, historically, the industry has always leaned more towards an experimental and scientific approach, “so it would be silly for brewers not to use the latest tech now”.
世界各地的啤酒酿造商正在转向人工智能来创造新的啤酒
早在2023年，就在ChatGPT向公众推出之后，菲律宾马尼拉Elias Wicked Ales & Spirits的联合创始人Raoul Masangcay认为实验是有序的。他让一个人工智能（AI）聊天机器人设计出一种朦胧的淡色啤酒，然后他的啤酒厂制造了这种啤酒。
Masangcay说：“人们认为这个想法很酷，即使啤酒相当标准，没有那么独特。”“我们甚至使用人工智能给啤酒命名[Foggy Daze]，并生成[促销]艺术品。显然，人工智能可以成为酿酒行业的工具，就像在其他行业一样。”
Masangcay也是英特尔的工程师，他并不是唯一能识别人工智能潜力的精酿啤酒商。去年，英国的St Austell啤酒厂推出了一种名为Hand Brewed by Robots的IPA，而在美国，Asbury Park Brewery使用人工智能帮助创造了一种新的印度淡色啤酒，它称之为AI-IPA。
日本Coedo啤酒厂今年与IT公司NEC合作，利用人工智能根据20多岁、30多岁、40多岁和50多岁的人的特点和偏好开发了四种精酿啤酒。
人工智能也吸引了大炮的酿造者。虽然Beck’s拥有Autonomous啤酒，这家啤酒巨头称这是由人工智能创造的，但喜力啤酒最近在新加坡开设了其全球GenAI实验室。它可能还没有参与食谱的创建，但计划是让实验室通过研究特定人群对特定啤酒的感受来提高生产力。
事实上，人工智能制造更好啤酒的潜力才刚刚开始被探索。在比利时的一所大学鲁汶大学，Kevin Verstrepen正在研究500种比利时啤酒中发现的风味化合物，根据训练有素的品尝小组和18万条在线评论对其进行评分，允许他使用数据构建一个机器学习模型，仅根据啤酒的成分预测啤酒的味道。然后，人工智能可用于对食谱进行精确但关键的改进。他现在正在将这项研究扩展到葡萄酒和烈酒。
Verstrepen曾与许多中国啤酒厂以及白酒蒸馏厂合作过，他说：“仅从化学学上，我们无法确切地知道啤酒的味道如何，因为香气化合物的相互作用是如此复杂，但机器学习可以帮助我们更好地连接点，并预测味道，以及是否有人会或多或少地喜欢它。”
“模型永远不会完美，因为品味是个人的。没有适合每个人的啤酒。但该模型可用于调整啤酒配方，使其吸引更多人或某些人，啤酒厂已经要求我们这样做。”
但是，根据Verstrepen的说法，由于酒精化合物对啤酒的整体质量至关重要，人工智能的重大突破是更紧密地模仿酒精的香气，以改善饮用无酒精啤酒的体验。他关于这个主题的研究将于今年晚些时候发表。
在南澳大利亚州，Barossa Valley Brewing的创始人Denham D’Silva有人工智能，部分要感谢他公司的成功，他与阿德莱德大学的澳大利亚机器学习研究所联手，帮助创立了Deep Liquid，这是一家使用生成人工智能提供定制啤酒和最近提供健康饮料的咨询公司。
D’Silva说，人工智能的真正力量在于“它能够高效、快速地处理每个人的投入和反馈，尤其是现在人们喜欢跳上社交媒体来表达他们对产品的想法。”
“不是人工智能接管啤酒的开发，而是倾听客户的意见，并提炼出对啤酒的见解。”D’Silva还认为，由于小型酿酒商往往更敏捷，对这些见解做出回应可以让他们比全球巨头更有优势。
“最终还是酿酒师必须决定正确的配方”
德纳姆·德西尔瓦
虽然有很多关于人工智能取代某些工作的讨论，但这不太可能意味着酿酒师技能的终结，几千年来一直受到重视。D’Silva说：“与人工智能不同，酿酒师只能考虑这么多变量。”“但最终还是酿酒师必须决定正确的配方。”
Asbury Park Brewery的联合创始人Bob McLynn表示同意。“我们的酿酒师对人工智能的食谱建议感到惊喜，但这个过程总是合作的。归根结底，你正在使用真正的原料，所以啤酒的味道，以及人类干预来提炼这种味道，将永远至关重要。”
然而，Verstrepen强调，无论啤酒营销商多么喜欢推动更传统的形象，“更多的马车，而不是那么多的白人实验室”，事实是，从历史上看，该行业一直更倾向于实验和科学的方法，“所以酿酒商现在不使用最新技术是愚蠢的”。