This winter took an oddly futurist bent.

While I was developing classes and material, Nora Young’s Spark got in touch from the CBC to ask about AI in beer. I went away and diligently researched what people were doing around the world. There are some good ideas out there, but it turned out AI didn’t mean quite what I thought it did; I played a lot of Deus Ex on PC when it came out. Judging from the research I put in, we’re a significant distance from the singularity or some nanoangstroms of hate. In a lot of cases, the machine learning that could be done in the podcast I’ve linked could be done with a card catalogue or a relational database, but it would take a hell of a lot longer. It’s just a question of comparative data sets and algorithms.

Towards the end of the interview, they asked me whether I knew of any AI assisted beer companies in Canada that were operational. I had heard a couple of rumblings about some University of Toronto based initiatives, but nothing concrete.

Sure enough, about four days after the interview for the podcast, an AI company got in touch with me to ask if I’d be interested in working with them. I like data sets and algorithms, so I said yes.

TasteGuru is an interesting concept. As they’re describing it on their website, it’s essentially an omni-facing recommendation platform for food and beverage which customizes recommendations based on your preferences. At the moment it’s a startup and they have started with beer. The first set of data and algorithms are just going into beta testing at the moment, so you now have the ability to click over to the website and try out their assessment questions.

I know better than anyone (having written and formatted two editions of the Ontario Craft Beer Guide), that the amount of information that exists in the world of beer has expanded exponentially in the last decade. There are probably ~360 brewing entities in Ontario at the moment. They might make 4000 beers including one-offs this year. The number of potential ingredients has also expanded dramatically. Yesterday I had a Pina Colada Smoothie IPA. That did not used to be a thing.

Your guides to this market complexity are somewhat problematic since taste is subjective. Consider the popular rating program Untappd. On a long enough timeline on Untappd, everything is going to revert to a score of 3.56/5. Beer, then, if taken as an overall construct, is only a 3.56/5. Because the data input is subjective and based entirely on personal preference, it’s not very good data. Some people will give a Pliny the Elder a zero. Some people will give Coors Light a five. That might be true for them, but it’s not that helpful to a novice consumer.

What TasteGuru has done is attempt to use trained sensory experts to create data that is representative of the beer without taking branding, style, or personal preference into account. Based on a series of carefully chosen metrics, they’re attempting to align your flavour preferences with existing products in order to find something you’re going to enjoy and eventually to tell you where you can find it. It’s an ideal solution for people new to beer as a beverage and find the enormity of the selection a little daunting. It’s an exciting prospect that puts all products on an equal footing regardless of the size of the company making them.

In the long run, the data set will be a great deal larger than it is currently and it may be possible to tailor an evolving profile to the consumer. For the time being, they are just looking for feedback and testers for their initial foray into a consumer-facing portal.

Have a look here, and don’t be afraid to get in touch and provide feedback to them.