The term “machine learning” covers a grab bag of algorithms, techniques, and technology that are by now pretty much everywhere in modern life. However, machine intelligence has recently started to be used not just for identifying problems but to build better products. Amongst the first is the world’s only beers brewed with the help of machine intelligence, which went on sale a few weeks ago.
The machine learning algorithms uses a combination of reinforcement learning and bayesian optimisation to assist the brewer in deciding how to change the recipe of the beer, with the algorithms learning from experience and customer feedback.
Perhaps the most obvious intrusion of machine learning into the physical world is the voice recognition that drives Apple’s Siri, or Amazon’s Alexa. However, every time you search using Google, or take Netflix’s advice on what to watch this evening, you’re using it. Machine learning has proved itself useful in quickly identifying patterns that humans would overlook, or just be unable to find at all, in the vast amounts of data that we generate every day. Traditionally though, it does poorly when confronted with smaller amounts of data.
The machine learning approach used to brew the new beers is a Bayesian technique, in the form of Bayesian non-parametrics. This class of algorithm is efficient with data, but also very good at handling uncertainty.
Image: Alasdair Allan
“The critical problem you have when dealing with small amounts of data is that it generates uncertainty in areas where you have little experience. Many machine learning models don’t deal with uncertainty very well—instead they rely on a lot of data in order to be able to generalise and form good predictions,” says Rob McInerney, co-founder of IntelligentX.
He went on to say that, “The beauty of the real world is that the surface we are optimising over is very complex and has many hills and troughs. The reality is it’s very hard to know if you’ve hit a local minimum or whether something slightly better is just around the corner. I feel very strongly that a lot of data science doesn’t really accept this and so pushes for trying to find some sort of ‘perfect setting’ in massive parameter spaces.”
As the Internet of Things becomes more pervasive the amount of data we leave in our wake will multiply and, as the intrusion of machine learning into our lives more pervasive and more obvious, it’s likely the world around us starts to customise itself to us without us asking— With a thin layer of computing, of machine intelligence, standing between us and the world. In the same way augmented reality technologies will sit between our own vision of the world, mixing the simulated and the real world, machine intelligence will sit between our actions and the world.
Image: Alasdair Allan
“The brewer’s interaction with the algorithm is very much part of the technology and generates additional learning data—we like to think of the algorithm as being a bit like an apprentice, listening and learning from a master,” says McInerney, “However, unlike a human apprentice the tech can listen to multiple different master brewers as well as all the customers at the same time.”
Of the four beers beers released into the wild, I got to try three. The two I liked best were the Golden and the Pale Ale. The Golden is frothy, with a sharp taste, whilst the Pale has a taste almost that of sour apples. “The beer has taken a number of directions we didn’t anticipate. For example, [the algorithm] has a bank of wildcard ingredients, like adding fruit to a recipe, in a bid to create beer that pushes the boundaries of what’s possible within craft brewing. That led to one of the beers having a hint of grapefruit added to it,” says McInerney. The beers taste almost, but not entirely, unlike what you’d expect. I rather enjoyed them.
“What’s different for our use case is that we are seeing how this can help with creativity — ultimately creativity does have a lot to do with drawing from many different influencers,” says McInerney, “It’s not that creativity should just be the average of whatever data we receive from our customers — rather the machine intelligent and the brewer, as a cooperative, can take inspiration from these conversations to create something new. In many ways ‘art’ is all about the relationship between the creator and the consumer, so we think this is all compatible to the notion of creativity.”
Obviously it’s not all about beer. The same technologies used here to brew beer are equally applicable in other verticals, including ones that are a bit less anthropocentric — Google for instance has reduced energy usage in its data centres by 40 percent using similar techniques.
With all the warnings that the robots are coming to take our jobs, it’s perhaps then somewhat reassuring to know that amongst the first creative tasks that our new artificially intelligent overlords have been set is to brew better beer.
