All eyes on AI
Machine learning: Buzzword or big opportunity?
It’s often said that change is the only constant in life. But the current pace of change is truly unprecedented.
For evidence, hot on the heels of the Third Industrial Revolution – digitisation – we’re already onto our fourth. IndyRev 4.0 is bringing us quantum computing, nano-technology, 3-D printing and advanced robotics. While they’re all exciting in their own right, artificial intelligence and machine learning are creating the most noise in branding and design.
But what they both, exactly?
Well, to summarise: AI encompasses the various technologies that mimic ways humans process information or learn. Machine learning (ML) describes the processes that underlie AI. You might also have heard of Deep Learning – a kind of ML on steroids. We won’t go into that here, but I do want to have a look at how AI and ML are changing consumer outlooks, and what the biggest AI and machine learning business applications are.
The Amazon effect
You can’t understate the impact Amazon’s had on customer expectation. First, they removed the effort going to the bookshop. Then, with Prime, the wait for the book. Then, with Kindle, the book itself. With each step forward, they’ve made the obstacles to a consumer wish fulfilment vanishingly small. A key component of the Amazon model is their recommendation engine, that uses billions of data points to personalise suggestions to a level where they feel truly insightful, rather than an obvious up-sell. (Recommendations account for around 35% of total basket size on amazon.com.) The net result means we’ve all been trained to expect never to be more than one click or, with Dot and Echo, one voice command away from the products we want, might want, or sometimes didn’t even realise we wanted.
So, what can smart brands make of all this?
Optimal efficiency as standard
The first and greatest of our AI and machine learning business opportunities is improved efficiency. Airbnb Design is currently exploring how to reduce sketch to prototype time to zero – its developers are using ML to connect hand-drawn sketches with UI libraries to make working coded prototypes of web-pages in the blink of a second. Even the testing could be automated, using learnings from thousands of previous trials. Meaning what used to take several people days or even weeks to develop and test could take one person less than 24 hours.
Perhaps, in the near future, one single person will be able to hand-sketch a poster for a billboard ad for a new product. Almost instantly an AI will choose the best picture that aligns with that brand’s values and the specific message. It will also calibrate the size ratio between visual elements to achieve the right visual load balance, and the finished design will be put into performance testing with thousands of other billboards, to feedback as to how it will perform. A single person could iterate on a particular design multiple times in one day instead of a small studio of personnel, over several days, weeks or months.
Conjuring up content
There used to be a saying, ‘the camera never lies’. Hardly the case today, with Photoshop and Insta and Snapchat filters putting the power to edit images at everyone’s fingertips. Pretty convincing video content, or ‘deepfakes’, can also be created with tools like FakeApp and Voco (a kind of Photoshop for audio). They came to wider fame recently when the University of Washington and ‘Get Out’ director Jordan Peele synthesised eerily convincing fake messages by Barack Obama.
Pretty scary, when you consider the political possibilities, although artificial intelligence can also be trained to spot the tell-tale signs of AI fakery and flag the frauds out there. But it also has the potential to be highly cost-effective for brands wanting to satisfy content-hungry consumers without incurring massive production costs.
For your eyes only
Hyper-personalisation is another buzzword of the moment.
ML has the potential to reduce much of marketing’s scattergun wastefulness. Using behavioral data, it will increasingly enable marketers to segment audiences in almost unlimited ways – revealing and quantifying behaviours, perceptions, attitudes and needs, on a micro level. ASOS now uses members’ past purchase history to automatically recommend which size they might need. They also take into account whether the item was returned for not fitting properly. Additionally, they allow customers to make their suggestions even more accurate by evaluating how well past items fitted and adding details on their height, weight, tummy and hip shape, bra size and age. Interestingly, it’s had a mixed reaction from consumers. Fashion blogger Hannah Gale told her 13,000 followers she wasn’t keen on the update: “Stop trying to tell me what size I am, you don’t know me,” she wrote.
When it comes to building advocacy, ML can zero in, not just on the more obvious influencers, with their many thousands of followers, but on micro-influencers and even nano-influencers, with an audience of only few dozen. The outreach process could be automated, with a chatbot doing the work of recruiting them to the brand’s cause. This presents a great opportunity for brands to achieve an authentic and nuanced self-expression – represented not by tens, or even hundreds, but tens of thousands of active advocates.
Heart of the matter
Recently, I’ve also been impressed by Google’s messaging app, Allo, that they’ve pre-installed on their Pixel phones. Allo includes a virtual assistant that generates automatic replies which, in my experience, hit the nail on the head pretty much most of the time. In fact, about a year ago, it nailed it to the extent that it predicted (accurately) that I was falling in love with the person I was chatting to. I have to admit, this seemed a bit creepy. Even though, when I thought about it, I had to admit the app was right, it still felt pretty invasive. It’s one thing for a friend to say, “Hey Pedro, I think you’re falling for that girl you’re chatting to.” It’s quite another when Google gets that personal – and gets it right!
On more literal matters of the heart, a new model from Apple Watch, that can predict whether you might have a heart problem or diabetes, is currently pending approval from the FDA. Some of Google’s top artificial intelligence researchers are trying to predict people’s medical outcomes as soon as they’re admitted to hospital. A recently published research paper claims unparalleled accuracy at predicting a patient’s likelihood of dying in hospital, being discharged or readmitted and their final diagnosis. They used data from nearly a quarter of a million patients of two major Californian medical centres, with more than 46 billion data points between them. Their biggest claim being the ability to predict patient deaths 24-48 hours before current methods – which could in future allow time for doctors to administer life-saving procedures.
Many of us think of ourselves as creatures of habit, but another machine learning business application is that it has the scope to steer our behaviour in previously unimaginable ways. Unless you’ve been hiding under a rock for the last two years, the terms ‘Cambridge Analytica’, ‘election meddling’, and ‘fake news’ can hardly have passed you by. As a society, we’re increasingly aware of just how energetically hidden forces are working to single us out and tell us exactly what (they think) we want to hear. To win our hearts and minds, our votes and, of course, our cash.
In response, both Facebook and Google are working equally hard to rehabilitate their reputations. Google has produced a ‘leaked’ video to measure the public’s response to the prospect of a future world in which idealistic super-intelligent AIs could help usher us towards broadly beneficial outcomes by shaping the behaviour of the entire populace. Sounds freaky? Make up your own mind below.
They’re also adding deep learning on their Android operating system, to predict the apps you may want to use, or even stop using, as part of a drive to give everyone ‘the tools they need to develop their own sense of digital wellbeing. So that life, not the technology in it, stays front and centre’.
A less risky business
To sum up, the core purpose of AI and ML is to use data to reduce the personal bias in decision making, and thereby reduce risk. Starbucks manages to open new branches cheek by jowl with existing ones, almost counter-intuitively. (Normally, when expanding a business, it’s needlessly risky to open a new location just round the corner from another one.) Starbucks uses big data to determine the potential success of every site prior to expanding their operations. Using location-based data, traffic data, demographic data, and customer data, they’re able to estimate the likely profitability of each location, with astonishing accuracy.
So, what are the big out-takes in the short and the medium term?
What’s clear is that, from both from a business and a consumer point of view, the role of brand is becoming more important than ever, if they’re increasingly going to be able to get as close to us as our most trusted friends. Only a clear, engaging and authentic brand identity will earn itself the right to make truly personal observations and recommendations, without over-stepping the mark.
I mean, it’s one thing taking relationship advice from your best buddy. But from your breakfast cereal? Someday, maybe. But we’re not there yet.