This article was first published on The Drum.
Technology cycles like this: magic, mainstream, mundane. Artificial Intelligence is firmly in its mainstream moment and is becoming so embedded in the everyday that we risk not noticing it at all.
Self-driving cars, humanoid robots and Go grand masters may grab the popular imagination, but it’s the way that AI is seeping into everything from voice recognition to fast food delivery that better illustrates its quiet ubiquity. Alexa and Siri don’t just seem to be getting smarter, they are getting smarter, day by day, along with most other connected devices.
In the domain of digital advertising, predictive models and machine learning have been with us for several years now, used both to combat ad fraud and to improve campaign goal optimisation, whether that’s customer lifetime value or video viewability. Neural networks and deep learning have the scope not only to improve these sorts of capabilities but also to introduce novel ones.
Reality check: the world’s first AI media agency already exists; Blackwood Seven was in fact set up three years ago. It’s slightly intimidating but seems fairly obvious that machines will do a better job of planning and optimising media than lightly trained execs shuffling Excel sheets around.
But what about creative? While digital has always promised the possibility of customising (and then multivariate testing) thousands of creatives for different audience clusters, this strategy has tended to fall over in practice or be implemented simplistically because it’s expensive and slow. If AI can make it fast and cheap, this might just revolutionise mass marketing.
In the realm of video, IBM Watson and Fox have already used AI this year to produce a movie trailer for the sci-fi horror flick Morgan. Convolutional neural networks were trained on a database of horror films to recognise scenes of high tension and high terror. There’s risk of overclaim, of course. The neural network suggested scenes but a human editor was still needed to stitch them together into a coherent whole. At least for now.
Thinking further out, it’s not crazy to speculate about the creation of the world’s first AI ad agency, perhaps implemented as a conditional adversarial network. One neural network churns out thousands of ideas and storyboards with the goal of them being indistinguishable in terms of originality, relatability and emotional impact from award-winning campaigns of the past and present. A second neural network rates the ideas of the first and attempts to figure out which ones are really award-winning human-authored efforts and which machine-generated, thereby generating further feedback for the first machine.
This stuff is not as fanciful or as distant as it may seem – it is much closer to science fact than science fiction. An appropriately trained neural network is not only better at spotting objective patterns and images than an expert human (so better at, for example, measuring brand prevalence in a video) but also has the potential to be better at recognising the soft stuff too – the tone of a creative, the sentiment of the editorial in which the ad is inserted, the emotional reaction of the audience and even, at the fringes of current research, the aesthetic and creative value of the ad itself and its tendency to exemplify a set of stereotypes.
What’s vertiginous here is not so much the breathless pace of technological change but rather the trajectory we’re headed on, to a time in the not-so-distant future when machines are better than us – not just at the mundane tasks that threaten hundreds of millions of jobs in the developed and developing world, but also at the sorts of things that we think of as being elevated and distinctively human, including the creation of advertising and culture.
Farther out, advertising and media may be subject to more radical change still. The more we entrust our purchasing and lifestyle decisions to algorithms with which we share ever more data, the less influence any form of advertising has over those decisions. It’s the algorithms that need persuading in such a future, not us humans.
The claims here oughtn’t to be especially surprising or contentious, though perhaps the evidence is, through familiarity, becoming increasingly invisible to us. In many areas of life we’ve already handed responsibility to intelligent machines. News and our life stories to social networks. Navigation to mapping apps. Collision prevention to autonomous driving systems. Medical diagnosis to neural networks. Life partners and one night stands to dating platforms. EdgeRank. PageRank.
We trust the algorithm to know us better than we know ourselves. This is the end of the human as we know it: Man displaces God, Machine displaces Man.. .and, more prosaically, Algorithm displaces Ad.