Country Manager, WPP France
CEO, GroupM France
The intelligent revolution on our doorstep
We’ve grown accustomed to the digital revolution, but the next stage will be led by machines, and we have to prepare for it.
Artificial intelligence became part of mainstream conversation with the rise of AlphaGo, and before that, there was IBM's Deep Blue. This was the technology that, having been beaten by chess master Gary Kasparov in 1996, went on to defeat Kasparov, finally, in 1997. We all said then that machines had become intelligent, yet it took another 20 years before we saw true artificial intelligence.
AlphaGo learned by studying players of the game “Go”, and then played – and defeated – human Go players. It became European Go champion 2015, then world champion in 2016.
And then we said to ourselves: That's it, the machine is more intelligent than us! It had been two decades in the waiting, then something happened six months later that is the real AlphaGo revolution.
It's AlphaGo Zero, which has a key additional feature: it doesn’t just play against humans, it plays against itself.
This time, the computer taught itself the game of Go. It was given the three basic rules and then it played and played – millions of times, until eventually it beat the old AlphaGo technology 100 games to zero! It played shots that no human could have imagined; shots that had never before been played.
So imagine if we took programmatic buying of digital media, which is already effective, and then added the kind of artificial intelligence behind Deep Blue and AlphaGo Zero. We could give it a set of rules to play by, set the goal, and then leave the computer to do what it “thinks” is best.
IBM had the means to do exactly that, and equipped computer technology to spend 18 months learning by making 53 million media purchases in the UK market. It learnt so much that, just like AlphaGo Zero, it became much more effective at media buying than IBM’s media agency in that market.
And here are the results: the cost per click was reduced by 35 percent; an amazing performance! This technology learnt to optimize buying according to five parameters: time, device, language, format and location. There was just one objective: improve click-through rates. A complex challenge, but one that IBM’s Watson technology made look incredibly simple.
The future of this is in what IBM calls the “dark data”, that is, by adding parameters that are a little more subjective: using casual language in communications to match the context in which it appears.
This might sound futuristic, but the US lingerie e-commerce site Cosabella is already doing this. This 35-year-old brand decided to affect a digital transformation, and did this using artificial intelligence – a startup called Albert, to be precise. The result: 336 percent growth in sales for an e-commerce site that was already leading its sector. This marks a turning point in the world of e-commerce. Albert has achieved autonomous targeted media buying.
So, what’s the next step? Video. This is where advertiser investment is moving. Wise Air, an Israeli startup, has already developed an extremely strong system of optimizing online video advertising buys using artificial intelligence, with a self-learning algorithm. Still, we are in the early stages of artificial intelligence.
Within a year, though, we expect this kind of technology to be fairly commonplace. At WPP, these kinds of algorithms have already begun to be implemented into programmatic buying platforms. When we add to that the human touch, we can ensure relevance. This way, we can stop bludgeoning people with advertising messages they don’t want, and can optimize the machines to achieve something more meaningful and effective than just reach and repetition.