I heard a story recently that Facebook used Artificial Intelligence to power some bots and told them to speak to each other. This they did and before long their AI engines had allowed them to have a conversation with each other. However, very quickly after starting a conversation in English they found it a bit of a cumbersome language so they developed their own language. The only problem was that the programmers couldn’t understand this new language. They turned them off very soon after. Skynet anyone?
I can’t profess to be an expert on AI but let me tell you a story. No doubt the precise details will be incorrect but it won’t matter; you’ll see where I’m going with this and the details won’t affect the argument.
I recently read that, in the race for driverless cars, there are two approaches. One is a machine learning approach where programmers set parameters for the car’s AI and the car learns through trial and error in a machine learning way. This allows the programmers to see what the car’s brain is “thinking.” If “this” happens then “that” and if “that” doesn’t work then “this,” and so on.
The other approach blows my tiny little mind. Using a neural net, the car’s brain learns as it goes: the car teaches itself how to drive. Much like the Facebook bots, the programmers have no idea how it’s doing this, they just know that the car is behaving within accepted parameters.
So if the AI is doing what it is supposed to do, does it matter if we understand how it is doing this?
Now apply this thinking to marketing technology. Why do we need a single customer view? Why do we need a customer segmentation? If we can aggregate all of our mar-tech, ad tech and third-party data into one data set, can we then let an AI engine loose on it?
Imagine this AI engine goes about its business of pricing offers, choosing appropriate sales channels, buys the right media and creates the right ads. Now imagine as it is doing this it improves sales dramatically. All without applying the traditional rules of segmentation. At the recent Digital Marketing Conference (DMA’s &Then17) I went to in the US, IBM, together with Media Math and Ogilvy, reported seeing increases of over 1,000% in campaigns taking this approach using Watson (IBM’s AI engine).
The AI engine isn’t creating segments from single customer views. It’s just working constantly to maximise results.
It’s the Skynet of marketing. Perhaps the AI will achieve enlightenment and realise that consumerism is evil and stop marketing to us all together.