The CMO’s AI: 5 Next Frontiersby Frank Delmelle - June 15, 2016
Since all of marketing’s challenges can ultimately be quantified
– i.e. taken care of by learning machines, which still is marketers’ predominant belief, it seems, hadn’t CMO’s better begin to imagine what they want the bots’ next major, android milestones to be?
“Over the past several years, the biggest tech companies in Silicon Valley have aggressively pursued an approach to computing called machine learning.” The recent reminder is Wired’s Jason Tanz’, who continues to put AI into context as follows: “With machine learning, programmers don’t encode computers with instructions. They train them. If you want to teach a neural network to recognize a cat, for instance, you don’t tell it to look for whiskers, ears, fur, and eyes. You simply show it thousands and thousands of photos of cats, and eventually it works things out. If it keeps misclassifying foxes as cats, you don’t rewrite the code. You just keep coaching it.”
Now imagine that coach would be you. If tech were all it would take to lift your marketing onto whatever next level, where would you, as a CMO, want tech to take the rest of us? Below are five random, personal suggestions, put on a – scify-ish? – ‘marketing machine learning timeline’… that just might contribute to an inevitable debate.
Remember Tay? “In the 24 hours it took Microsoft to shut its chatbot down, Tay had abused President Obama, suggested Hitler was right, called feminism a disease and delivered a stream of online hate.” Could Tay be cured of its propensity to mimic ‘nazi’ comments? Can machines be taught good manners?
As a matter of fact, they can. At least according to the passionate Quora user called Robert Stone, even if it would take training the AI ad infinitum: “Manners are part of a cultural system that identifies generally expected behavior responses. Manners can be encoded as beliefs for response. Manner responses are going to be decisions by an entity whether to use a rule or modify it depending on circumstances. To train an ANN-based AI manners one would have to present encoded situations and train it on a variety, until the rules are embedded. (…) So, yes manners could be taught,” Stone concludes, ‘but the setup is not very simple.”
Hyper Island’s Shirley Sarker recently dug up a curious piece of reporting entitled ‘If TiVo Thinks You Are Gay, Here’s How to Set It Straight’ by the WSJ’s Jeffrey Zaslow: “Zaslow interviews a gay man who purchased a popular gay tv show series on Amazon and was inundated with gay-related calendars and book suggestions. Having then purchased a baby book for a pregnant friend, Amazon donned him a “a pregnant gay man”. The customer then proceeded to trick the data set to set new guidelines of personalization by inundating it with additional data. Zaslow writes how the gentleman “searched for other stuff — on politics, computers — so it would stop throwing baby books” at him. The customer now suggests his current profile is a man who has abandoned his baby and is preparing for a career in politics.”
Will machines ever acquire the ‘common sense’ required to rule out “pregnant gay men”? “Write enough rules,” the AI optimist would answer, “and eventually we’d create a system sophisticated enough to understand the world.”
Next up, however, could be questions such as ‘how much of humanity will be sacrificed along the way?’ “We are twelve in my body,” the philosopher Achille Mbembe once wrote, “We are packed like sardines.” In his newest book, by the way, Mbembe issues another dire warning, he wouldn’t mind being summarized as “reduced to numbers we’re all ‘niggers’.”
The extent to which tomorrow’s learning machines will circumnavigate pitfalls such as reduction or abstraction remains to be seen. And that sure is another issue worth looking into. We wouldn’t want to risk, after all, ending up marketing to a tattered ‘segment of one’ resembling Van Doesburg’s famous Cow, that “painting literally demonstrating the meaning of "abstracted" in that it simplifies and reduces the thing depicted.”
What could be next on a CMO’s AI wishlist? Provided growing levels of consumption and/or investment still are marketers’ key aspiration by, let’s say, 2027, optimism would be a great skill for AI to acquire. Imagine tech-controlled consumer optimism or even bots being optimistic on our behalf…?
“Have you noticed, Yanis Varoufakis recently asked a keen audience of Googlers, “that in the United States today, we have the following very interesting yet very worrying phenomenon: we have extremely low rates of interest, extremely high profit rates and very low levels of investment. How is this possible?”
“If investors fear that the level of investment will be low, then the level of investment will be low,” the former Greek finance minister / economics scholar explained. “The key parameter determining whether we go from the good to the bad, or vice versa – is average optimism.” Omit optimism, in sum, and economies crumble. Hence the desirability of optimism asap being quantified, … especially since Nielsen sees “Worry Lines Deepen for Many Around The World.”
As made painfully clear by the recent revelation of the human editors Facebook’s been hiding in its basement to curate its ‘Facebook Trends’, algorithms so far rely on human intuition to get – and keep – their mathematical magic going.
“Algorithms are in fact full of people and the decisions they make,” was Microsoft Research’s Tarleton Gillespie’s kind and concise reminder. “Any algorithm that has to make choices has criteria that are specified by its designers. And those criteria are expressions of human values. Engineers may think they are “neutral”, but long experience has shown us they are babes in the woods of politics, economics and ideology.”
Artificial intuition. Machines capable of formulating adequate questions for their own computing power to answer. Will they forever remain science fiction? And, subsequently: will we ever rely on bots to come up with new ideas? We won’t, it seems, if today’s most eminent human physicist, Stephen Hawking, is right stating that "there is no prescribed route to follow to arrive at a new idea. You have to make the intuitive leap."
Hawking’s own scientific reputation, by the way, is barely owed to mathematics. For the past several years he hasn’t even put any more math on paper. If he’s still considered a superior physicist nonetheless, that’s due to his unrivaled – human – talent to ask math and machines the questions that propel science forward.
That sheer weight of intuition also explains, by the way, why Hawking used to systematically drag his students to the opera every other month, to have their sense of wonder sharpened by Wagner’s Götterdämmerung among other works of art. (Professor Thomas Hertog of the Institute for Theoretical Physics recently mentioned these nights at the opera as among the most memorable moments of his time as a student of Hawking.)
Ask ‘the class of Hawking’ and every single great physicist will confirm: “If you cannot measure it, that doesn’t mean it doesn’t exist.”
If artificial intuition is problematic, love by AI, by definition, seems even more unlikely. Even when “robots are so mainstream, we're now having sex with them,” as CatchNews recently reported, love – that different beast altogether – implies levels of illusion utterly alien to even the sexiest machines. (See e.g. Jens Christian Grøndahl’s novel Jernporten (The Iron Gate) for a great read providing ample evidence.)
Still, that doesn’t seem to stop CMO’s industriously automating their ‘relationship marketing’ and/or rejoicing in retargeting, marketers putting all their hopes in predictive analytics or companies diligently decoding their (future) customers’ most intimate ties. Facebook’s Mark Zuckerberg has gone so far as to suggest, Wired’s Jason Tanz writes, there might be a “fundamental mathematical law underlying human relationships that governs the balance of who and what we all care about.”
What good could it be to believe even artificial love will eventually be a mere matter of feeding the machine enough whiffs of illusion?
“Our industry has become data first, human second,” Leo Burnett’s Ali Amarsy recently concluded at the Dubai Lynx. Tech won’t do the trick in and of itself, unfortunately. Tomorrow’s great leaps forward will not be delivered by the self-flying drone alone. And chances are, even in 2030, it will still be up to us, unquantifiably human marketers, to ask ourselves what (not) to ask our beloved bots.