PSFK speaks with Creative Technologist Claudia Perlich about how machine learning is impacting advertising
Collaboration, at times, is at the core of innovation. Though most thinkers yearn for a “Eureka!” moment, most solutions result from a series of incremental ideas that build upon one other, often inspired by the creative thinking around them. This building process has a name, The Adjacent Possible, and is the theme of the 2015 CreateTech Conference, the annual gathering of creative technologists, developers, executives, and innovators hosted by the American Association of Advertising Agencies (4A).
PSFK sat down with Claudia Perlich, Chief Scientist at Dstillery, a digital intelligence company helping brands discover better ways to tell stories.
Perlich develops and optimizes machine learning that drives digital advertising and technology, operating at the intersection of big data, artificial intelligence and predictive modeling. We learn more about the promises and pitfalls of predictive marketing, and what that means for the future of digital advertising.
PSFK: Can you tell us a bit about your background, and how you got into machine learning and predictive modeling?
Claudia Perlich: In 1995, as a German exchange student in Boulder, Colorado, I met a German professor who taught a course on artificial neural networks. I had no idea what this was but it sounded very cool, and so I signed up for the course. That was basically the beginning of my involvement in data modeling techniques, and ever since I’ve been really excited about working with data and trying to tease out the stories it tells.
After I graduated, I went to the IBM Watson Research Lab to work at the Predictive Modeling Group with IBM data, as well as with clients who had data analysis challenges.
PSFK: Given your history at IBM, how has Watson changed the advertising landscape? How has it changed public perception of AI?
CP: Watson really changed public perception of what is possible with data. On the business side, practitioners in my field have been saying for a long time, “Oh, there’s more data than ever. We should be doing something with it.” Then, all of a sudden, everybody thought, “Man, if I don’t pay attention to this, I will probably be missing out.”
Watson involves training a machine to beat humans arguably at their best. The public is split into those who feel there’s a bright and exciting future as to how we will interact with machines, and those who are a bit worried about predictions of what artificial intelligence might be able to do that we might not be able to govern or control. I’m on the optimistic side: I think it will affect many parts of life for the better.
PSFK: How have the processes for machine learning changed in the past year and what does that mean for advertisers?
CP: The advances on the advertising side are basically driven by the rise of programmatic media, or the ability to buy and sell advertising in real time. In the early history of advertising, talking about billboards or TV, advertisers had very little information about their audiences. Even if they did, they couldn’t change the advertising to accommodate even minor audience dynamics. Programmatic media allows us to change what we do and what we message you in real time.
On the data side, I know enough about you to make an educated guess or to have a machine make an educated guess about what the right message is. We know where the person browses on the Web, we know which app the person is using, and the physical locations a person visits. That type of information is structurally very primary and very informative about who you are. We’ve moved away from demographic indicators; your actions and experience now speak more than gender or age.
Most of our work at Dstillery is leveraging machine learning to try to describe people better. This technology is known as predictive modeling (which is also one of the tools that Watson relies on), where we anticipate the needs and future behaviors of the user by looking at their recent past behavior. We can see, for instance, patterns that indicate you might have recently bought a home, which is the perfect time to show you an ad for home insurance.
At Dstillery, we’re looking at this wealth of data in its most original form without rolling it out into descriptive categories that lead to conclusions.
PSFK: Can you speak to some of the trends that you’re seeing right now within the realm of data science and marketing?
CP: I don’t know if I would call it a trend yet, but I’m personally interested in bringing out the creative side of machinery. I want to teach the machinery to ask: can we see how people react to that message? Can we message smarter? We know so much about people, so shouldn’t we better understand on the very large scale and on the individual level how messaging can be done best? This will bring the brand and the creators back into the optimization side of data.
PSFK: Share your thoughts on the growth of consumer AI tools—what will change about usability, data collection, and marketing?
CP: I think AI will touch our lives in many places. I already see the recommendation systems in Netflix and Amazon suggesting the next movie you should be considering. They do not feel natural yet, but we will see advancements in many of these places.
As Alan Turing said, “I will call the machine smart if it can fool me into believing that it’s human.” In that same sense, you reach successful applications of AI in data analysis if you forget that you’re dealing with a machine. Ultimately, I think it will touch every part of our lives, wherever we go and whatever we do.
It might be the security services at the airport, customer support for retailers or your next movie recommendation. These businesses will push AI into being real, but to us it will be as natural as speaking to a human.
PSFK: Do you find that your best ideas are recycled or borrowed from old ideas? What’s your best recycled idea?
CP: Many of the techniques we have packaged under predictive marketing have been “rediscovered” or developed in parallel with different disciplines. It is only with great reluctance that the statisticians finally agree with the computer scientists that, yeah, this is kind of the same thing, and that our methods are similar.
In my team, we don’t reuse ideas so much as we rebut them. The process of debate is what I enjoy most with my colleagues; it’s in debating our ideas where the breakthroughs happen.
Interview has been edited for brevity and clarity.
4A’s CreateTech Conference is an annual gathering of creative technologists, developers, executives, and innovators in advertising, media, and other digital industries. Tickets for CreateTech 2015 taking place on November 11 and November 12 in NYC are available now.