Artificial Intelligence Is Finally the Real Thing
By John Patrick Pullen
Imagine it's a sunny weekend in May. Unless you're in the raincoat industry, this kind of weather is poised to make your bottom line giddy, because if the sun is shining, your customers are likely out enjoying it — lacing up some sneakers, drinking a soda, uploading photos via an app, or even just reading a book in the park. Whatever the case, the concept of using a local forecast to buy relevant advertising isn't necessarily the sales secret it once was. "An advertiser who's selling gardening tools and has been doing it for 20 years already understands this," says Ritesh Soni, VP of data science at SapientRazorfish. "That's what you may consider tribal knowledge."
Now, instead, imagine it's not the weekend. Maybe it's June. Perhaps it's snowing. Or maybe it's something really strange, like cloudy with a chance of meatballs. When you're trying to plan around changing key variables to the ad targeting and conversion equation, you're much less likely to have the same success you did previously — after all, you're only human. "It's very difficult for a human analyst to sift through the thousands of factors to come up with the perfect decision engine," Soni says. Yes, but not difficult for a computer.
Over the past 50 years, computer scientists have promised that this thing called "artificial intelligence" would soon come along and save people from making complex and confounding decisions. Seemingly always five years away, suddenly AI has arrived, most notably in the form of personal assistants on mobile and in-home devices. Now the question on every C-suiter's lips is "What do I do with it?"
"We're now in this period of really quite significant change where society and, in fact, a whole bunch of engineers and entrepreneurs are all racing in a slightly confused-but-excited fashion, because things are changing so fast," says Andrew Moore, dean of the School of Computer Science at Carnegie Mellon.
A former VP of Google Commerce, Moore led initiatives in advertising and shopping for the search giant, and is one of the preeminent AI experts in the world, having been named a fellow of the Association for the Advancement of Artificial Intelligence for his work with data mining, machine learning, and algorithms. "There's nothing magic in AI," he says. "AI is simply a very large scale computation."
What Exactly Is Artificial Intelligence?
According to Moore, artificial intelligence really consists of just two key parts: data and an automated process that can make decisions based on that data. Every marketer who has waded into the pool of online and interactive advertising knows the wealth of data that has long existed in this space, but they probably hadn't thought about it in terms of AI until only recently.
The obvious change bringing AI to the forefront has been the recent emergence of products from major technology companies: Amazon's algorithms and Alexa assistant, Apple's Siri, Facebook's various tools and algorithms, Google's Assistant and its keyword-based ad system, IBM's Watson, and Microsoft's Cortana assistant. While these various programs all perform varying tasks, they also each have the potential to make advertisers and marketers more efficient and innovative moving forward.
Take, for example, Google's AdWords program. From roughly 2000 to 2010, large-scale statistics were being applied to keywords and text, with Google's computers looking for and optimizing the relationship between the terms that users had typed in. "That I would call very large-scale machine learning or very large-scale statistical analysis," Moore says. Pair that insight with analytics derived from cookies and other markers, and you're able to compound the intelligence gleaned from simple keyword searches exponentially.
Industry insiders might think of this approach simply as online advertising or programmatic ad buying, but what's actually going on behind the scenes is algorithms trying to understand the meaning of the queries received so they can give a meaningful response, says Moore. Or, in other words, this myriad of practices that the ad and tech industries have been refining for close to two decades is actually, finally, artificial intelligence.
Where Did AI Come From?
In varying forms, research in AI has been around since the mid-1960s, though it was initially used in areas outside of commerce for things like software verification. It wasn't until the mid-1990s, when new theories in statistical learning — combined with a drop in data storage prices and new methods of handling very large data sets by Yahoo! and Google — made gathering actionable information from data possible. Since then, the combination of those three elements have led to the commercial applications we're seeing today in products like Amazon Alexa.
Technology developed in computational advertising faster than almost any other field. "If you think about the earlier adopters of AI — large ad networks like Google, Yahoo!, or Bing — it's always been in the background for them to balance things like consumer value for advertisers," Soni says. "That was the beginning of the AI renaissance back in the mid-1990s."
And Moore, with his post at Google, was right in the thick of things. "Some of the big Internet companies, the machine learning systems they had for click-through prediction and conversion prediction, were at the time the biggest, most advanced machine learning applications on the entire planet," he says, adding that technology developed in computational advertising faster than almost any other field.
Advertising's AI advantage all had to do with the vast cache of reliable data it had collected through the years. From around 2000 to 2005, the world of search engine–based and click-based advertising, along with the natural growth of the web, generated huge amounts of data at rates other industries were not able to match. "Even in parts of science and tech — which you think are integral to the economy and the well-being of the human race — there was less data available," Moore says. Or if it was around, it was stored in silos, unable to be utilized by machine learning.
"Advertising was the first of the major planetary implementations of machine learning where you have systems learning from what's going on around the planet in real time," Moore says. "You think of things like science or physics or biology or cancer research as being the places where people are going to be doing the most sophisticated machine learning, but in fact it really happened big time in advertising first."
What Can Advertisers Do with AI?
Opportunities for advertisers to wield the massive power of AI are about to explode. "We've just gone up a steep improvement curve," says Moore, who now predicts a technological plateau. Even so, he says, while the advancements level out, new innovations will emerge. "We're going to see lots of applications which people haven't thought of yet based on the existing plateau."
The most noteworthy example of this plateau is the feverish development currently underway in the voice-enabled assistant space. This began with Apple's introduction of Siri, followed by Google Now, Microsoft's Cortana, and now Amazon Alexa. "The big push is don't just look for keywords," Moore says. "Figure out what the user wants and try to provide them with a solution — that's the new stuff."
Currently, Amazon Alexa is the hottest space for developers. With fourth quarter 2016 unit sales growing nine times over the previous year (the company has never provided specific numbers), the Amazon Echo smart-speaker was the breakout gadget of the most recent holiday shopping season. And now home to some 4,000 "skills" (the Alexa platform's name for apps), it has also become a must-use for brands, too.
Companies like Fidelity Investments leaped on to the platform early when, in late 2015, the financial firm built a simple stock-quote lookup tool that got the brand's name into the fray. But that was only the start. Listening to how Alexa owners were using the tool, Fidelity was able to gradually make adjustments and add more features. "We had our core capabilities up in no time," says Brady Frost, Fidelity's director of mobile personal and workplace investing, in a post on Amazon's developer blog. "Subsequent releases were just a matter of tweaking based on what we learned to ensure the best possible customer experience." Moving forward, the company can use its Alexa experience to build more complex AI applications, like customer authentication tools.
Whether it's in an app, an ad, or a skill, AI should be leveraged by marketers now, if they haven't done so already. Taking AI farther afield from voice assistant technologies we've already become familiar with, IBM Watson has recently introduced another emerging AI opportunity for advertisers. Watson Ads, a program that lets advertisers connect with consumers directly via one-to-one voice and text interaction, leverages IBM's heralded AI to give brands access to the platform's vast caches of data.
Watson Ads' initial campaign rolled out between Campbell's Soups and IBM-owned The Weather Compay (TWC). In it, Campbell's placed ads on TWC's website and apps that featured an AI assistant that could help deliver personalized recipes for the consumer. By simply tapping on the IBM Watson logo, consumers could summon Watson, say aloud a few ingredients that they already had in their cupboard, and the AI would return a dinner suggestion — complete with directions and a recommended Campbell's product to accompany it, all suited to the day's weather, of course.
"While in this increasingly cluttered world it is getting harder and harder to break through and engage people, technology is helping to find ways to connect with more relevant content," Keith Weed, chief marketing and communications officer at Unilever in the U.K., said of the Campbell's ads in a statement. "This will help us to create better, more engaging content that matches our consumers' interests and unique preferences." The campaign rolled out in June 2016, and Watson Ads has since hooked on with more than a half dozen other brands.
AI's advertising genie is clearly out of the bottle, and it's not just vested industry insiders making this claim. According to an International Data Corporation prediction, more than half of all apps developed in 2018 will incorporate AI in some way, a trend that bodes well for its adoption across industries, Soni says. "Increasingly it's becoming much easier to incorporate AI into the development process," he adds.
And whether it's in an app, an ad, or a skill, AI should be leveraged by marketers now, if they haven't done so already, Moore advises. "As we see the rise of Alexa and the other systems, when a person asks for help they're no longer going to be looking for a web page full of results and they choose a result to click on," he says. "I absolutely do not think you can sit on your hands and [wait to] see how it turns out and expect keyword or even display ads to be carrying on as they were before."
Image Credit: Alex Nabaum for ANA magazine
You must be logged in to submit a comment.