A Marketer’s Guide To Artificial Intelligence
How AI Thinks, Listens, and Sees
This week, the Interactive Advertising Bureau (IAB) launched a working group focusing on artificial intelligence and machine learning in response to “significant member interest.” Make no mistake, AI is quickly coming off the pages of science fiction and being implemented across the marketing, media, and advertising industries. So, what is AI and how exactly can brands make the best use of it?
How normal people (used to) think of AI:
What Is AI For Marketers
Artificial Intelligence (AI) is an umbrella term for a large number of technologies that can train software to think and learn on its own. While this might sound rather lofty, if you look around in the digital world, low-level AI is already being deployed in things from simple spam filters to the recommendation engines on ecommerce sites. AI is the brain that will power the smart robots, but it doesn’t necessarily need to mimic the human form; in fact, most of the AIs in use today and in the near future are just advanced algorithms.
The Status Quo of AI
Artificial General Intelligence (AI that has similar capacity to the human brain) is coming, but a ways off. Artificial Specialized Intelligence (AI that has deep domain knowledge or expertise) is here today.
As Tim Urban of Wait But Why puts it, what we have now is called “Artificial Narrow Intelligence” or “Weak AI.” This is the kind of AI that can beat the world chess champion in chess, but is useless when it comes to other things. After that comes “Artificial General Intelligence” or “Strong AI,” which matches human level intelligence. Then after that, experts predict there could come an “Artificial Superintelligence” that exceeds the best human brains in practically every field. (Whether the Super AI will lead to a robot uprising against humans is still up for debate.)
Obviously, today we are still in the stage of gradually moving from “Weak AI” to “Strong AI.” Capable as the likes of Siri and Alexa appear to be, they are still very much in the category of “Weak AI,” or rather, a bundle of specialized AI functions that are branded under a cohesive virtual persona. As AI technology continues to advance at an exponential rate, it is set to trigger the next round of industrial revolution, just as the invention of the steam engine, electricity, and the internet did previously, pushing human productivity to new heights and significantly impacting the way we work and live and consume.
Given its significance, it’s no wonder that most of the major tech companies have developed their own AI platforms. Companies like Google, Apple, Facebook, IBM, and Baidu are leading the charge in bringing AI into consumers’ hands, while AI startups like Deepmind, which was acquired by Google in 2014, and OpenAI are leading the pack in AI research. More importantly, they’re also making many of their tools available to third-party developers and brands. By leveraging those tools, marketers can tap into the power of AI to enable more precision in brand messaging and targeting, uncover better insights, and optimize our customer experiences.
Given the magnitude of disciplines that AI covers, there is a myriad of ways that brands will be able to make use of it. Today, however, the biggest AI opportunities for consumer brands lies in machine learning, natural language processing, and computer vision. Amazon, for example, offers developer tools like Lex (a chatbot framework), Polly (for text-to-speech), Rekognition (deep learning-based image analysis), and Machine Learning in its cloud service AWS. AI demands a huge amount of data to develop and improve, and by building on these platforms, we can benefit from their scale.
A sidebar on IBM’s Watson: As one AI expert and industry insider summed up in a recent Wired article: “[Watson has] really over-claimed what they can deliver in a big way; the only intelligent thing about Watson is their PR department.”
Machine Learning — How AI Thinks
As the core component that backs most of AI, machine learning is one of the most straightforward forms of narrow AI that brands can implement today.
Traditional computer software is strictly defined — the programmer creates the logic that will allow the computer to generate a response, based on the data or user input it is presented with. This doesn’t change until there’s an update to the software, with the programmer’s manually-created improvements.
In comparison, machine learning allows software to think more like humans: taking feedback on its performance, and incorporating that feedback into future decisions, without intervention by a human programmer. The programmer moves from being a drill sergeant to being a kindergarten teacher, improving how the software learns, rather than dictating what it does. But, as with children, some of that decision making will remain a black box.
The most notable implementation of machine learning is currently in retail marketing. Whether it’s search optimization, recommendations, or programmatic ads, retailers have been making good use of AI to interpret a large amount of consumer data and fine-tuning its messaging accordingly.
eBay, for instance, is supercharging its own email marketing platform with machine learning so as to send individually personalized offers to subscribers based on their browsing history on eBay’s site. By dividing deals into virtual “buckets” such as shoes, consumer electronics, or collectibles, eBay’s proprietary algorithm is able to insert the relevant deals from one of the “buckets” that a user frequently views.
Beyond retail, machine learning is a great tool for any brand seeking to understand their customers better. In a sense, it is essentially the same thing that marketers were calling “big data” or “data science” two years ago, just with a sleeker name. One key difference here is that machine learning tools will soon evolve beyond simply processing and interpreting data, and start to be able to handle some of the creative work in advertising. Indeed, Mariano Bosaz, the global senior digital director of Coca-Cola, has expressed interest in leveraging machine learning to improve content, media, and commerce, especially in streamlining the ad creation process and experimenting with automated narratives. The company also recently revealed that it is working on an AI-powered vending machine with an accompanying mobile app, but it hasn’t shared many details on how exactly it will work.
Natural Language Processing — How AI Listens
Natural language processing (NLP) is the ability for software to understand normal human language, whether it’s input via voice or text. In other words, it allows AI to communicate with humans in the human way.
NLP is often deployed to power a conversational interface — a chat or voice interface, which is in contrast to a graphical user interface, as we normally use on our PCs and smartphones. It is a particularly important component of AI for marketers because it directly impacts brand-consumer interactions. For example, as voice search starts to take off (Gartner predicts that 30% of all searches will be done without a screen by 2020.), marketers will soon be confronted with a brave new world where the old rules of SEO and SEM no longer apply.
A conversational interaction feels inherently more personal and intimate than the browsing-based or command-and-response models that marketers typically use when designing digital experiences, so it will be crucial for brands to start developing an authentic brand voice and leveraging NLP to engage with customers in an intuitive way.
In addition, mastering NLP will be especially important for brands with an international presence, considering that the next billion of internet users/digital consumers — a swath of underprivileged and less-literate people in the global markets — will likely favor voice recognition over text.
While NLP platforms like chatbots are still emerging for marketers, even those not currently deploying them are excited about the customer service applications. And, indeed, customer service is merely table stakes for NLP — it can offer deeper experiences, but customers expect anything that can talk can help them with their questions and concerns.
The NiroBot that we created in collaboration with Ansible for Kia is a good example of leveraging NLP to engage with potential car buyers via a chatbot. Available on Facebook Messenger, the bot could answer questions about the vehicle features and customization, and also delivered pithy answers to off-topic queries.
Computer Vision — How AI Sees
Computer vision is the ability for software to understand photos and videos just as easily as it’s always been able to understand text. If NLP allows our AIs to hear and speak, computer vision allows them to see.
The proliferation of smartphones is turning the camera from a media capture method to a new kind of input device. And the recent development in computer vision is making the smartphone camera capable of recognizing faces, objects, clothing, and even the layout of the room, allowing brands to gather information from facial expression, contexts, and other nonverbal cues so as to better serve their customers. For example, the Hawaii Tourism Authority worked with Expedia to launch an online campaign that leveraged facial recognition technology to gauge prospective travelers’ interests and offer personalized recommendations.
As with NLP giving rise to voice search, computer vision is quickly making visual search increasingly useful and reliable. Both Pinterest and Google are launching their respective “Lens” visual search product this year to allow for easy, seamless search as simple as snapping a picture. Needless to say, this will vastly expedite the product discovery process and make the future of search marketing even more complicated. Pinterest Lens, for example, aims to capitalize on visual search by surfacing similar-looking items that people can click to buy when they search for certain apparel and fashion items.
Another exciting new area that is emerging on the back of computer vision is augmented reality. As with machine learning, major tech players are all invested in AR. Microsoft currently leads the pack on the hardware front with the HoloLens headset, and Google officially brought back Google Glass this week as an enterprise use-focused professional device. To reach a broader audience, companies such as Facebook, Apple, and Snapchat are all launching general mobile AR platforms, using the smartphone as a window for layering digital data onto the real world. We’ve written extensively about the opportunities and challenges that AR will unleash for brands, which you can check out here.
Looking further, computer vision will help usher in even more futuristic devices like autonomous cars and wearable AR glasses that will soon emerge as a new smartphone accessory, both of which will open up new time to media experiences. When the AI is able to see and fully understand what it sees, the amount of data it can collect increases exponentially, which, in turn, will make the AI even smarter and more useful.
One gadget that perhaps best indicates how AI-powered devices will impact the daily lives of consumers is the Amazon Echo Look. As the latest addition to the Alexa-powered Echo lineup, the voice-controlled selfie cam combines the power of machine learning, NLP, and computer vision to offer users outfit recommendations (and, of course, purchase recommendations) by taking selfies of you in different outfits and learning your wardrobe over time. The accompanying app even has a feature called Style Check, which utilizes machine learning to compare different outfits and tell you which one looks better.
As we enter the age of AI, it is paramount that we marketers keep up with the developments, especially the available AI platforms that brands can build on, and figure out how to make use of the intelligence that AI brings to marketing.
