An early draft of what ultimately became Unpacking Artificial Intelligence.
–
Early in 2017, Hebrew University of Jerusalem historian and Sapiens author Yuval Noah Harari released Homo Deus, a book projecting what the future may hold for our species. Among the most provocative suggestions revealed toward the end of the text, Harari’s affirmation of the existential risk of artificial intelligence seemed to validate the intellectual hobby horses of Elon Musk, Bill Gates and Stephen Hawking — all of whom have been keen to preach about the existential dangers of artificially intelligent technology.
We are, of course, no stranger to these ideas. Our feeds have been inundated this year with think pieces about the professional, industrial and existential implications of AI. Although the extent to which marketing executives actually buy into any of this hype isn’t entirely clear, Publicis CEO Arthur Sadoun’s decision to opt out of Cannes next year—instead using the savings to create an intelligent marketing platform called Marcel—signaled perhaps most glaringly a belief among some marketing executives that ongoing advances in AI technology will have colossal implications for marketing and business more broadly.
Hype-Checking
To be clear: There were undoubtedly some significant advancements in AI research this year. Equally important are trends such as the rise of cloud-based AI services being offered the tech titans, coupled with the growing popularity of conversational interfaces. Collectively, all these represent significant evolutions in our socio-cultural fabric being driven by AI technology. These macro trends are most important to keep in mind when considering how AI will shape the business playbook moving forward.
Because although there are emerging applications of AI technology in marketing, in truth these services are still in their infancy; many CRMs and other marketing tools are simply AI-washing their products, tacking the term onto their solutions to leverage the hype. In this context, marketers are wise not conflate stories pertaining to advancements in deep neural network technology—such as headlines about AI’s ability to master our most challenging strategy games, predict your sexual preference, and invent its own language—with mystifyingly game-changing technologies that diminish the value of our profession (alongside that of our species).
Not only is the history of AI riddled with “AI winters”, wherein the technological advances failed to live up to the hype — but it doesn’t take a veteran marketer to recognize that overhyping is a perennial issue for new business technologies (and that’s all to say nothing about the sensationalism incumbent upon our hyper-saturated media ecosystem.) These caveats don’t preclude us from a candid assessment of recent advances in this technology, but it’s important to start with some background of the technology before assessing the state of AI from both macro and micro perspectives.
AI Abundance and The Skills Grid
We don’t have to go to square one when considering the role of AI in marketing — in fact, AI is already commonplace in both consumer and marketing tech. For instance, Google search is powered by AI. In addition, many popular recommendation engines—Facebook’s Newsfeed, Spotify’s Discovery Weekly—use AI to surface relevant content; intelligent assistants rely on AI to recognize the sound of your voice and understand your intent; Snapchat’s facial recognition technology is based on AI technology; even Pinterest has a bumpin’ AI lab. Arguably on the forefront of these technologies, new computer-vision products such as self-driving cars and the facial unlock feature on the iPhone X also rely on AI.
When it comes to marketing, it’s foremost important to be clear that there are no shortage of cognitive routine tasks in modern marketing operations. In fact, wasted cognitive labor spent on calculating media metrics is practically incumbent upon of our hyper-saturated media ecosystem. Given this context, it follows that most future AI-related marketing services will naturally involve assisting with these sorts of rote tasks that amount to chores for most marketers. Today, these types of AI are already manifest in things like programmatic advertising, email lead nurturing and dynamic content optimization. In addition, things like landing page development, analytics storytelling and media mix calibration are quickly becoming garden variety marketing tools, all of which technically incorporate some AI technology.
At the clip that these features are evolving and coming to market, they’re becoming so commonplace that we don’t even consider them AI any longer. This is one of the reasons that AI has historically had a definition problem: As soon as a capability is classified as AI, it’s not long before it can no longer be considered AI, since humans no longer perform the task. For example, the ability to make large calculations and keep track of data was once considered something that would constitute AI (insofar as large calculations were previously a domain exclusive to humans), but today a single cell within an Excel spreadsheet can comprise more computing power than an entire row of back-office personnel. Therefore, such a routine cognitive task is no longer a consideration when discussing the capabilities of AI.
Increasingly, routine cognitive aspects of the traditional digital marketing profession will be most poised for complementation and disruption by AI technology, as services like SEO, SEM, and standard-issue web design and development for small to medium-sized businesses will fall into the domain of neural network software augmenting human capabilities.
Parsing the Jargon
In its most straightforward sense, AI is a collection of techniques through which computers can be programmed to reason similar to how humans reason. Machine learning is a popular AI technique by which an algorithm is trained to constantly improve toward certain outcomes based on large amounts of inputs. A type of machine learning called neural networking is the technology is typically behind the most jarring headlines you’ll come across. In short, today’s technologists apply big data to neural networks, letting the computer decide what’s important and what’s not through a process called deep learning. Generating some truly novel insights about the limits of human understanding and the nature of human reasoning, this is what’s comprising most of the high-level AI buzz these days. At root, it involves processing colossal amounts of data through an algorithm in manner whereby different inputs are automatically calibrated to identify patterns, loosely analogous to how a human brain functions (hence the name, neural networks.)
One of the biggest issues with neural networks involves something known as interpretability, or the ability to deconstruct a neural network’s architecture in order to understand exactly why it is behaving the way it is. Put simply, we still don’t know exactly why neural networks make the decisions they do — we can make assumptions and manually weight different variables differently within the network architecture, but there is no surefire way to accurately dissect certain outcomes; they lack full interpretability. This is, for example, what had many observers attributing AlphaGo’s 2016 defeat of Lee Sedol to creativity. Because we couldn’t understand why AlphaGo’s heuristics, many chalked it up to ostensibly human intuition.
Business, Interpretability, AI Data Lakes and Moats
Currently Google, Amazon, Microsoft and IBM all offer cloud-based AI services. Of these four, Google is the only one so far which has announced its own AI hardware, unveiled at data centers earlier this year. Intel also recently released new AI-based chips, and the new iPhone will feature a custom chip for running native neural networks. All the social platforms are also outwardly signaling big moves toward AI — everything from generative AI plans coming out of Pinterest Labs to emotion-based Newsfeed calibration — while conversational interfaces are becoming smarter and more ubiquitous. With these developments, it’s clear that AI capabilities will soon be an intractable aspect of digital life. And while emerging technology will naturally always bear applications for novel (if not sometimes gimmicky) campaign activations, what’s most significant for business is that the tech titans perceive AI’s primary utility as a horizontal enabling layer for businesses and organizations.
Currently at the forefront of this new environment are companies positioning themselves at this insertion level, leveraging data as a moat — that is, vendors using AI techniques to gather and/or process data such that it affords a novel layer of intelligence, deriving insights by aggregating new forms of business and marketing data. The ingrained nature of this intelligence layer is crucial because it reflects the importance of interpretability: In effect, neural networks lacking interpretability means that companies which are able to aggregate and process large amounts of data via neural networks can build conceivably impenetrable moats around their business. A budding crop of martech companies such as LiftIgniter, Amplero, and the Oracle-acquired (and aptly named) Moat, all derive their core value from these AI-powered data moats (it’s also inconceivable that data moats weren’t a factor in Arthur Saoun’s Marcel push.)
Of course, there are still companies which leverage their own data and neural networks in order to provide custom solutions to known marketing issues, but these activities more closely resemble data mining than data moating. For marketers seeking to get firsthand experience with these technologies, social listening platforms such as Crimson Hexagon are increasingly developing their own visual listening capabilities, influencer reports from vendors such as Influential.co leverage IBM Watson, and indeed chat bots also stand to represent some elements of these technologies. Most significant, however, is the implication that AI can be used to construct data moats which insulate companies from their competitors. This holds big implications for the future of business and the role of data in shaping value, product and the customer experience.
Bear in Mind
Don’t Leave Data on the Table When AI is Table Stakes for Your Competitors
For business, most pressing is the notion of unexplored data being left on the table. Just as technology has disrupted media so violently over the last decade, it will be increasingly important for industries with equity built on broadcast media to leverage data moats in order to protect their brand. It follows that these businesses will also increasingly find themselves in need of novel data in order to compete. Lest they realize vertical or horizontal integration as a viable avenue for success, agile tech-first competitors which are able to build data moats via AI technology (such as Amazon) stand as increasingly formidable threats.
Beware AI-Washing and End-of-Theory Visions
Just as some vendors will simply rebrand their products as AI-centric, others will resort to making bold claims about their capabilities and their long-term vision. In a world enabled by AI, educational vigilance is key not just to parse through AI-washing, but also for understanding the limitations of AI technology. This is because we are continually learning more about the power of AI, and so it follows that speculating about its future utility can be fraught. Coupled with the fact that advances in deep neural networks are primarily driven by top tech talent in industry and academia, marketers are wise to be circumspect about vendors touting AI product based on deep learning. If such companies aren’t also using some combination of proprietary data and neural networks to power a novel intelligence layer, it follows that the value of their product may be limited.
Believe it or not: Humans Are Also Good at Things
As the technology progresses, education will be imperative for marketers, especially as companies like Microsoft are making their frameworks completely open source. Moreover, familiarity with AI-based reasoning will be important as issues such as judgment will become more crucial. Writing in the MIT Sloan Management Review earlier this year, Jeanne Ross explained how human judgment will become increasingly important when working with these new technologies. For instance, when two AI-based platforms are providing two different metrics, which one will you go with? Moreover, when an algorithm’s accuracy is closer to 50%, how will this affect human decision-making?
Final Thought: The Transformation Imperative
AI may not mean the end of marketing, but it does mean that marketers and businesses must internalize the new business playbook centered around data moats and neural network heuristics. On the professional level, we are all charged to strive for constant reinvention. Just because many marketing professionals are relegated to the dominant workplace quadrant, this doesn’t mean we are free to skate by, insulating ourselves from these large-scale changes with our Big Ideas. In fact, the naked potential for AI to disrupt so many industries makes it imperative that marketers understand the root of such dynamism as it marches forward.
To illustrate this point, consider that there are currently more than 100 value chains in the global economy. Coupled with AI, some scenarios of technological advancement posit that there may be as few as 12 value chains by the end of the century, largely due to advances in AI and sector-border dissolution driven by technological disruption. Granted these will be multi-trillion dollar value chains, this vibrancy nonetheless speaks to the extent to which we must orient around professional and organizational reinvention. Indeed, this is one of the key takeaways Yuval Harari puts forward as a result of the rise of AI in Homo Deus.
Because for all it can currently do and all its future applications, it’s important to be clear that there are a bounty of things AI can’t do as well. It cannot be creative in a humanistic sense. It cannot generate Big Ideas. It cannot develop truly innovative and groundbreaking concepts. Such activities will doubtless remain the domain of humans for the foreseeable future. AI can help with these things—if only by freeing us from the immense amounts of wasted cognitive labor which plagues so many professions—but in failing to internalize the changes this technology entails, we’d risk falling victim to an utmost most banal version of technological antagonism.