How risky is AI, really?

It is a decidedly grim view of a technology infiltrating the marketing industry via channels like customer service, search and data analysis. So how worried are marketers exactly?

For her part, Jenna Niven, creative director at ad agency R/GA, noted AI has been around for a long time and is already integrated into processes we use every day, like Gmail sorting spam or Netflix recommending content, but only now are we starting to see AI in the forefront of user experiences like Alexa and self-driving cars, which means it is assuming a more obvious role in our lives.

A misnomer

And fear about AI is rooted at least in part in a misunderstanding of what it is exactly.

“The industry has come to use AI as a catch-all phrase for data warehouse, data science, analytics and machine learning, which, at times, either oversimplifies or exaggerates the general views,” said Sargi Mann, executive vice president of digital strategy for investments at Havas Media, the media division of the global marketing and communications group Havas.

AI is indeed often an umbrella term for concepts like machine learning, deep learning, intelligence, consciousness, sentience, singularity and super intelligence.

But, according to Eran Abramson, head of marketing at AI startup Knowmail, at this point, AI is really just processing power to help us compute more data, at greater speeds and with greater accuracy and insights.

Indeed, marketers like Matt Yorke, chief marketing officer of media company SourceMedia, point to data processing as one of the biggest benefits of AI in marketing today.

‘Fearing a rise of killer robots is like worrying about overpopulation on Mars’

However, Pete Meyers, marketing scientist at SEO software firm Moz, said even though we seem to be on the brink of a golden age of machine learning, which may lead us to think progress will continue to until we achieve singularity, this may be jumping the gun.

“Machine learning in 2017 models very specific tasks. Take self-driving cars, for example. Self-driving cars have genuinely come a long way – and I'm optimistic about their progress, personally – but a self-driving car is a machine that is very good at one task,” Meyers said. “It's a complex task, but the entire ‘intelligence’ of a self-driving vehicle is built around a single context. A self-driving car has many technological advantages – eyes in the back of its head, superhuman reaction times, etc – but it's not an artificial intelligence. The fact that one machine can beat us at Go and another will eventually be a better driver than us doesn't mean those machines are smarter than us. They're just very, very good at one thing.”

In other words, it’s still early days in AI. In fact, many marketers liken AI to children.

That includes Eric Bee, executive producer at R/GA, who said if you left a child unchecked in a room full of things, the odds are the child would make a mess.

“Same thing goes for AI – if you develop a platform that just ingests data and reacts, completely unchecked and without a gentle nudge of cultivation and course-correction from a human being, the odds are exponentially higher that the AI will be ineffective and, ultimately, worthless,” Bee said. “When you start relying on AI to work with data that has very real consequences in its use, such as medical data for patients in a hospital or military applications, the outcome of an unchecked AI becomes very dangerous. We must raise, educate, form and shape AI to be responsible in its decisions like we would expect from how we raise a human being, especially as we rely on them in matters of life and death.”

Niven agreed AI, like a child, is trained with whatever data it receives, pointing to Watson, who was fed the slang dictionary Urban Dictionary and began swearing unabated. But, unlike a child, the team behind Watson could simply wipe Urban Dictionary from its memory to solve the problem.

Other problems aren’t so simple. To wit: Mike King, managing director at digital marketing agency iPullRank, noted a more dangerous and immediate risk than AI killing us all is that AI reflects our biases.

“Whether that's in marketing or simply just in the classification use cases of AI, it reflects the conscious and unconscious biases of the people that build it and verify its results,” King said. “It's an axiom that gender and racial diversity are key things that the tech sector struggles with. In fact, the recent diatribe from the former junior Google engineer is a good indication that more voices and perspectives need to be accounted for in technology. So while AI could definitely spin out of control, draw conclusions and determine patterns that might result in human extinction, it could also potentially classify non-white males incorrectly largely due to the perspectives of other people being underrepresented.”

Industry perception

And yet most marketers are optimistic about AI overall.

According to Adam Jenkins, executive producer at data visualization firm Framestore Labs, it is marketers’ duty to create and elevate the industry through experimentation and discovery, which includes AI.

“The world is too big and too fascinating to stop pushing the limits of what’s possible,” he said. “As our data sets grow and become more complex, I’m bullish on AI’s ability to help harness this complex information to drive brand awareness. With AI, we can bring complex things to life in dynamic, relevant and unimagined ways at an increasing speed.”

John Marshall, chief strategy and innovation officer at creative consultancy Lippincott, too, said marketers should embrace help from AI.

“In the future, brands not knowing their customer will be unacceptable and nearly impossible. With constant access to the world’s knowledge base, we’ll know more than ever, shift whom and how we trust and change decision-making from a personal deliberation to a collaborative and connected feedback loop,” he said. “Marketers will have exponential intelligence and [AI] will help us make smart decisions. Robots and intelligent devices will instantly interpret diverse information sources and deep learning robots will be able to make all sorts of decisions for us on the fly.”

Bee, too, said the industry is cautiously optimistic because of the value AI presents in data science.

“The marketing industry relies on accurate data to make decisions and having toolsets to increase responsiveness or even automate that responsiveness...is a very real, very valuable use case for AI,” he said. “Where it gets foggy is when we start to introduce AI as a consumer-facing platform. For example, chatbots have come a long way, but still react using a set of curated data against structured inputs rather than making a decision on their own. I don’t think any brand is ready to cede complete control of a consumer experience to [AI].”

‘Man-computer symbiosis’

And so, for the time being anyway, a combination of man and machine may be the best way forward.

“We don’t want to distract from progressing this technology to really truly reach the potential to transform our lives from the way we diagnose illnesses to the way that…we communicate…all of these things can potentially massively transform [our lives] in a positive way,” Niven said. “What we don’t want to do is take the Elon Musk perspective…and then not progress it as far as it should because, really, it’s an extreme perspective.”

Benefits and challenges

Antoine Amann, chief executive of Echobox, the company behind Larry, the AI-driven assistant for online publishing, said AI has huge positive potential for marketers in the short- and medium-term because it will gradually take over repetitive tasks so marketers can focus on strategic and creative work, as well as increase the number of client accounts they manage.

Sarim Haque, growth hacker at AI startup Caspy, agreed, saying AIs will play the role of assistants.

“By removing grunt work from our lives, AIs will help us be more productive doing things we actually care about,” he added.

At the same time, Adam Buhler, vice president and group architect of creative technology and innovation at ad agency DigitasLBi, noted that as consumers delegate more of decision-making to increasingly capable digital concierges, the question becomes how marketers pitch these AIs.

“B2B and B2C, move over for M2M, machine-to-machine, requiring a wholly new strategy,” he said. “On the flip side, can you translate your brand experience into an algorithm? This won't be an abstract question.”

Hannah Redmond, group director of strategy and innovation at marketing firm the Marketing Arm, agreed it’s exciting “we are looking at a technology that, through its nature and how makers discuss it, aims to assist humans first and sell things second. This approach poses an interesting challenge to marketers, as it’s not just being involved in AI or jumping on board. Rather, how do we use it to make things for people better?”