Our robot overlords are not all they cracked up to be, but they are everywhere. At least one recent ad campaign for Toyota was co-written by a computer and another for The Natori Co. used artificial intelligence to target paid social media to sell bras; but for the most part, AI is becoming the most high-tech intern at the office: sorting photos and studying data.
Still, this doesn’t mean artificial intelligence is doing menial labor. Many of the tasks being handled by AI tools such as machine learning, deep learning and natural language processing are powering marketers to work faster and more effectively that ever before, building personalized experiences in real time.
“In 50 years from now, artificial intelligence is going to be a part of the DNA of technology,” said Felicity Carson, CMO of IBM Watson Customer Engagement.
Carson noted that Performance Bicycle uses AI to identify best customer experiences and connect them online and offline; doing that increased sales by 11%. AI can also identify what needs fixing, she said, noting Pay Pal uses AI marketing apps to detect customers struggling at key moments of their journey, such as the checkout; by fixing its checkout, it saved $3 million in sales and $6 million in revenue.
AI is doing far more than housekeeping in the user experience; it is helping everywhere, from speeding cancer drugs to market, to making better wines. Carson noted that wine growers in California had used AI to detect pests and choose where to plan their vines, which led to a 25% reduction in water consumption and a better product.
“Virtually every size of organization in every industry is starting to use machine learning,” said Matt Wood, general manager, deep learning and AI at Amazon Web Services.
Tens of thousands of marketers are now using machine learning to sort out data in order to make better uses of it, noted Wood. GE Healthcare uses it to drive better knowledge from medical images, looking inside chest X-rays and head CTs; the NFL runs all its player statistics, including telemetry from players and videos to predict plays and understand how plays interact; and “every single swipe and every single connection at Tinder is driven under the hood by a machine learning model,” he said.
In one case, Moody’s worked with AWS to apply machine learning to the content of the thousands of financial reports it had on file in order to make a better analysis of the data in them. “Financial services is a goldmine of printed and scanned documents,” and machine learning can extract the data in real-time, Wood explained.
Trending in the Hype Cycle
AI is the next step in software’s evolution, Carson told the recent WP Engine Summit. The technology is becoming the way to get the information marketers need to perform better, she said.
“It can search through oceans of data, petabytes,” she said. “But more important than that, (it) can recognize and learn patterns and present it back in a meaningful and digestible way.” AI also has visual perception, speech recognition, and even decision-making abilities to aid marketers, Carson noted.
“Marketers are being asked to land on the moon,” said Carson. They face an excess of data and too many tools for content, campaign management, and other functions to manage at once. They are also facing rising customer expectations and customers reluctant to share their data, Carson said; but they can “rise above the chaos with AI-powered marketing.”
Even as the CMO’s budgets have started falling back after a period of growth led by technology, they are still finding ways to add to their AI spending and baking it into the customer experience. Marketers were early and enthusiastic adopters of software-as-service and cloud computing to expand their technology use within their budget limitations and are doing the same with AI. Gartner identified “Democratized AI” as as one of their top 5 trends in this year’s “Hype Cycle” and marketing departments have been big trendsetters on that cycle.
Machine learning is still a relatively young technology, but it “is experiencing something of a renaissance” thanks to the rise of cloud computing, said Wood. The discipline had been around a while, but it was out of reach to all but the most sophisticated organizations until all sizes of companies were able to access cloud computing. Developers of AI apps now can store as much data as needed to make their computer models work and use as much computing power as they need on demand; this allows them to build on the many APIs made available as open-source tools.
Speaking to an enthusiastic crowd at a recent AWS developers summit, Wood said the growth of machine learning developers working on Amazon’s platform alone has increased 250% year over year. “Machine learning has truly arrived for every developer,” he explained.
Stupid math vs. more thought
Some uses of machine learning will lead products to market faster. Wood noted pharmaceutical company Celgene, which specializes in cancer drugs, is now using machine learning to predict the toxicity of drugs in development, “dramatically accelerating the path to market,” he said. Snapchat is using deep learning to parse the images being posted by the hundreds of thousands on their platform: “They’re able to look inside and identify the relationship of objects, people and places inside those photos,” said Wood.
AI enables data to flow freely across the marketing ecosystem, Carson explained. It creates a personalized experience across multiple devices and channels and allows marketers to work smarter.
On the creative side, AI has proved to be more Wall-E than The Terminator, assisting rather than replacing. A recent survey found almost two-thirds of enterprise marketers expect to use AI in their content marketing strategy this year, but it also found they plan to use it mostly to personalize messages, not to create content. That is still the province of humans, insiders agree.
Integrating neuroscience and advances in quantum computing offer promise for expanding AI in the future, by fine-tuning the data sets that feed AI and freeing it from the binary if/then algorithms and allowing it more free rein. But those integrations are still a few years away from popular use, especially quantum computers. Experts say their mainstream adoption isn’t still almost a decade away.
In the short term, “this convergence of magical technologies” still needs to be policed by humans, said Anthony Scriffignano, senior VP and chief data scientist at Dun & Bradstreet.
Massive connectivity can render data sets useless and increase the risk of making wrong calls from them, he told an audience of digital officers. He singled out the case of investors using AI to pick stocks, who wiped out incremental returns because they all mined similar data pools.
“What we need to protect the machines from the stupid math is more thought,” he said.