i3 | January 19, 2018

Thinking About Things That Think

Gary Arlen
Two men, thinking

Businesses incorporate artificial intelligence processes as consumer opportunities grow

Beyond self-driving cars, drones, intuitive home appliances, healthcare devices, dating apps and dozens of emerging products, there’s a new corporate world of artificial intelligence. AI and its allied technologies of machine learning, deep learning, cognitive computing, computer vision, natural language processing and neural networks are reshaping business practices and retail strategies. 

Predictive engines (such as those that generate Netflix viewing recommendations) and conversational interfaces — such as Amazon’s Echo and Google Home — are at the consumer-facing vanguard of AI. Plus, they are also finding enterprise homes. Companies are implementing AI systems for business processes and management decision-making as well as customer relationship management and retail/inventory planning.

AI is driving “the transformation of firms from being product-centric to customer-centric,” says Dr. Peter Fader, a marketing professor at the University of Pennsylvania’s Wharton School and co-director of the Wharton Customer Analytics Initiative. “As we let the computer make these decisions for us, it gives us more capability to do things at a customer level and to let algorithms drive that decision making in a way that humans are not very comfortable doing right now.”

Fader is also a co-founder of Zodiac Inc., a New York B2B software platform provider that uses customer analytics and predictive behavioral models for marketing and management processes. Zodiac works with traditional and ecommerce retailers. “We ingest historical transaction information and tell companies when customers will return, how often they will return, and how much they will spend.” Fader explains the predictions are “at the individual customer level, not at a segment or aggregate level. 

“With this degree of granularity, businesses can make better decisions when it comes to budgeting, acquisition and retention,” he says. Zodiac’s “Customer Lifetime Value” (CLV) calculates the net value of the relationship a business has with a customer. Its AI features enable a retailer to determine how an individual shopper will remain loyal compared to what Fader calls the 60 percent-plus customers who are one-time buyers. “The core model for our predictions requires nothing more than RFM (recency, frequency, monetary value),” Fader says.

Joshua Montgomery, CEO of Mycroft AI, believes, “Voice assistants are the purest form of artificial intelligence.” His company  foresees AI opportunities in both the enterprise and consumer markets. Montgomery characterizes its product as “the world’s first open source assistant,” able to run on a desktop computer, inside an automobile or on a Raspberry Pi. 

“AI is going to take over any applications where you have a human translating for a computer,” Montgomery says, citing call centers, retail stores and the automotive industry — “anywhere there are repetitive actions.” He expects that as applications evolve, “We’ll get extra features, including “virtual assistants capable of interacting with each other.” In the process, there will be several silos, he says.

“We’re building a full AI [system] that interacts exactly like a person,” he explains. “There are a lot of deep learning technologies behind it.” To accelerate development, Mycroft has plunged into the hardware sector, building a stand-alone device with an anthropomorphic base. The Mark 1 device is now available, and Mark 2 with advanced (but not yet disclosed) features will be out in late 2018.

“It’s extremely early in the AI market,” Montgomery agrees. “The types of assistance will change radically in the coming decade.”

Digging into the Numbers

Corporate enthusiasm for AI has spurred a spate of research. A September Capgemini study, Turning AI into Concrete Value, found that 79 percent of senior executives worldwide agreed that AI will bring “new insights and better data analysis.” Similar levels said that AI would make “our organization more creative” (74 percent) and “make better management decisions” (71 percent).

Artificial Intelligence for Marketers 2018: Finding Value Beyond the Hype, an eMarketer/emarsys research report in October, cited the “explosion of internet-connected devices collecting and sharing various types of structured and unstructured data including text, speech, images and videos” as a critical factor in training AI systems. It identified current and near-term AI applications in marketing intelligence, lead generation and customer acquisition, marketing optimization, customer experience management, brand building and “content creation and dynamic creative.” The latter includes automated writing and image video production that “creates specific content for targeted audiences based on learning algorithms.”

A McKinsey & Co. analysis found that global corporate investment in AI projects totaled about $12 billion in 2016 — a figure that is growing. The McKinsey study listed natural language processing, natural language generation, speech recognition, machine learning (including deep learning), decision management, virtual agents (including chatbots and digital virtual assistants), robotic process automation and computer vision as the top categories for AI efforts. 

 AI at Home, On the Road and Everywhere

As the business opportunities for AI take shape, consumer AI products are proliferating — most visibly as voice response systems in the home and in smartphones and smart cars. AI is also at work in daily digital activities, from Netflix recommendations to matchmaking and dating sites. Spotify uses a combination of AI tools to create its personalized recommendations. Its Discover Weekly feature uses collaborative filtering, natural language processing and audio recognition, which analyzes the raw audio tracks of new music. 

Olivier Malafronte, founder/CEO of PocketConfidant AI, a French developer, sees AI’s value on several levels. “Personalization will be ubiquitous and show up in ways we cannot yet imagine,” he says, adding that “customer experience will be the focus of all innovation.”

The PocketConfidant AI system offers “a conversational system to support individuals as they navigate personal, professional or academic transitions, helping them develop the competencies they need to enhance their critical thinking and problemsolving skills, become more resilient and self-confident.” He says PocketConfidant can “respond with empathy and smart questions that get to the heart of the challenge the user is facing to help them find their own solutions. AI will become a complement of human beings, not a replacement,” Malafronte says. 

Knowmail, based in Woburn, MA, uses the phrase “We’ve heard you hate email, so we fixed it!” to describe its AI focus. The venture capital-backed startup builds personalized and anonymized user profiles through data collected from communications services and device sensors to identify a customer’s communication patterns, social preferences, timetables, favorites (people/topics), narratives and preferred actions. From this data, it prepares a prediction about each email or message’s priority, urgency, estimated handling time and next-best-action. 

CEO/Founder Haim Senior says Knowmail combines cognitive analytics and advanced human-to-computer interfaces, which are being used in “several joint developments with major tech companies to further improve our service and offering to users and businesses.” He says Knowmail softwareas-a-service (SaaS) reaches customers at more than 4,000 companies. Its initial market is firms with more than 200 employees, where it is used “to increase productivity, save time and improve engagement.”

The three-year-old company uses “cognitive research, neural network and online learning methods to compose our personalized artificial intelligence (PAI) engine,” Senior explains. He lists “trust, privacy and loyalty” as critical challenges for AI to succeed. 

At Deepmind, an AI company that Google acquired in 2014, one of the key projects is Deepmind Health, which is working with medical clinicians to provide more accurate analyses and enable patients to get faster treatment. In initial trials, nurses said that the AI functions are saving them more than two hours per day, meaning they can spend more time with other patients. One goal of the AI process is to empower patients to care for themselves and their families’ health. 

In a similar vein, GetWell.ai, a stealth-mode startup, is using speech recognition and natural language in its AI venture, aimed at helping people monitor and treat chronic health conditions. And IBM’s Watson AI system took just 10 minutes to analyze a brain cancer patient’s genome and suggest a treatment plan compared to 160 hours for human experts to make a comparable plan. 

An AI system is being used to fight counterfeiting, initially focused on the $900 billion luxury goods sector. Entrupy is a hardware-enabled SaaS company that uses computer vision algorithms and microscopy to identify counterfeit luxury products from brands such as Louis Vuitton, Chanel and Hermès. The two-year-old company has authenticated $14 million worth of goods using a deep learning-driven database with millions of micro-scopic images. An Entrupy device takes microscopic photographs of different areas of a high-value item and runs them through a computer that uses proprietary algorithms. It claims authentication accuracy above 96.4 percent, valuable to merchants and resellers of luxury products.

The proliferating efforts to use AI tools are reaching into more industries and the daily lives of more individuals. Software developer SAS is putting visual data mining, machine learning and visual text analytics into its SAS Platform and SAS Viya products, which embed AI capabilities into products used by its customers.


For example, SciSports finds football stars via AI. The Netherlands-based company uses SAS Viya to identify the influence of individual players on team results, track player development, determine potential market value for a player and predict game results. It’s a codified version of the human expertise of the Moneyball book and movie — but useable on a much wider scale. The AI process is used in conjunction with SciSport’s BallJames camera system for object detection to access player movements through 3D images. The company says that the system can identify rising stars or undervalued players compared to others in their league.  

  At the Massachusetts Institute of Technology’s Media Lab, an AI program called “Shelley” (a paean to the author of Frankenstein) recently wrote a horror story that was deemed sufficiently scary. The Lab’s deep learning algorithm read many terror stories and trained itself to write horror fiction. 

Finding AI Experts

These and dozens of other business and consumer projects illustrate the potential range of AI projects. While customer experiences with AI are proliferating, thanks to Alexa, Google Home, drones and self-driving cars, experts agree that the AI juggernaut is just beginning. Aside from the inevitable debates about privacy and employment impact, AI is facing another rollout challenge. 

"The biggest barriers to the successful and wide adoption of AI will be finding talent who know how to build these algorithms and managers who know how to ask the right questions,” says Zodiac’s Fader, who cites computers’ ability to make more accurate diagnostic decisions. “We’ve seen a reluctance with some businesses to trust the algorithms and the data in the face of lots of ‘noise’ [about] changes in the economy and new competitors in the landscape.” 

Meanwhile experts are upbeat that AI is entering an era of explosive growth, which will trigger even more innovations. For example, Intel is deeply involved in “neuromorphic computing,” which will mimic the way the human brain works rather than rely on transistor-based circuits that process data. In addition to faster processing, Intel Circuit Research Labs expects that neuromorphic designs can cut energy consumption by up to 300 times compared to today’s circuits.

Although there’s no timetable yet, it’s easy to imagine how such digital brainpower will be used in wearables, sensors and the Internet of Things without requiring continual network connections. Think about that.

 AI at CES 2018

The AI opportunity is certainly big enough for all kinds of aspirations, from smart cars, robots and drones to medical diagnostics and sophisticated personal digital assistants. In CTA’s member roster, more than two dozen young companies identify themselves as “artificial intelligence” providers. They have plunged  into a field where venerated technology  firms such as Intel, NVIDIA, IBM and Google plus countless academic institutions are  accelerating their work. 

 At CES 2018, the new Artificial Intelligence Marketplace showcased about a dozen companies, presenting products ranging from automotive to smart home to smart cities applications. AI Marketplace exhibitors range from NovuMind, which is working on AI projects and heterogeneous computing  for the IoT and what it calls “the Intelligent Internet of Things (I²oT),” to Backyard Brains, which makes neuroscience experiment kits for students (elementary to grad school) who want a hands-on education in electrophysiology. 

 Another three dozen exhibitors are in the Robotics Marketplace, with an expanding array of personal, education, healthcare and recreational robots; the Drones Marketplace is home to about 50 unmanned aerial vehicles.


In the rapidly evolving AI ecosystem, new terms pop up for technologies  and products. Here are some of the fundamental factors.

ARTIFICIAL INTELLIGENCE: Software and systems  that perceive an environment and take actions that mimic cognitive functions of human minds, such as learning and problem solving. Sometimes called “cognitive computing.” 

COMPUTER VISION  OR MACHINE VISION: Emulating the human visual system to view and interpret digital images; it includes image processing, pattern/ facial recognition and image understanding.

DEEP LEARNING: A branch of machine learning for building and training neural networks, allowing network layers to find patterns in the output of higher layers. 

MACHINE LEARNING: Systems that train algorithms to perform tasks by learning from previous data and examples rather than explicit commands programmed by humans. 

NEURAL NETWORKS: Machine learning algorithms and computational models designed to function like neurons in the human brain. They progressively “learn” from data without being explicitly programmed.

PREDICTIVE ANALYTICS: Programs that combine  techniques from data science, statistics and AI to analyze  sets of structured and unstructured data, uncover patterns and relationships, and use them to make predictions about probable future outcomes and events and identify risks and opportunities. 

PRESCRIPTIVE ANALYTICS:  Produce actionable data  and a feedback system that  tracks outcomes. 

RECOMMENDATION ENGINES OR RECOMMENDER SYSTEMS: AI-driven information filtering systems that predict user preferences and responses to queries based on past behavior, relationships to other users, similarity among items being compared and context.

VOICE-ENABLED DIGITAL ASSISTANTS: (or Intelligent Agents, Virtual Personal Assistants, Automated Assistants  or Virtual Agents) Programs that  can organize, store and output information that helps them conduct a conversation with  a human. They answer voice queries with information from  a multitude of online sources. Examples include Apple’s Siri, Google Now, Amazon’s Alexa  and Microsoft’s Cortana.

January/February 2018 i3 Cover Issue

Subscribe to i3 Magazine

i3, the flagship magazine from the Consumer Technology Association (CTA)®, focuses on innovation in technology, policy and business as well as the entrepreneurs, industry leaders and startups that grow the consumer technology industry. Subscriptions to i3 are available free to qualified participants in the consumer electronics industry.