i3 | February 14, 2017

Digital Brainpower: Anticipating What You Want

Gary Arlen
Artificial Intelligence

International Data Corp. foresees a $40 billion market for AI by 2020. IBM calls cognitive computing a “$2 trillion opportunity.”

Behind the buzzwords “artificial intelligence” (AI) and “machine learning” are even more buzzwords: cognitive computing, neural networks, deep learning, relationship intelligence, augmented intelligence, voice/image recognition, data science, Bayesian optimization, predictive analysis and a growing lexicon of techno-verbiage to describe the fastemerging tools of the smart ecosystem. Even the term “Jetson” is officially in the rhetoric. (It’s the name of an NVIDIA Corp. embedded hardware developers’ kit).

Beyond this vocabulary are business altering systems quickly creeping into the retail, automotive, entertainment, medical, education and agribusiness world. International Data Corp. foresees a $40 billion market for AI by 2020. IBM calls cognitive computing a “$2 trillion opportunity.”

TechSci Research expects the U.S. AI market will grow at a 75 percent compound annual growth rate during the next five years, fueled by consumer technology devices, drones, self-driving cars and health care research. McKinsey Global Institute estimates that by 2025, AI will generate $6.8 trillion in “knowledge work,” $4.5 trillion in advanced robotics and $1.9 trillion in smart vehicles.

Accenture expects that in less than 20 years, AI will impact the U.S. economy by two percent and will boost labor productivity by up to 37 percent. A General Electric study envisions that machine learning will boost production capacity by 20 percent while lowering material consumption rates by four percent. NVIDIA says deep learning creators are flocking to AI, with the number of graphic processing unit (GPU) developers growing 25-fold in the past two years. It tallies at least 1,500 AI startups now underway.

A Gartner analysis in October focused on “intelligence everywhere,” putting AI and advanced machine learning at the top of its list of technologies that “will be embedded in everything in the digital business of the future.” Gartner emphasized that these capabilities are crucial for the anticipated boom in Internet of Things and digital security.

Giving Intelligent Machines Common Sense

The vast number of AI projects aim to help machines think and respond in human-like ways, with countless levels of instant deliberation. Whether it’s a truly comprehensive virtual personal assistant, an intelligent camera that can alert you to pictures you should shoot, drones and robots for warehouse navigation/product picking or portable medical instruments that can diagnose blood samples instantly, AI devices need to behave sensibly – not just crunch big data.

“Machine learning technology amplifies and extends the reach of big data analytics and can help create an exceptional shopping experience,” explained John O’Rourke, vice president of marketing at Indix, a Seattle Data as a Service provider that focuses on retailing. “Innovative retailers can tap into the power of machine learning algorithms to do things like determine available products from outside vendors or recommend the quantity, price, shelf placement and marketing channel that would reach the right customer in a particular area,” he wrote.

One human-like factor of machine learning is its capability for constant improvement: the more it works, the more it learns for future decisions. And in a connected environment with billions of devices networked together, the devices become more intelligent by sharing experiences. So a personal device can be tuned to an individual’s tastes or a vehicle can leverage information from all the cars that have driven along a road under various conditions.

That’s why “Macy’s On Call,” a mobile Web tool, can respond aurally to customers’ queries about how to find a product in a store or online. Built on the IBM Watson platform, it draws on the experience of previous customers to analyze unstructured data and develop a natural language response. Similarly, North Face has developed a personalized shopping experience, based on Watson’s intelligence, which tracks a customer’s transaction history to deliver relevant, realtime recommendations via a mobile app.

In other words, IBM’s Watson can be far more customized than what it is recognized for: its debut as the victorious know-it-all competitor on a 2011 Jeopardy TV game show. IBM is now aggressively marketing its Watson Developer Cloud, which provides various application program interfaces for language, speech, vision and data development.

Cristene Gonzalez-Wertz, leader of electronics of Cognitive Manufacturing at the IBM Institute for Business Value, points to several ways that the company’s Watson AI platform is being used in manufacturing, product management and security.

“Watson uses machine learning to improve visual inspection, aiding both humans and robots on the production line to prevent defective or damaged devices from getting out of the plant,” she explains. “This works well in high velocity environments, such as manufacturing mobile devices.” She also cites the ability to ensure that products meet regulatory and safety compliance standards on a global scale. On the security front, she notes that IBM tracks 20 billion security events every day.

“We have developed deep knowledge on patterns and approaches to identify and protect our clients from risk at the speed that cognitive enterprises need,” she adds.

Drilling into the integration of AI and IoT, Gonzalez-Wertz points out that Whirlpool is using Watson IoT for connectivity among home appliances to reshape the appliance maker’s understanding of how products are used and for diagnostics. “I don’t think there are any restrictions on where AI can go,” Gonzalez-Wertz says. “Watson can work with people, data and systems. ‘He’ really lives for the data – eats it for breakfast, lunch and dinner at a rate of 800 million pages per second. While assembling that data is critical, integrating into a business solution is where the value is.

“It’s totally different to have usage data aggregated across wmillions of devices,” she explains. “Your ability to detect patterns earlier, even with spotty or incomplete data, improves dramatically. You can respond faster and with a much better answer. At the end of the day, that not only improves operations and the bottom line, it improves customer brand loyalty.”

Next Industrial Revolution

Jen-Hsun Huang, founder, president and CEO of NVIDIA, a technology platform provider, calls AI “the fourth industrial revolution” (after steam, mass production and automation) and expects that “AI will revolutionize every industry” from smart cities and transportation to health care and entertainment.

Huang, who is delivering the opening keynote at CES 2017, cites the central role of the GPU, a specialized electronic circuit to manipulate and alter memory, as the core of “deep neural networks that are trained to recognize patterns from massive amounts of data.”

“Software writes itself and machines learn,” Huang explains. “Soon hundreds of billions of devices will be infused with intelligence.” He points to academic research that has confi rmed, “the larger the network, the more it can learn.” NVIDIA researchers worked with Stanford University experts to “develop a method for training networks.”

Among the skills of deep learning are image and speech recognition, which provide the basics on how to learn, perceive, reason and solve problems. Huang says NVIDIA’s GPU began as an “engine for simulating human imagination,” especially for conjuring virtual worlds for videogames and Hollywood productions.

“Now, it runs deep learning algorithms, simulating human intelligence that can perceive and understand the world,” Huang says.

In AI, algorithms learn from real-world examples. As opposed to computer programming, which is about coding instructions, deep learning is intended to create and train neural networks for data centers, where they can infer, predict and classify new input.

“Networks can also be deployed into intelligent devices like cameras, cars and robots to understand the world,” explains Huang. As new experiences accumulate, the network becomes more intelligent. NVIDIA has created an energy-efficient AI supercomputer called “Jetson TX1” – a credit-card-sized module that can process at 1 TeraFLOP FP16 using 10 watts of energy – for intelligent IoT devices.

The company has also developed “Xavier,” which Huang calls “the most ambitious single-chip computer – the world’s first AI supercomputer chip” with seven billion transistors. It can handle 20 trillion operations per second of deep learning performance, the kind of expertise needed in self-driving cars and other dynamic products.

“From real-world data, computers can learn to recognize patterns too complex, too massive or too subtle for handcrafted software or even humans,” Huang says. He cites the $10 trillion transportation industry and the healthcare sector, where AI can detect diseases such as Parkinson’s, Alzheimer’s and cancer “to understand the human genome” for early treatment.

Obviously, such massive AI projects don’t materialize out of nowhere. For more than a decade, the consumer technology world has seen the onslaught of “recommendation engines” (such as the popular Netflix and Amazon algorithms which make suggestions based on what similar viewers/readers have enjoyed) and voice-activated response systems (such as Apple’s SIRI and Amazon’s Echo), which (sort of) know what you’ll want to do next. Stitch Fix, Spotify, Waze and dozens of other services interpret information about your lifestyle, tastes and preferences – all of which can be blended into a personalized service once a provider wrangles all the “big data” input into a digital “you.”

Even iRobot’s 15-year-old Roomba autonomous home vacuuming device, with an estimated 10 million in service worldwide, has a limited AI capability, able to sense dirt spots, use its “cliff sensor” to avoid stairs and mechanisms so that it doesn’t “miss a spot” as it traverses a rug.

Google’s “Daydream” and the Pixel Handset Assistant software plus Samsung’s recently-acquired Viv open AI platform have put more attention on conversational digital assistants that integrate natural language as an interface to everything else your mobile device knows about you.

“Viv was built with both consumers and developers in mind,” explains Injong Rhee, CTO of Samsung’s Mobile Communications business group. “This dual focus is also what attracted us to Viv as an ideal candidate to integrate with Samsung home appliances, wearables and more, as the paradigm of how we interact with technology shifts to intelligent interfaces and voice control.”

The latest developments anticipate what you’ll do – even before you think about it. For example, you may start considering a new place to live after you get a new job offer. But a networked AI system can start planning as soon as you go job-hunting (monitoring your LinkedIn searches or your calendar – calculating whether the job you seek may lead to a relocation. And that sets off “anticipatory triggers” that you may soon be looking for new appliances, insurance and telecom service. This “early alert” capability may smooth the process and pressure when decision-time comes. It also can set up relationships with potential suppliers of everything you’ll need for the move.

Next Generation Assistants

Wiidii, a hybrid personal assistant, typifies the implementation of machine learning and the growing appeal of personifying the machine.

“The more you use Wiidii, the more he gets to know you and provide you with personalized answers,” explains Cédric Dumas, founder and CEO of the three-year-old Bordeaux, France,company, which is making its first CES appearance at Eureka Park. The software uses machine learning, automation and AI to enable “a real conversation with your personal assistant for the first time,” Dumas adds.

Users can enter queries by voice or text to conduct a “real conversation with their hybrid personal assistant, who provides a relevant and personalized answer,” he says.

Wiidii (“he” as Dumas calls the digital personality) keeps track of personal documents securely, including passport, driver’s license, blood type card, vaccination card and membership cards that may be needed for transactions. Unlike similar systems, Wiidii is a hybrid combination of AI and human intelligence, using qualified concierges.

“Most requests can be handled by the AI,” Dumas says. “However, if the AI cannot answer requests considered as ‘complex’, the human will answer instead. Also, if the user wants an appointment at the doctor’s for instance, the concierge will call the doctor to make the appointment.”

“Thanks to Wiidii’s mix between AI and human know-how, all their requests are carefully handled,” Dumas adds.

Initially Wiidii is focused on travel and daily scheduling activities, including price comparisons and reservations, accessing existing extensive databases. Wiidii does not store credit/debit card information or financial data, which are useful but confidential Dumas says. It also does not keep users’ private medical information other than blood type for emergency use. It encrypts other data.

Dumas is targeting this “B2B2C solution” to travel companies (hotels, airlines, tour operators) and automakers, among others. Air France, Hertz Europe, the Palais des Festivals et des Congrés in Cannes and other venues are using or testing Wiidii.

The growing roster of AI products has triggered concerns about privacy, security and simply letting people think for themselves. Indeed, one hurdle that concerns some is an unwarranted backlash from technophobes and some careful warnings about managing AI development.

“The thing that could hold back technology the most is to make some horrible early AI mistakes,” says Andrew W. Moore, dean of the Carnegie Mellon University School of Computer Science. He warns that being too cautious during experimentation could send a signal to society to give up on AI because of worries about the danger of the technology. Moore says that wouldn’t be smart. Or a philosophical AI device, who has studied DesCartes, may simply intone: “Cogito ergo sum.”

January/February 2017 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.