|
Get these Alerts by e-mail -- FREE for a limited time  |
||||
|
September 30, 2015
Greetings: In this issue: Coming advances in printed electronics. Faster chips for mobile devices. An interview with Alex Nugent, founder and CEO of Knowm Inc. Thinner, More Flexible ElectronicsPrinted electronics technology -- imprinting electrical components on a wide variety of thin substrates -- is poised for big advances. The relatively inexpensive manufacturing process will quickly find uses in a variety of applications, from product labels and medical tracking devices to smart fabrics and robotic sensors. “The printed electronics industry is going to see a lot of attention and growth" in coming years, says Jennifer Ernst, chief strategy officer at Thin Film Electronics ASA and chair of the governing board for FlexTech Alliance. One of the first products stemming from printed electronics technology: Electronic labels that will help combat counterfeit merchandise. Introduced by Xerox and developed by Thin Film, the labels are more secure than bar codes and cost less than radio frequency identification chips. They consist of thin, flexible memory chips that can be attached to product labels and then read by a smartphone or other handheld device. The labels can be updated with new data as a product moves through the supply chain. Makers of high-end goods with high profit margins will be first to adopt electronic labels. Think electronics, pharmaceuticals and liquor. They're the sorts of pricey products that counterfeiters often target -- a growing concern as Internet sales rise. (Last year alone, Uncle Sam confiscated $1.2 billion in phony goods.) Eventually, electronic labels will become ubiquitous, used even on low-cost items. Other types of electronic labels will be used to relay information to the makers of products bearing such labels. The producer of a bottle of gin can be notified when one of its bottles is opened, for example, providing data previously impossible to access. The technology will track other cues, such as temperature readings for food or pharmaceuticals, letting consumers know whether products have been subject to spoilage. Among makers of printed electronics: GSI Technologies, based in Illinois; Quest Integrated, based in Washington; and Cambrios Technologies, based in California. Other well-known firms working on printed electronics include Corning, Boeing and DuPont. These and other players are betting on cheaper, better ways to add printed electronics to billions of products, many of which are increasingly being connected to each other via the Internet of Things. Expect to see printed electronics manufacturing facilities sprout in many parts of the U.S. where demand will be greatest. According to Ernst, a printed electronics plant costs about $20 million to $30 million to get up and running, far less than it takes to put up a semiconductor manufacturing facility. The latter can cost well over $10 billion to build. Mobile ChipsAlso coming soon: Liquid cooling systems to enable faster chips for mobile devices -- a boon to users of such devices who increasingly expect their smartphones and tablets to handle complicated tasks, such as editing high-definition video, virtual reality gaming and 3-D movies. The computing power needed for those jobs can overheat a device's chip, leading to drained batteries and frozen programs. “Mobile systems have a unique challenge in dissipating heat," says Saibal Mukhopadhyay, associate professor of electrical and computer engineering at the Georgia Institute of Technology. Mobile gadgets must be thin, cheap and energy efficient, an incredibly challenging trio of constraints. Chip makers think they can keep processors cool by including microscopic channels for liquid coolant on future chips. That would boost both performance and battery life as well as extend the life of the device. The costs of production are high now. But ways to bring the expense down are emerging. Mukhopadhyay and a team of researchers at Georgia Tech have a promising design in the works that performed well on a Qualcomm-made chip. And Japanese chip maker Fujitsu wants to start selling a liquid-cooled chip in 2017. Expect Intel, Qualcomm, Samsung, ARM and other tech firms to be major buyers of successful cooling tech. Another step will involve designing software that can intelligently use the cooling systems only as needed. In addition to smartphones, other electronics that will benefit from the cooling technology include drones and robots. Interview with Alex Nugent, Founder and CEO, Knowm Inc.Alex Nugent is an inventor who is building “physically adaptive neuromorphic processors" -- in other words, brain-inspired computer chips. He founded his company in 2002 and has worked extensively on memristors, a groundbreaking element of integrated circuits that retains data even without power. From 2008 to 2011, Nugent supported the Defense Advanced Research Projects Agency's Systems of Neuromorphic Adaptive Plastic Scalable Electronics (SyNAPSE) and Physical Intelligence programs. Knowm Inc. has released hardware and software to developers of artificial intelligence systems for testing. We spoke to him about the future of computer chip design; what's next for deep learning, a branch of artificial intelligence in which computers learn from data; and the race to commercialize brainlike computer chips. Q: What's the problem with today's computer chips? Nugent: What we're seeing now is the maturation of digital computing and Moore's law. Back in 2001, when I started, that was when Intel was going to its dual core processor. Up until then, things were just getting faster and faster, and then they hit up against physics. Now we're hitting up against atoms, literally. The transistors are getting so small that the defects start to manifest, and they're costing more for every generation. Where do you go from there? Look at the brain. It's a highly distributed, parallel, very low power, adaptive learning system, which is almost the exact opposite of a high-speed microprocessor. Q: What could a computer chip designed to work like the brain be used for? Nugent: Computers solve all the problems that humans don't solve very well. Spreadsheets were the first killer app. Things that humans just don't do very well, like precise calculations, computers are amazing at. What's left? The things the brain does well but computers don't: Learning systems that perform inferences over data that they've never seen before in order to solve problems across a lot of different domains, such as speech processing, big data analysis and autonomous robotics. Q: What is the market potential for these new types of chips? Nugent: There are a lot of approaches out there for neuromorphic computing. The first one to nail something that can be deployed on a wide range of problems is going to take off. Eventually you are going to find these things in all computing systems. More of these technology companies are investing money into alternate approaches because we're running up against the wall. Q: When you hear about breakthroughs from Facebook, Google and others in the field of deep learning, you don't often hear about electrical power. Is energy efficiency one of the huge hurdles? Nugent: Oh absolutely. Their deep learning systems are still a factor of a billion less efficient than human brains. There's a place where you should just have devices that work like a neuron's synapses. And when you make that leap, you end up with another 10,000-fold reduction in energy. That's where we're going. Q: Where does Knowm stand today? Nugent: From a technology point of view, we appear to have everything that we need for this to work. Because it's compatible with today's semiconductor manufacturing process, there's not a whole lot stopping us from going from demonstration circuits to fully integrated systems. At this point, we're in a race with everybody else. Sincerely, View the archive of past Alerts issues
Copyright © 2015, The Kiplinger Washington Editors, Inc. |
||||
