From 3M Health Information Systems
AI Talk: Smart clothes, AI for AI and AI for IT
Forget wearable devices, these researchers from Purdue have a new vision for the future that is all about smart clothes. These clothes will monitor anything and everything about you. Here is a quote from their abstract: “This work describes an industrially scalable approach to transform conventional fabrics into smart textiles—wirelessly powered by omniphobic silk-based coils (OSCs).” Omniphobic? Turns out this term refers to clothing that is resistant to water and oil. The researchers have developed a fabric that is battery-free, machine-washable, reusable and contains electrophysiological sensors. Apparently, it is made of silk-nanocarbon composite coils that can be charged using magnetic resonance coupling. The engineering and science behind this technology is beyond me, but the researchers demonstrate the use of their fabric with a couple of examples. First, a wristband that uses light to measure blood flow using a technique called photoplethysmography (PPG). They have also developed a non-contact voltage detection glove. Wearable sensors have a lot of different applications, particularly when combined with AI. The future is bright for this line of research.
One source of pure joy in my life is watching my toddler-aged grandsons play. Like energizer bunnies, they constantly run around, exploring their surroundings and experimenting with everything. As I watch them play, I am amazed at the new tricks they learn. How kids learn is an unsolved mystery and constant source of research attention. It has been suggested that the mechanism that kids (and adults) employ is reinforcement learning. This basically means that when an action results in success you learn something—when it fails you also learn something.
But what about the environment that fosters that learning? My grandsons live in a world filled with toys. This MIT Technology Review article focuses on research that creates such environments for robots to learn to navigate. Uber researcher Rui Wang has created an environment called Paired Open-Ended Trailblazer (POET) that simulates an environment of hills, ravines and obstacles that helps a robot (in this case, a stick figure) learn to navigate. It is an endless simulation where the robot slowly learns to overcome obstacles without being told anything—literally. Here is a video that captures the movement of this robot.
Jeff Clune, another researcher who worked with Wang initially and is now part of the Open AI team, has a similar goal: Using AI to bootstrap agents to learn on their own. Clune believes this is the pathway to achieving Artificial General Intelligence (AGI).
This line of reasoning is becoming quite popular. DeepMind researchers have just published a paper in the Artificial Intelligence Journal, titled: “Reward is enough.” Their claim: “..reward is enough to drive behavior that exhibits abilities studied in natural and artificial intelligence, including knowledge, learning, perception, social intelligence, language, generalization and imitation.” An agent given virtually any task learns by simply seeking to maximize rewards. The trick is to give it the right rewards and guardrails to learn. Intelligence is framed as a reinforcement learning problem in a rich variety of environments. Whether these researchers are proven right, only time will tell. In the interim, there are plenty of applications that can benefit from current AI capabilities.
AI for IT
An article in VentureBeat caught my attention this week. It quotes a Gartner report that basically says by 2024 (that is just three years away) roughly 80 percent of tech products will be built and deployed by non-technology professionals. That is indeed a bold statement. The tools, powered by AI, are powering a range of new applications that require little or no coding. It is unlikely the software development profession will become obsolete – but the profession will undoubtedly morph and become powered by increasingly sophisticated tools.
I am always looking for feedback and if you would like me to cover a story, please let me know! Leave me a comment below or ask a question on my blogger profile page.
V. “Juggy” Jagannathan, PhD, is Director of Research for 3M M*Modal and is an AI Evangelist with four decades of experience in AI and Computer Science research.