Artificial Intelligence and Machine Learning
Total Page:16
File Type:pdf, Size:1020Kb
ISSUE 1 · 2018 TECHNOLOGY TODAY Highlighting Raytheon’s Engineering & Technology Innovations SPOTLIGHT EYE ON TECHNOLOGY SPECIAL INTEREST Artificial Intelligence Mechanical the invention engine Raytheon receives the 10 millionth and Machine Learning Modular Open Systems U.S. Patent in history at raytheon Architectures Discussing industry shifts toward open standards designs A MESSAGE FROM Welcome to the newly formatted Technology Today magazine. MARK E. While the layout has been updated, the content remains focused on critical Raytheon engineering and technology developments. This edition features Raytheon’s advances in Artificial Intelligence RUSSELL and Machine Learning. Commercial applications of AI and ML — including facial recognition technology for mobile phones and social applications, virtual personal assistants, and mapping service applications that predict traffic congestion Technology Today is published by the Office of — are becoming ubiquitous in today’s society. Furthermore, ML design Engineering, Technology and Mission Assurance. tools provide developers the ability to create and test their own ML-based applications without requiring expertise in the underlying complex VICE PRESIDENT mathematics and computer science. Additionally, in its 2018 National Mark E. Russell Defense Strategy, the United States Department of Defense has recognized the importance of AI and ML as an enabler for maintaining CHIEF TECHNOLOGY OFFICER Bill Kiczuk competitive military advantage. MANAGING EDITORS Raytheon understands the importance of these technologies and Tony Pandiscio is applying AI and ML to solutions where they provide benefit to our Tony Curreri customers, such as in areas of predictive equipment maintenance, SENIOR EDITORS language classification of handwriting, and automatic target recognition. Corey Daniels Not only does ML improve Raytheon products, it also can enhance Eve Hofert our business operations and manufacturing efficiencies by identifying DESIGN, PHOTOGRAPHY AND WEB complex patterns in historical data that result in process improvements. TBG VICE PRESIDENT OF This issue of Technology Today highlights some of Raytheon’s innovative Raytheon Advanced Media ENGINEERING , developments and applications of AI and ML. PUBLICATION DISTRIBUTION TECHNOLOGY AND Rose McGovern MISSION ASSURANCE In our Leaders Corner, Raytheon Missile Systems’ Technical Director CONTRIBUTORS Dr. Jeff Vollin answers questions about his role, and what excites him Paul Bailey about technology development at Raytheon. Our Eye on Technology Steve Klepper section highlights our Mechanical, Materials, and Structures technology Tony Marinilli network and how the industry shift toward open standards designs Nora Tgavalekos exerts influence on this technology area. ON THE COVER Artist’s depiction of a deep neural network The Special Interest section highlights Raytheon’s White House visit to celebrate the 10 millionth U.S. Patent, granted to Raytheon engineer Joe Marron. Finally, the People section spotlights Raytheon women engineers encouraging and mentoring young women in high school on Science, Technology, Engineering and Math careers during a FIRST® Robotics Competition in California. Mark E. Russell This document does not contain technology or technical data controlled under either the U.S. International Traffic in Arms Regulations or Thisthe U.S. document Export Administration does not contain Regulations. technology or This document does not contain technology or technical data controlled under either the U.S. technical data controlled under either the U.S. International Traffic in Arms Regulations or the International Traffic in Arms Regulations or the U.S. Export Administration Regulations. U.S. Export Administration Regulations. TECHNOLOGY TODAy 2018 | 1 CONTENTS FEATURE LEADERS CORNER SPECIAL INTEREST 52 Jeff Vollin Talking technology with Jeff Vollin, Technical Director, 4 Raytheon Missile Systems SPOTLIGHT EYE ON TECHNOLOGY Artificial Intelligence 32 54 Mechanical Modular 58 and Machine Open Systems Raytheon Predictive Architectures The Invention Engine Learning Discussing industry shifts Maintenance toward open standards designs Raytheon receives the 10 millionth at Raytheon and their influence on the U.S. Patent in history Predictive Maintenance is as much Mechanical, Materials and Raytheon’s customers are a traditional engineering problem Structures technology domain embracing the intelligent as a data science problem machine era with the knowledge that AI will PATENTS profoundly affect the 36 Strategies for character of war Rapid Prototyping PEOPLE 59 Patents Issued Machine Learning to Raytheon Addressing the challenges of U.S. Patents issued from January rapidly applying machine learning 56 2016 through June 2018 capabilities in dynamic environments 10 14 20 26 of limited training data FEATURE FEATURE FEATURE FEATURE 42 Detection, Extraction Automatic Proactive Machine Machine and Language Target Emerging Learning in Learning for Classification of Recognition Threat the Factory Patterns Handwriting Extracting and classifying Systems Detection Machine learning of Life handwriting of unknown location, can provide From One Robot-Maker ATR is designed to ( ) Raytheon is at the size, color, content, and language PREVENT predictive power enhance the utility forefront of pattern To Another An analytical tool and actionable of military systems of life discovery, that assesses insight to nearly Women engineers mentor by interpreting detection, tracking, unusual patterns or all aspects of 100 robot-makers at the data faster and and prediction behaviors, as they manufacturing FIRST® Robotics Competition more accurately happen, to detect than human threat activities analysis alone before they occur 2 | TECHNOLOGY TODAy 2018 TECHNOLOGY TODAy 2018 | 3 SPOTLIGHT Artificial Intelligence and Machine Learning at Raytheon Raytheon’s customers are embracing the intelligent machine era with the knowledge that AI will profoundly affect the character of war. AI AND Machine Learning 4 | TECHNOLOGY TODAy 2018 TECHNOLOGY TODAy 2018 | 5 SPOTLIGHT AI and Machine Learning automatic target recognition (ATR) trained cases, perhaps even seconds. Similarly, The 2018 National Defense Strategy MODEL TRAINING PREDICTION has listed the military application of OPACITY to identify tanks in a desert environment AI technology is becoming ever more AI and ML as a key element of the OPACITY will likely not correctly recognize tanks in widely accessible, causing the cycle of U.S. military modernization strategy.1 OPACITY a dense forest. The implications are that model obsolescence to further accelerate. Artificial Intelligence (AI) and Machine the technology does not yet provide the The average model lifespan will likely TRAINING DATA Learning (ML) will serve as an enabling human-like, general intelligence AI (or grow ever shorter as the AI/ML space technology of next generation battle MODEL INFERENCE “Strong AI”) that one might see depicted matures and becomes ubiquitous across networks; human-machine collaboration in movies. the battlefield. Beginning with one set of & combat teaming; human-assisted algorithms, once the battle is underway, As AI and ML is introduced into systems, operations; and network-enabled Figure 1: Machine Learning Development Cycle combatants will introduce new algorithms especially those that perform missions autonomous weaponry. Raytheon NEW, UNSEEN DATA while evolving and deprecating the of high consequence, developers and continues to strategically invest in old. The cycle of this “AI arms race” Technological cognification has accelerated users will have to deal with a range of AI and ML technology development will ultimately collapse into second-by- in recent years through the potent challenges, including opacity, perpetual as a key contributor in delivering and second contests of updating and evolving combination of ever more cost effective, upgrades, and Operational Testing and operationalizing cognified capabilities algorithmic models as each side attempts accessible and powerful computing Evaluation (OT&E); Figure 2. to achieve technological overmatch PERPETUAL to counter, nullify and obfuscate each hardware; massive volume and availability for the United States and its allies. UPGRADES Opacity. AI and ML aligns with a other’s capabilities. The key discriminator of data; virtually unlimited scalability of PERPETUAL UPGRADES2.0 common trend across many of today’s in this environment will be automation: AI and ML technologies are reshaping the PERPETUAL software platforms and architectures; UPGRADES advanced technologies, the danger automated model development, testing commercial world as well as the military. 2.0 and AI and ML algorithms and software. of overwhelming complexity. Humans and deployment. Combat-effective AI In fact, commercial industry and academia 2.0 ML is often considered an enabler for now build machines that we cannot capabilities of the future will have the are leading much of the basic AI and ML the broader notion of AI that includes 1.0 fully comprehend and AI and ML ability to continuously update through technology developments. In his book, man-machine teaming and autonomous exacerbates the issue many times over.3 automated mechanisms to out-sense, The Inevitable, futurist and internet pioneer, operations. Unlike traditional rules-based 1.0 1.0 A compounding factor in ML systems out-think and ultimately overmatch the Kevin Kelly, describes the forces and inherent analytics that are explicitly programmed, is