Harnessing AI - NPS Online
In response to the Department of the Navy’s strategic concerns around AI, and for advanced AI development, the Naval Postgraduate School (NPS) is integrating the ongoing efforts of researchers and students involved in over 100 current projects with a unique consortium focused solely on AI applications. Through the university’s education and thesis research initiatives, NPS students will be primary agents translating AI into naval operations, bringing new technology into their respective fields of study, promoting trustworthy AI technologies and instructing others on their use. Through partnerships on NPS current projects and with other institutions, NPS students and researchers will advance discovery and establish new channels and methods that influence the rapid adoption of AI technologies.
Click here for more information about the Harnessing AI series.
06 Jun 2023
Spring quarter pre-graduation awards ceremony
09 Jun 2023
Spring quarter last day of classes
13 Jun 2023
Spring quarter final examinations begin
What is AI? Goals, History, Machines, Trustworthiness
Having made steady progress since its founding in the 1950s, the AI field has now accelerated to the forefront of computer science, propelled by super-fast microprocessors, availability of big data, and our supreme appetite for automating human cognitive tasks.
Automata are machines that can do jobs faster and better than humans. Traditional automata may have adaptive feedback, but they do not learn new functions. AI is transforming automation by providing automata that do learn new functions.
Rule-Based Deductive AI
Expert systems were among the first machines aimed at learning the skills of experts. They encode rules described by experts into logic formulas in a database so that a machine can follow the same logic and solve the same problems. However, because human experts take actions not describable as rules, no software expert system has become an expert.
Supervised Learning AI
Recognizing faces in images is a human function that we do not know how to describe with rules. With a database of 100 million labeled images, we can teach an artificial neural network to name the faces when shown the images. Unfortunately, these networks cannot explain what they do and are very sensitive to small, pixel-level changes in images.
Unsupervised Learning AI
Large, quality data sets sufficient to train neural networks are difficult to find or expensive to gather. We have figured out how to get machines to become grandmasters at Chess or Go by playing millions of games against each other and rewarding themselves for winning. No external supervision of learning is needed. Can this translate to other AI?
Human-Machine Interaction AI
When IBM Blue beat him in 1997, Chess grandmaster Kasparov invented a new kind of chess played by human-computer teams. The teams beat the best machines. Finding ways to use machines to augment rather than replace human intelligence is a central question in AI.
Machines that can carry on intelligent conversations, think, understand, create, care, be self aware, or be sentient are well beyond our current understanding. Such aspirations have inspired perseverance in the search for intelligent machines. Our best near-term bet is to concentrate on machines that team with humans, augmenting and amplifying us.
Security and Adversarial AI
By using biometrics and monitoring user actions for deviations from profiles of authorized users, AI has enhanced cyber security. But new AI tools have vulnerabilities we do not yet understand, such as the neural network that mistakes a stop sign for a speed limit sign when just a few pixels of the stop sign image are changed. Will adversarial AI get so good that no one can rely on AI tools for military operations?
Data Science and AI
Statistical inference tools that find patterns and trends in large data sets have come extremely useful. Making sense of the vast troves of sensor data from our ever-growing network of Internet "things" requires enormous computing power. Working synergistically, data science provides the methods to improve quality of training data and new algorithms to help AI learn from the data.
Management of AI Projects
Traditional models for producing software systems work relatively well when the requirements on the software are stable. When the system incorporates AI, the requirements can be subject to rapid change because of the central role of data in these systems. In this case it is essential to have a strong team whose expertise covers everything that comes together in the software, now including AI. It is a myth that AI project teams require strong managers but not strong AI team members.
Computer Vision and AI
Automatically identifying persons and objects in images has been a long quest back to the 1950s. Much has been learned about feature extraction – patterns of edges and areas that group into identifiable objects. That is being combined with the "convolutional neural network" to provide sophisticated image recognition.
Development Operations Cycle for AI
The biggest challenge for large software system development is that the systems are developed by multiple persons to service multiple users over multiple versions. In recent years software developers have turned to “agile” teams that can adapt rapidly to changing requirements. This process, called DevOps for development-operations, has been depicted as a constantly cycling figure-8 loop linking developers and users. The DevOps process has been adapted for software that depends on machine learning.
Logistics and AI
Moving materials from suppliers to recipients is long standing concern in all sectors, civilian, government, and military. Military supply networks are in constant flux and are under constant threat. AI planning tools are finding their way into logistics, where they help plan routes, specify stockpiles and depots, and project where supplies will be need in the near future. They are becoming indispensable for modern supply networks, which are in constant change and evolution.
Natural Language and AI
Finding ways that computers can understand text and speech and translate between languages has been a long quest in computing. Progress has been slow but steady. We now have limited AI tools such as Alexa and Siri as our interface to the computer, language translators, dictation recorders, and real-time voice translation between languages.
Robotics and AI
Robots have fascinated human beings since well before the electronic computing age. You may remember the famous story of the Mechanical Turk, a robot invented in 1770 to play chess. (And it was pretty good.) Eventually it was exposed as a hoax – there was a human chess player hiding inside the cabinet. But even so, it inspired interest in the question of whether a chess-playing robot could be built. It took two hundred years to reach a positive answer to that question.
Ethics and AI
AI has raised numerous ethical dilemmas. Who is responsible when an AI machine fails and causes harm? Should weapons systems be allowed to on full automatic, making kill decisions without consulting with a human? Is the programmer of an AI tool used in a military operation a combatant? Is it wise to search for intelligent machines without knowing that we can control them?
AI, Strategy, and Power
Military strategy is concerned with setting guidelines within the context of technology and geopolitics that minimize the chance of war but maximize the chance of winning should war occur. AI has assumed a central role in the context of great powers competition. What kind of AI would be most useful? How should the US, whose political system encourages cooperation between government and industry, compete with China and Russia, whose systems mandate alignment between government and industry?
AI in the DOD
Motivated by a concern to deal realistically with newly emerging great powers competition, the US DOD has issued various strategy documents for AI research, implementation, and industry cooperation. What offices and programs has DOD set up to accelerate adoption of AI technologies?
Risks of AI systems
AI can be alluringly attractive because it can do certain jobs well beyond the capabilities of humans. It can also be alarmingly unattractive because it can make serious mistakes so fast that no human can intervene. Where are the risks in the kinds of AI we have been discussing in this course?
Bias and Trust in AI
The nature of geopolitical conflict today is to engage in actions that gain ground for the aggressor but fall short of starting a war. The decision systems to plan and execute actions depend increasingly on AI. AI systems can make serious mistakes if attacked by an adversary, if trained on data containing unseen biases, or if dependent on fragile AI technology. Trustworthy AI is crucial not only for success but to avoid mistakes that precipitate war.
The Next Face of Battle
In 1979 John Keegan published a book, The Face of Battle, in which he analyzed the practical mechanics of battle and how they affect outcomes as much or more than strategy. AI is radically transforming the space of possibilities available to commanders and warfighters. It emphasizes small, networked, distributed, swarming, autonomous agents over large platforms. AI also adds a new dimension, where the agents have their own decision-making authority.
This course has sought to cut through AI hype by exposing the base principles of AI. Although we currently have no intelligent machines, we do have six varieties of machines that can rapidly learn to do complex human tasks. These machines have produced significant advances and vulnerabilities in important domains such as vision, robotics, natural language processing, and cyber security. The quest to make these machines reliable and secure has unearthed a host of dilemmas that implementers must face.