Artificial Intelligence: Past, Present, Capabilities and Limitations
(Room 320GH)
02 Dec 19
12:45 PM
-
2:15 PM
Tracks:
Full Schedule, Monday Schedule
Many in the political, industrial and defense communities are expecting current artificial intelligence technologies (deep learning and deep neural networks) to solve a wide array of problems. Others are deeply concerned that adversaries investing heavily in these technologies will produce highly autonomous and adaptive weapons that will overmatch any known defenses. This reaction is not surprising given that deep neural networks and deep learning systems have been remarkably successful at tasks long believed to require high levels of (human) intelligence. These technologies are enjoying great success because of two enabling developments. The availability of large amounts of appropriately labeled training data and the continued growth in sheer computing power permit the decades-old neural network technologies to now reach impressive performance levels. These success stories beg answers to questions about the limits of performance and potential.This tutorial describes artificial intelligence in its historical context of boom and bust cycles. The AI discipline has a 60-year record of remarkable achievements fueling heightened expectations that were followed by disillusionment (“AI Winters”) when the technologies failed to satisfy popular expectations or generalize to wider application. The tutorial develops parallels between the current deep neural network requirements for success and those of previous intelligent technologies that were once inspiring but are now less widely used. The tutorial also identifies application areas where deep neural network technologies have been applied, highlighting both successes and limitations to develop, frame and temper expectations. Finally, the tutorial will examine the state-of-the-art in terms of methods and tools for testing AI-enabled autonomous unmanned systems. The tutorial is open to any who would benefit from developing an appreciation of the larger context surrounding current AI achievement. It provides an overview of the field. It is not intended to teach use of available deep learning utilities or to provide detailed information about constructing deep neural networks.