• Email

• AI,
• Computers,
• Robert Hummel,
• Hollan
• Military

Details

Published: Wednesday, 06 March 2024 19:36

Written by Lois Hollan and Robert Hummel

Hits: 367

Lois Hollan, Senior Fellow, Potomac Institute for Policy Studies

Robert Hummel, Chief Scientist, Potomac Institute for Policy Studies

Introduction

Artificial intelligence (AI) is expected to transform the way wars are fought and revolutionize the enterprise of national security. However, it is still unclear how this technology can be successfully leveraged for national security purposes. The problem stems from the ambiguity of the term “intelligence.” Intelligence is generally taken to mean: “the ability to learn or understand or to deal with new or trying situations: reason; also the skilled use of reason.” But current AI systems are “artificial” and neither perform reasoning beyond their training nor adapt to novel situations. The value of AI to national security will be in accessing data to provide relevant, confident, and reliable information to operators, analysts, and commanders in a real and uncertain world. In this article, we examine the kinds of data that AI technology might address, the challenges of exploiting that data, an approach by which AI could enable a new dimension in the recognition of threats, and why we should develop those capabilities now.

Automation technologies are already supplanting human analysis of vast amounts of sensor data to understand “the battlespace.” Techniques have been developed to perform “automated (sometimes assisted) target recognition” (ATR) to identify tanks, other military ground vehicles, aircraft, ships, submarines, and objects of significance to military operations. Exquisite sensor systems have been developed to collect data to feed into recognition systems. Such sensor data supply both human and machine recognition systems, with the latter employing both classical and emerging AI techniques to recognize threats.

Yet, these elaborate systems have failed to adapt to two new realities:

1. A massive amount of timely data is available for public consumption, which is considered unconventional and separate from the capabilities of exquisite sensor systems designed to collect (conventional) battlespace information; and
2. The kinds of items, threats, and events that must be recognized are distinctly different from the artifacts of war that have been modeled and taught to existing recognition systems.

Related to (1), there is a great deal of accessible digital data (such as social media, cell phone data with images or videos, Twitter [now X] content, news commentaries, and search engine queries). These timelier sources dominate traditional intelligence-gathering sources.

Regarding (2), it is important to recognize that the battlespace is increasingly shaped by influence operations, psychological techniques, civilian technologies, and economic and political dynamics. These novel operations elude current recognition systems; can be engaged before, during, and after kinetic conflict; and can replace kinetic warfare. Recognizing propaganda, deep fakes, nefarious ideological intent, and foreign influence has become as important as tracking troop movements or detecting tank convoys.

So maybe we’ve been doing it wrong or, at least, not keeping up with the times. AI may be the panacea, but likely not in the way that we have been expecting.

See Full Article PDF Here

Twitter