Page 75 - Artificial Intelligence for the Internet of Everything

Intelligent Autonomous Things on the Battlefield 61

A possible approach to developing the necessary capabilities—both
human and AI—is to train a human-agent team in immersive artificial envi-
ronments. This training requires building realistic, intelligent entities in
immersive simulations. Training (for humans) and learning (for agents)
experiences must exhibit a high degree of realism to match operational
demands. Immersive simulations for human training and machine learning
must have physical and sociocultural interactions with high fidelity and a
realistic complexity of the operational environment. These include realistic
behaviors of human actors (friendly warfighters, enemies, noncombatants),
and interactions and teaming with robots and other intelligent agents. In
today’s video games these interactions are limited and not suitable for sim-
ulating the real battlefield. Advances in AI are needed to drive character
behaviors that are truly realistic, diverse, and intelligent.
To this end, some of the cutting-edge efforts in the computer-generation
of realistic virtual characters are moving toward what would be needed to
enable realistic interactions in an artificial immersive battlefield. For exam-
ple, Hollywood studios sought out the army-sponsored Institute for Crea-
tive Technologies (http://ict.usc.edu/) on multiple occasions to create
realistic avatars of actors. These technologies enable film creators to digitally
insert an actor into scenes, even if that actor is unavailable, much older or
younger, or deceased. This is how the actor Paul Walker was able to appear
in “Fast and Furious 7,” even though he died partway into filming (CBS
News, 2017).

3.8 SUMMARY

Intelligent things—networked and teamed with human warfighters—will
be a ubiquitous presence on future battlefields. Their appearances, roles
and functions will be highly diverse. The AI required for such things will
have to be significantly more sophisticated than that provided by today’s
AI and machine-learning technologies. The adversarial—strategically and
not randomly dangerous—nature of the battlefield is a key driver of these
requirements. Complexity of the battlefield, including the complexity of
collaboration with humans, is another major driver. Cyber warfare will
assume a far greater importance, and AI will have to fight cyber adversaries.
Major advances in areas such as adversarial learning and adversarial reasoning
will be required. Simulated immersive environments may help to train
humans and AI.