The Department of Defense is deeply interested in the man-machine combination, but when it comes to killing, the decision will always be made by the human component. This was stated by Air Force General Paul Selva, vice president of the Joint Chiefs of Staff during a meeting at the Center for Strategic and International Studies, in Washington. His statements have been around the world.
Although there is nothing classified (the timeline communicated would be less than that actually achieved by the military), Selva has in any case drawn attention to the theoretical and ethical approach of the Department of Defense on the development of autonomous weapon.
"Scientists estimate that within the next ten years, we will be able to develop an autonomous weapon, able to decide who or what to destroy. The United States has already decided that they will never fight that way and that man, when it comes to Warfighting, will always be part of the equation. The pitfalls that lie behind autonomous technology and artificial intelligence are many, as are the legal and ethical implications. We are military and in war we are forced to kill the enemy. I don't hide, it's what we do. The methods we use, within that context, are always regulated by law and convention. We are constantly confronted to determine the impact of the tools we are developing, but man must always be decisive in the final decision in automated systems. It is a line that we will not cross ”.
Selva, however, is aware that the line could be surpassed by other countries. "We need rules to govern the war run by AI, but I am aware that not all world powers will respect it. Before even building a Terminator, we should write laws that will bind all the countries in the world. No one will have to build totally self-conscious machines, but I'm sure this won't happen. We need to understand what the limit is, before creating something like a Terminator physical or IT. Man must always be included in the decision-making cycle ". Yet a few hours ago, in a study by the prestigious Defense Science Board, the conclusions expressed are almost antipodal to those of General Selva. "The Department of Defense must accelerate the development of automated systems, both for their military value and for maintaining technological advantage over their opponents. Machines and computers that can process much more data than humans can, will allow the United States to disrupt an opponent's asset. This is why it is vital for the United States to maintain a military advantage. " And if "Skynet" became self-conscious? "Sooner or later something will go wrong. We must be able to allow human teams and not to intervene, correct or stop actions in a timely and appropriate manner. The machine must be verifiable and able to preserve and communicate an immutable and understandable reasoning behind its decisions, in support of the actions performed ".
In the Defense Science Board study, we advise the Department of Defense, of "carry out a series of experiments on prototypes that could demonstrate a clear operational value ”.
The progress of the Air Force Research Laboratory It is called ALPHA and is an artificial intelligence created by Psibernetix, a company founded by the University of Cincinnati in collaboration with the Air Force Research Laboratory. According to the reports published by the Journal of Defense Management, ALPHA is one of the best fighter pilots on the planet. In simulations dogfight with some former Air Force instructors flying thousands of hours behind them, ALPHA would never have been shot down. Indeed, he would have gone on the counterattack of his human counterpart. ALPHA is currently defined as "the most aggressive, reactive, dynamic and credible AI ever created".
During the first tests, ALPHA fought with several simulators, the same ones currently used by the Pentagon to train its pilots. Specifically designed for the dogfight, his archive collects all the maneuvers of his digital and human adversaries: statistics, simple and complex decision models, real and simulated maneuvers. What distinguishes ALPHA is that the transition from defensive elusive actions to offensive ones takes place instantly. ALPHA is based on a decision-making system called "genetic fuzzy tree", the fuzzy logic algorithm. His approach involves the subdivision of a complex problem X into sub-activities. These include all the tactics adopted by man. By relating to variables, artificial intelligence makes complex decisions with extreme speed.
ALPHA, according to the data, calculates the best maneuver in a complex and dynamic environment, 250 times faster than its human opponent. The aspect that most fascinates researchers is the reactive capacity. The adaptive AI reacts instantly by calculating the possible intentions of the opposing driver. ALPHA appears to have been designed for a cyclical flight profile for constant defensive and offensive actions. In fuzzy logic, the system basically calculates an immense web of predictions for a possible action / reaction. Priority is provided by context analysis.
In the Journal of Defense Management it is specified that "Alpha's primary goal is to serve as an intelligent hostile force for pilot training in the Advanced Framework for SIMulation". The Air Force Research Laboratory envisages its implementation in a real cell within the 2025. Meanwhile, new aerodynamic models for the physical platform are being evaluated. The AI, meanwhile, will continue to expand its capabilities, serving as testers for the Air Force Research Laboratory prototypes. The first operational step within the next five years when ALPHA will serve as a follower in a mixed crew, so as to learn the tactical decisions in real contexts.
Considering that between the development and experimentation of the military and what is shown to public opinion there is a gap of about 20 / 25 years, ALPHA could already fly for some time and in secret in a physical version even more powerful.
(images taken from the series Terminator and from the movie Stealth)
