What if future artificial intelligence could predict future events several days in advance? The United States is experimenting with this scenario.
See more articles from iGuRu.gr when you search for news on Google.
A scenario that first became a movie and apparently inspired the Pentagon hawks. It sounds like a last resort to prevent a war. A visionary idea that would lead US military commanders and senior politicians to know in advance the most likely scenario that could happen. Both are quickly adopting the idea of introducing artificial intelligence (AI) into the military services.
In July 2021, the North American Aerospace Defense Command (NORAD) and U.S. Northern Command (NORTHCOM) conducted a third series of tests called the Global Information Dominance Experiments (GIDE) , in collaboration with leaders from 11 command forces. The first and second series of tests were conducted in December 2020 and March 2021 , respectively.
The tests were designed to be conducted in phases, each demonstrating the current capabilities of three interconnected AI-enabled tools, called Cosmos , Lattice , and Gaia.
Gaia provides real-time status awareness for any geographic location, consisting of many different classified and unsorted data sources, such as huge volumes of satellite imagery, communications data, information reports, and a variety of sensor data.
Lattice offers real-time monitoring and threat response options. Cosmos enables strategy and cloud-based collaboration, with many different commands.
Together, these decision-making tools are supposed to predict what opponents will do in advance, allowing U.S. military leaders to anticipate their opponents' actions before any conflict arises.
Tools such as the use of artificial intelligence on the battlefield are particularly attractive to the US military leadership team, as they will prepare them in the future to make decisions in tight times.
They also cite many popular issues that sound loud, such as the dominance of information, the supremacy of decisions, the complete deterrence and the joint command and control of all sectors (JADC2).
Speaking at a one-day conference of the National Security Committee on Artificial Intelligence (NACAI), US Secretary of Defense Lloyd Austin supported the importance of Artificial Intelligence for the possibility of a comprehensive deterrence, expressing his intention to use “ the right mix of technology, operational concepts and capabilities, all woven together in a networked manner that is so reliable, flexible and powerful, that it will stop any adversary ”.
These AI platforms are expected to go beyond simply raising awareness and providing better early warning.
They will offer US military leaders what is considered the holy grail of business planning, producing a strategic warning for hostile actions in the gray area (ie, in the phase of political rivalry), before any irreversible move.
Such a development would allow decision-makers to make precautionary choices (rather than reactionary ones, as they used to) and allow for much faster decisions.
Here is a tempting question: What could go wrong?
Everyone knows that in the basic scenario of science fiction novels and films that explore the possibilities of artificial intelligence, such as Minority Report , The Forbin Project , War Games , etc., something always goes wrong.
The idea is also strangely reminiscent of a Soviet intelligence program, known as RYaN , which was designed to predict a nuclear attack based on data indicators and computational estimates.
Gathering a truly unbiased data set, designed to predict specific results, remains a major challenge, especially for life and death situations and in areas with sparse data availability, such as in a nuclear conflict.
During the 1980s, the KGB wanted to predict the onset of a nuclear war, six months to a full year in advance, from a wide variety of indicators – e.g. unplanned movements of senior officials, FEMA preparations, military exercises and alerts, scheduled weapons maintenance, denials of leave to soldiers, visa approvals and travel information, and activities of US foreign intelligence services.
They even considered the removal of documents related to the American Revolution from public display as a possible indicator of war. Massive amounts of data were fed into a computer model to “ calculate and track the correlation of forces, including military, economic, and psychological factors, to assign probabilities .” Ryan’s findings contributed to Soviet paranoia about a possible nuclear attack by the United States in 1983 and nearly led their leadership to start a nuclear war.
Although such an idea was well ahead of its time, today’s machine learning technologies are now capable of detecting subtle patterns in seemingly random data and could begin to make accurate short-term predictions about adversaries. Amid the excitement over AI-enabled decision-making tools, American defense leaders hope to allay any concerns by insisting that their adoption will be responsible, that humans will remain in control, and that any systems that produce unintended consequences will be taken offline.
However, national security experts such as Paul Scharre, Michael Horowitz and many others point out the critical technical hurdles that need to be overcome before the benefits of using artificial intelligence tools outweigh the potential risks.
Although there is already a lot of useful data for linking machine learning algorithms, aggregating a truly unbiased data set designed to predict specific outcomes remains a major challenge, especially for life and death situations and in areas with sparse data, such as a nuclear conflict.
The complexity of the real world offers another significant obstacle. To work properly, machine learning tools require accurate models of how the world works, but their accuracy depends heavily on human understanding of the world and how it evolves.
Since such complexity often defies human understanding ( Stanislav Yevgrafovich Petrov is a prime example ), AI-enabled systems are likely to behave in unexpected ways. And even if a machine learning tool overcomes these hurdles and works well, the problem of explainability can prevent policymakers from trusting it if they are unable to understand how the tool produced the various results.
Leveraging AI-enabled tools to make better decisions is one thing, but using them to predict enemy actions in order to prevent them is a whole different ballgame.
Aside from raising philosophical questions about free will and the inevitable, it is not clear whether any precautionary measures taken in response to the predicted hostile behavior could be perceived by the other side as aggressive and catalyze a war that the former tried to avoid.
Dimitris hates Mondays...
