To put it in a generic but understandable way, Artificial Intelligence (AI) is a technologically mediated ability (but always present in a digital computer or in a computer-controlled robot) to carry out activities usually typical of an intelligent being.
In this case, the intelligence is the traditional one of the definitions born in the twentieth century in the framework of empirical psychology: logical ability, in the sense of abstraction from the characteristics that science considers “secondary” and hence subjective; understanding, that is the thought correctly imitating the future behavior of the human and non-human movements and reactions present in the external world; emotional knowledge; design, in the absence of an image already present in the external world; finally creativity and problem solving.
As the American pragmatist Charles S. Peirce used to say, understanding or thinking is a form of “talking to oneself” and of symbolically representing – not necessarily reflecting – the inferences that can be found in the external reality.
In Peirce’s mind, all these inferences were probabilistic.
According to the Austrian physicist and philosopher, Ernst Mach, science is instead the process replacing experience with representations and images through which “it becomes easier to handle and manage the experience itself”.
This means that in the transition phase between the nineteenth and the twentieth century, science was no longer interested in the “essence of reality” – interpreted in a reductionist sense – but it created a new reality on its own, easier and more adapted to the human mind and to societal needs.
It was Ernst Mach who applied the criteria for analyzing the data which developed between the nineteenth and the twentieth century in human sciences to physical and chemical science.
In essence, at the end of its epistemological program, the Artificial Intelligence (AI) can reach a complete simulation of the human brain and, in some respects, even outperforms it in its results. Possibly even in the forms of information processing-understanding and hence transformation of what – in a long-standing Western philosophical tradition – is called “reality”.
Hence a “Hyperman”, technologically reminding us of the “Beyond Man” or “Overman” (the Übermensch never to be translated as “Superman”) theorized by Nietzsche, since the homo sapiens sapiens is evolutionarily unstable. Again to use Nietzsche’s words, what we call “man” is a “a rope, tied between beast and overman – a rope over an abyss”.
However, let us revert to military technology.
The IA technology includes not so much the replacement of man with the thinking machine – an idea probably harboured in some people’s minds – but rather more specific techniques: the Virtual Agents; the processing of Natural Language; the platforms for the “self-learning” machines; robotics; the processing of human and computer perception; neural networks.
Incidentally, the Intelligent Virtual Agents (IVAs) are the programs providing pre-established interactions with human beings, especially on the Web. The Natural Language Processing deals with the computer treatment of natural human and non-human language.
The platforms for the “self-learning” machines use the recognition of external patterns and the computational theory of learning – hence they create algorithms that can learn new rules from a wide set of data and make predictions starting from the already pre-defined patterns and from the data sets that grow indefinitely.
In the current phase of this complex “research project” – just to use the terminology of the epistemologist Imre Lakatos – we have reached the following levels: a) we can build systems and robots that are already faster, more capable and more powerful than us. The AI systems are expected to reach our same analytical (and creative) power within 2045. The level of singularity, as this point of no return is currently defined.
Moreover, b) we will have robots permanently taking care of us, interacting with our body and reading our emotions. But this already happens. Google Home, the Jibo control center and the Roomba “social” robot are already among us.
Furthermore, c) also on the basis of a huge and always updatable universe of data, we can predict the great global phenomena at natural, cosmological, medical-epidemiological and human-statistical levels or even at economic level.
We will soon be able to predict also the human behavior in larger populations – often with great accuracy.
On top of it, d) we will have such exoskeletons or extracorporeal extensions as to improve – as never before – our physical and even intellectual/perceptive abilities.
We will shortly become super-human – not in Nietzsche’s meaning of the concept – but rather in the sense of the most popular science fiction comics of the 1960s. Before technological singularity we will record a merger between AI apparata and our mind-body whole.
In the near future, there will be a stable connection between the human brain and computer networks, as already planned and designed by Neuralink or by the Californian company Kernel, which even study the implantation of AI interfaces in the human cerebral cortex.
Finally, e) the end of work.
However, what will the exchange value of the items processed by the AI machines be, considering that our society is based on Smith’s labor theory of value?
How can we set prices, including the non-monetary ones, if there are no values – the classic theme of political economy and also of Marx, his main critic?
In all likelihood, at strategic and military levels we will have 4 types of applicative Artificial Intelligence: the Artificial Narrow Intelligence (ANI), sometimes also referred to as “weak AI”, is an intelligence working within a very limited context, only for specific and routine functions, that cannot take on tasks beyond its field. It is a specific type of AI in which a technology outperforms humans quantitatively in some very narrowly defined task. It focuses on a single subset of cognitive abilities and advances in that spectrum..
Then there is the Artificial General Intelligence (AGI), also known as strong AI, that can successfully perform any intellectual task that a human being can. Finally we achieve the Super Artificial Intelligence (ASI), when AI becomes much smarter that the best human brain in practically every field, including scientific creativity, general wisdom and social skills.
Hence, in principle, the application of Artificial Intelligence in Defense operations and programming will generally regard a) real military organizations; b) the network of political and intelligence organizations developing around real military structures; c) the whole governmental organization, which is a defense structure in itself.
Finally it will regard the whole system of Defense and Security within society, i.e. the whole network of sensors and AI networks that can be used for protecting infrastructure, the territory, as well as economic, strategic and intellectual resources.
Currently the major powers’ research focuses on level 1, namely the Artificial Narrow Intelligence – the level at which AI outperforms the human mind and perceptions only in some sectors and only quantitatively.
Nowadays ANI is used to apply Artificial Intelligence mainly to the battlefield and the integration of forces, intelligence and tactical decision-making within “industrial age” technologies.
An evolution which is still included in and confined to the US Revolution in Military Affairs (RMA) that had its true baptism of fire during the two Iraqi wars.
It was at the core of the doctrinal and technological transformations of the Chinese, Arab, NATO and, later, Russian Armed Forces.
It is worth recalling that the RMA was based on the central idea of the Network-Centric Warfare, which sought to translate an information advantage, enabled in part by information technology, into a competitive advantage through the robust networking of well-informed geographically dispersed forces. Therefore the network and the integration between weapons and sectoral and regional commands – hence the de facto merging between the political-military decision-making and the activities on the ground.
This means that there will be a two-fold approach in modern and future warfare: high-technology strikes, which determines the strategic superiority on the field, as well as the whole new low or medium-intensity panoply of political warfare – which operates with the apparent opposite of the Special Forces, on the one hand, and of parallel, civil and rank-and file organizations on the other, including armed citizens, mass operations and operations of influence, the use of local criminal and non-criminal organizations, and the stable “black”, “white” and “grey” propaganda.
Future wars will be more widespread and characterized by swarming, because many regional and local actors, including non-State ones, can afford attack and defense panoplies on the basis of Artificial Intelligence – systems more connected to the link between propaganda and politics and less Clausewitz-style: the separation between warfare and non-warfare will disappear in the future and the armed clash will not be “the continuation of politics by other means” – as maintained by the Prussian general and military theorist – but, if anything, there will be a continuum between armed action and political and economic-social operations.
With new and extraordinarily relevant legal issues: who is responsible for an AI or cyber attack? Can we rely on probabilistic analysis of enemy operations or will there be a “cyber or robotic declaration of war”?
Moreover, AI is a way of rethinking, reformulating and reducing military spending, with smarter and more flexible technologies and better cost effectiveness. Hence we will witness the gradual end of the oligopolistic market of technological and military acquisitions – typical of a traditional industrial world – and the emergence of some sort of market economy, open to even the smallest states, in the old political-military establishment that, as early as the 1950s, Eisenhower accused of directing the Western countries’ foreign policies.
In the market of strategic acquisitions there will be specific room for commercial algorithms which, however, can be applied also to the military universe. Nevertheless, in what we have called the “third level” of AI, that is the integration between government and strategic operations, we will have to deal with algorithms that will rationalize bureaucracy and the decision-making process, both at governmental and operational levels.
This will happen also for what we have defined as the Fourth Level, i.e. the dimension of the ecosystem between politics, technology and the rest of society, which is not normally interested in military operations.
As already noted with reference to China and to the Russian Federation, here AI will deal with social prevention (which is the new way of avoiding the post-Clausewitz mix of clash and political representation) and with social resilience, namely the stability of “civil societies” and their critical infrastructure. Not to mention counter-propaganda.
It was Napoleon-style Blitzkrieg.
However, in the future, it will no longer be sustainable, economically and politically, given the military forces’ economic and social limits we are already experiencing today. Hence the link between the AI-Defense Fourth Level and the previous ones will be between Deep Learning, new wide databases, as well as high-speed and highly performing computers.
Within the framework of the NATO countries’ current defense doctrines, the main AI military actors have paid the utmost attention to information and computer technologies that bring together – quickly and easily – the “effectors”, i.e. those or the things that perform the operations with human or artificial “sensors” – the so-called Network Enabled Capabilities, in the NATO jargon.
Nevertheless, how do the old and new superpowers respond to the challenge of military Artificial Intelligence?
China – the country that best stands up to the USA in this field – established the National Laboratory for Deep Learning, which has been operational since February 2017.
Moreover the Chinese company Baidu and the other Chinese web giants have been entrusted with the task of working with the State in sectors such as automatic visual recognition; recognition based on a vast evolutionary database; voice recognition; the new automated models of Man/Machine interaction; intellectual property in the sphere of deep learning.
In this field everything is based on supercomputers, which China can currently manufacture on its own, after the USA blocked the sale of the Intel Xeon processors, up to even producing their autonomous superprocessor, present in the advanced computer Sunway Taihu Light, which is so far the fastest computer in the world, at least in the field of complex computer networks.
Furthermore China’s 13th Five-Year Plan envisages an expansion of the national AI market to the tune of 100 billion renmimbi, with two specific plans: the China Brain Plan, that is AI military-civilian planning for unmanned networks, as well as cybersecurity and complex society’s governance.
The other level, the second one planned by China, is the use of Artificial Intelligence only for military and strategic superiority.
This is President Xi Jinping’s policy line of the progressive “military-civilian integration”.
The Chinese Armed Forces have also established an Unmanned Systems and Systems of Systems Science and Technology Domain Expert Group, in addition to working hard in the sector of visual recognition for the Navy – above all for operations in “disputed waters” – as well as dealing specifically with the command and control of large-range Unmanned Aerial Vehicles (UAVs).
Missiles are another important sector in China’s AI military policy – above all to respond to the recently-deployed US Long Range Anti-Ship Missile program, which has replaced Harpoon.
In China all this is included in the broader theory of Remote Warfare, which is based on drones and advanced missile networks.
This is currently at the basis of many Chinese strategic choices. Hence – at any distance – hitting targets which are a greater danger for China, as well as for the Forces on the ground and for the politics-warfare link – a danger that cannot be replaced or postponed in response.
The Chinese Armed Forces are and will always be used “to defend the Party”.
Moreover, Israel was the first country – even before the USA – to use fully automated robots and unmanned military vehicles in warfare, besides building Harpy, the anti-radiation UAV searching, targeting and destroying enemy radar centers without human control and supervision.
In the near future, the Israeli decision-makers plan to deploy “mixed” battalions, with robots and human soldiers operating together.
Moreover the Israeli Armed Forces have already put in place the system called Automatic Decision Making, employing robotics, AI and deep learning and operating with almost instantaneous speed, which is strategically unavoidable for Israel.
Aeronautics Ltd, an IDF contractor, has already built a series of UAVs having complex Artificial Intelligence algorithms.
IA systems to support political decision-making, as well as techniques for the immediate transfer of data from one computer platform to another, and finally AI technologies for the camouflage of networks and human and non-human operators are already operating in Israel.
In the Russian Federation, military Artificial Intelligence has a limit, that is the availability of ultra-fast processors for supercomputers – a problem which, however, is being solved.
Currently Russia is mainly interested in developing the Unmanned Ground Vehicles (UGVs), in addition to the robotic platforms for the integration of the various aspects of the battlefield.
In January 2017, President Vladimir Putin ordered the creation of “autonomous robotic complexes” – just to use the Russian government’s terminology – but for military use only.
With the creation of the National Center for the Development of Robotic Technologies and Basic Robotic Components, Russia is implementing a careful policy of acquisition and independent research in the field of military Artificial Intelligence. A network that already operates for acquisitions throughout the world market.
This is an organization which has been operating since 2015-2016.
Russia has already developed unmanned helicopters and the use of remotely-controlled robot-terminators which target alone – again without human supervision – the targets they have autonomously selected by severity of threat and response to actions on the field.
Finally, after the good results reached with its unmanned operating platforms in Syria, Russia is interested in developing AI systems for border protection, with a series of neural networks automatically referring and reporting to cameras, seismic and human sensors, as well as UAV networks, for an immediate response to threats.
For the USA, the first country to be permanently committed to military AI, the next developments will be in the following areas: a) autonomous machines for deep learning, capable of collecting and processing data and making choices, especially in the framework of the current “hybrid warfare”; b) the development of AI strategic doctrines in the field of man-machine collaboration, with the final implementation of the Centaur network; c) the creation of joint man-machine combat units; d) web-connected semi-automatic weapons to survive cyberattacks.
All these systems will be obviously online and interconnected.
Certainly, nowadays, all the major operators of strategic Artificial Intelligence need to use these networks for the crypto-preservation of real intelligence data, as well as for their classification and also for conflict prevention which, as China maintains, must be “predictive, preventive, participatory and shared between political and military decision makers”.
Moreover, the goals of the new AI military networks include “social resilience”, i.e. the stability of the non-military, namely of the members of the “civil society”, faced with any unexpected shocks and actions of “covert or hybrid warfare”.
What about Italy? It has no real document on National Security to be updated every year or for major crises.
This is already a severe limitation. The “White Papers” already drafted in Italy, however, are always political and government documents. They are often drafted by “external” people not involved in the Defense mechanism, or possibly, by the General Staff and they are more focused on the rather vague spending plans and policy lines, as well as on the Grand Strategy, if any, than on the strictly geopolitical and military doctrine.
Little focus on National Interest and much unrequested loyalty to goals set by others.
Furthermore, at the Ministry for Foreign Affairs, there is a group of experts working on the Italian foreign policy challenges until 2020.
Hence, apart from the specific activities of the intelligence services, in Italy there is no doctrine or project for internal use of the AI technologies, also in view of stimulating our currently very scarce industries in the IT-AI sector.
Another great deficiency – among the many shortcomings of our defense criteria – and also scarce integration with the other NATO and Allied Armed Forces.
GIANCARLO ELIA VALORI
Honorable de l’Académie des Sciences de l’Institut de France