On the Future Perfect of Artificial Intelligence and War: War and Algorithm Review. Max Liljefors, Gregor Noll, and Daniel Steuer, War and Algorithm
Geoff Gordon · Journal of Conflict and Security Law · 2021
Earlier this year, the US National Security Commission on Artificial Intelligence (NSCAI) announced its conclusions in support of developing artificial intelligence as instruments of war and surveillance, including lethal autonomous weapons systems (or LAWS) for military use. This is not entirely surprising, given that the NSCAI hosts conferences called ‘Strength through Innovation: The Future of A.I. and U.S. National Security’.1 It is led by Eric Schmidt, former Google Chief Executive, a central figure in the interrelationship of the US defense community and Silicon Valley,2 presiding over representatives from Google, Oracle, Microsoft and Amazon Web Services, as well as current and former members of the US defense department and other interested parties. More surprising may have been that the NSCAI posited a moral mandate as part of its recommendation. The NSCAI’s vice chairman, Robert Work, former US Deputy Secretary of Defense, argued that it is ‘a moral imperative to at least pursue this hypothesis’, the hypothesis being that weaponized artificial intelligence and LAWS constitute a positive humanitarian good.3 There is a problem, however, with calling this argument a hypothesis: namely, we cannot test it. As Gregor Noll points out in his contribution to War and Algorithm , we do not have access to a parallel universe in which we have the time and capacity to test the ways in which LAWS will change the battlefield, for better or worse. In lieu of an untestable hypothesis, War and Algorithm offers a probing interrogation applicable to Work’s claim, along with a number of snapshots of the sorts of AI technologies under development for security purposes. These observations are framed in the introduction of War and Algorithm as part of a conversation across disciplines. The conversation has apparently been conducted slowly, steadily, and on an occasional basis, with the result that the principal authors no longer recall in some places precisely what material originated with whom. The product of their conversation presents the reader with a meditation on the nature and stakes of contemporary algorithmic technologies that have enrolled and been enrolled in the violence of war. I say enrolled and been enrolled because the book depicts a symbiotic relationship between these technologies and the militaries that deploy them; sometimes the technology fights for the military, sometimes the military fights for the technology.4 The book is written by three principal authors from three different academic fields—Max Liljefors, Gregor Noll and Daniel Steuer—each of them engaged by one of three interlocutors, Allen Feldman, Sara Kendall and Howard Caygill, respectively. Steuer and Caygill hold appointments in philosophy; Noll and Kendall in international law; Liljefors in art history, Feldman in media, culture and communication. The diversity of academic fields makes clear that the book is not principally a legal text. Moreover, the authors expressly disavow an ‘interdisciplinary’ aim.5 They aim instead at ‘disciplinary unruliness’, which they associate with finding ways out of their ‘professional confines’.6 Understanding that ambition is key to appreciating the work and its salience for a legal journal tailored to conflict and security law. The authors communicate a widely-held concern that their several disciplines are each challenged at their limits by new technologies. They are looking for ways to meet those challenges. Their method is to push one another past disciplinary limits by confrontations from beyond each individual discipline. This does not mean that the authors abandon disciplinary constraints: each proceeds from a deep grounding in the dogmas of their respective fields. Nor will their unruliness satisfy every reader interested in surpassing disciplinary boundaries: these liminal exercises are situated and provoked by specific confrontations. The combination of political philosophy, law, art history and media studies here yields partial and particular demarcations of the borders among them. The choice of disciplines for this particular exercise is clear enough: the authors observe a radical change in the valence of international law and politics when practiced in a security environment defined by contemporary information media. Consider, in this light, Fleur Johns’s description of a redistribution of the sensible in international law, which refers in part to the algorithm- and data-driven ways of knowing and constituting the world in which international law is practiced.7 Johns describes techniques of pattern recognition applied recursively and reiteratively to constant streams of information and images, to construct an actionable international space in any given institutional context. The actionable construction is enabled by a technological distribution of sensors and processors, sensors to generate information about the world, processors to assemble it into actionable substance. Such information systems, featuring cables, satellites, sensors, computers and screens, among other things, are part of the military technologies including LAWS and drones that drive the concerns in War and Algorithm. The technologies that are reshaping the construction of international spaces are not available equally to all, and the same technologies are reshaping the ways of securing the world they help to construct. Combinations of biometric and GPS technologies control flows of people at borders in new ways, while also enabling movements of weaponized drones across the same borders, changing the valence of borders in the process.8 In this new distribution, there is a live debate as to how Charter principles of territorial integrity apply in cases of drone operation,9 or even whether international humanitarian law can apply to autonomous weapons systems.10 As Robert Work’s quote already underscores, there is a sense in which technologies have pushed our legal-institutional devices to the limits of their applicability, so that high-level policy committees resort to extra-legal hypotheses to justify the emerging security architecture they are building for the conduct of war. The authors in War and Algorithm meet this sense of frustrated limits by drawing vocabularies of law and politics into confrontation with vocabularies of media and art. In this review, I will try to respect the authors’ ambitions to defy professional limitations, while explaining in brief and for a legal audience the different conceptual devices on which they continue to rely. As a result of the unusual set-up of the book—three principal authors, each paired off with one interlocutor—the book has a double structure: the principal authors in conversation with one another; and each principal author in conversation with an interlocutor: Caygill with Steuer, Kendall with Noll and Feldman with Liljefors. Despite this double structure, or in addition to it, common themes and leitmotifs run through every chapter, and it is not hard to draw connections across any of the contributions. Algorithmic techniques of war-making, of course, make up the common objects of analysis. Within this broadly but clearly delimited analytical space, a space saturated in this book with a vocabulary of governance, one common theme explores the limits of human perspective as we know it—this is a situated we, assuming a perspective constructed under conditions associated with the late-Enlightenment west. Another common theme investigates the character of cybernetic intelligence predicated on information and signal strength in neural networks. These are just two prominent examples—there are more.11 Despite the integration, however, each contribution maintains a distinct voice, and the balance of preoccupations changes from one to another. Below, I will schematically treat each chapter in turn, in their order in the book, but recommend reading the book as a whole (though that hardly needs to be linear). The next section, focusing on the chapters by Steuer and Caygill, sets the stage, framing the prosecution of war in a world defined by information technologies. The section thereafter, focusing on the chapters by Noll and Kendall, addresses the dilemmas posed as a matter of law in this emergent security environment. The section after, focusing on the chapters by Liljefors and Feldman, considers where the emergent technologies and norms appear to be leading. Finally, I offer some summary observations of my own in conclusion. Daniel Steuer sets a tone with the first chapter, entitled Prolegomena to Any Future Attempt at Understanding Our Emerging World of War, urgently critiquing a failure of critical in security the failure by the world as a The world as a is a world of one in which artificial intelligence is to Steuer sets off the world of information with to it from a world that it, and world, respectively. does a of work for Steuer, and what it it Steuer to in the of international by in the fields of defense and The that is by information technologies is a product and of and security on work on new are with the of new work for Steuer a and environment of among and with while war as The is the the of military in the character of that conflict a for war on Steuer the of the from of the in which the the of in Steuer the of the with the help of Steuer and but also as well as the of to and The in which are however, have the for which are no longer but in In with the technological and character that the contemporary conflict of new are and and this of by among of technology and Steuer to on The of and the makes between and military work points to the of information as a military which a theme of chapter, and the In addition to Steuer on as and for the ways in which information broadly has a vocabulary of governance, security and over the world as information the of or even this is not on Eric and at the As Sara Kendall will this is in the the this is not a to the one The AI technology at work by signal strength and by pattern recursively for with to some and AI by not on the of positive about in the world, but on the of cybernetic techniques to and points up in information technologies neural over and algorithmic and that out of its information The to Steuer, is ‘a world that the of This a to War and in the chapter, by Feldman, the of as the of In chapter, however, the principal is the of critical the of the world or is one that no space for of the work by weaponized of artificial here I have and with respect to The of AI into a for critical is my is in the argument is out of of of It is not clear how well they Steuer to try to any by of In it as the argument is being by Moreover, of that that the the of however, the for of the Howard entitled and the Emerging World of War, off from the sense of conflict that In this Caygill an to and The however, a critical in the of and Caygill in the that as the an of the of critical for to a to or through and the of the preoccupations at work in chapter, that of are in that Steuer to out of the by systems and the development of cybernetic These and the that also a for on from the security associated with to the associated with the Caygill does not the the and limits of the of including the cybernetic and world systems by the and engaged by by including of the as a whole a of of its into as an on the of is not to the is that is in a that it can that not entirely In out of security concerns by the development of to specific and technologies of systems and cybernetic information technologies. These technologies constitute the emerging world of war by Caygill as a which as a world of the this the to human posed by the of systems of and in an emergent world of war needs to be by to changes in the of the It needs also to the of the of not its own and the with the that hosts There is however, in this conflict is in the nature of war. is at the of and of the of that on an of in the of Caygill this with an in perspective and history from the of time this as an however, it is to know what conclusions to draw from to is and the of the that the and the have to the current techniques of war and while Caygill that the human is in on the of time to which how that will in the the and with and over and that Caygill under have at The next chapter, by Gregor entitled War by The of and the of cybernetic to what framing of autonomous weapons by for our to them through The is is not to algorithmic of to the to the at this is also a on my as Noll makes his the that the and or or beyond contemporary These with a over technology or the Noll to be at a in contemporary weapons systems an of with human in an architecture in which of them can be from any that of law may be to the of in ways that cannot be to a It is surprising, in this light, not to or work into this chapter or and other have been with of the that Noll surpassing the and namely, in how the in which a a and a of law are with the world while Noll an of human and is not at of human law and the law of in hold in the because law in the as Noll it is an of a to a The of law that Noll on is a in which the law is from any one instead in a that be argued over and by with an of the neural a from the in the former law by is in the with on and the is As by Steuer, there is no for critical in the as in the In of critical there is a at predicated on the of and artificial intelligence least in its The of the former is to the artificial intelligence on a neural This on AI that the human cannot or even in of the nature of the what the of the no longer The of the neural the as pattern This the of at and the AI artificial neural on there being as emergent emergent the these emergent are into a of In this the world of information and available to the neural and a by and for which war is The of the of artificial intelligence to draw actionable and so to actionable out of so to a but that Noll with respect to with artificial there is no time to test the there is no in which to test the Noll an with the posed by in the US defense no space, featuring the of our for In AI it is time the space we In order to have an it is not so a parallel we as a parallel The of artificial intelligence the security conditions for its own a the weaponized cannot be applied and law to and changes in will apply to a world already by a next In this Noll the of artificial intelligence in war is a of in the law of the cybernetic which of the world, of its and a particular in history are into the Robert Work’s quote for the as a a hypothesis, the moral and humanitarian a with Sara entitled Algorithmic and the by in which law limits LAWS in or in which LAWS are a law on the of chapter, into for Kendall the to two of its and its aim The is the where law the aim is its The in chapter, is up with the between human and contribution offers a different of the by legal beyond the about that Kendall the by grounding it in a material history of and with and through international and humanitarian law. In this in which the law is an of the violence it LAWS are not but its chapter the of the chapter how to them of the political the a political on the of the on of the in the is a of and even over them. of course, is also the legal for to of the resort to the technology is the of violence in international law, for in the War and the of and so it is not to international law that Kendall for an instead to the capacity of the for the of is the at Google that the in as the Algorithmic is the first US to AI technologies for It that two of the for Robert at the US of and Eric at and of the Kendall the on The to from the recognition that even international law to be of it is already so is already violence that it This is a of the history of violence in which international law and humanitarian law are As the of the to the book, a central chapter is not in with but with each of the other as As one of the history of international Kendall points to of that under international law for defined by military to the of to capacity in security defined by to concern for in while Kendall Noll when an history and of the same offers to of material history makes the ways in which drone has that are in the of a that Liljefors will Kendall points to observations about the ambition to an one which the human in the aim of or securing with chapter that does not apply (or is not equally to applied to political to themes that Feldman the ways in which have technological that have the integrity of the This of the however, is part of a and specific history of in a world in between that are to control over their and that for to the of and this basis, Kendall points out that of LAWS will out to the of the former the of the as some are to the of drone in their this to the of this as an with LAWS will not test a humanitarian hypothesis, will the of the be equally in the and The chapter by Liljefors, entitled and its argument in at least three different with images, with a and with In the and a of images, a is and to its it by from The of with an featuring a but perspective of and with of the The in between including the of and Liljefors, however, also the one is The and the of the which Liljefors to from contemporary security artificial the other is a of the of and a by of a drone and drone at a two a of a of the and the of a situated to an is the same that the of The of the a different but and apparently in its lethal and for Liljefors a product of or a of that in to the that it Liljefors describes three of of a looking through of to at a to the technology that so that the of the technology This to the current of AI and technologies into of governance, and with observations Fleur Johns’s description of international data-driven one in which the particular work by information and technologies is The of a to or a failure of perspective at with the to at Liljefors refers to this as from the is the to a at technology no one the of the of In space, human being is entirely as information in that of This to and objects of drone and in a as in it is the Liljefors of the world of the world as and which LAWS will is the by it is hard to what of the legal in a space where perspective is for an in this Allen the the to the of War, up from Feldman describes the of the to algorithmic with the that it In this chapter a to the our in the technologies of algorithmic Feldman does through a of concerns an on a that to a with its it at they with the however, the is precisely because each can it and because each can also other the as it It is by this interrelationship of with one another and with the that the its to the the own as a of a This the in the of the its or a world which any is a The also Feldman to the contemporary of algorithmic a of and The of on the of and to the of and and with it the of and Feldman contemporary in a longer history of to the and of The of the contemporary in the of its including what and how that technology the technology on by has a the of to generate the as in drone on on which Feldman is a it is the of an through the of of information from so is into so and which can be and into sets pattern recognition and and spaces and The actionable are into sets in order for these sets to be as they the Steuer, that actionable information to a to which it is to a that I will to in just a however, that a This of to at the and that algorithmic Feldman, however, points out the posed by algorithmic to that of the as the algorithmic and of that which the the violence of actionable and as the of a are that do not through it is not just an the of algorithmic the of its own violence as part of the integrity of its I a that the information in the of world as or that theme to its in the book, Feldman how information an in is a drive that of world not through its as a of that to and security its own and as to the world it to and as its chapter, I is and It from the that in the of algorithmic into and the of the analysis. Feldman offers at least one in addition to the analysis. The to the The is a not to any one human while of them. Feldman also that algorithmic is not any The of the of algorithmic is the that space the the the of the is the of its AI does not and in and the of AI is hardly the first to AI from other technologies for These chapters to a number of in one of the the about an emergent technology in an emergent world of war. There is a double at The technology in its world of war into the technology on it as part of a and in this sense is to the of and in a to a constant of and The of is to make is to the of to the one that is The to AI refers to its as part of a of sensors and information to over and over a In this weaponized AI can be to three pattern recognition out from the The of is the of the of is also just another for into the next of the as part of a that is people and cables, and As out by Feldman, the actionable out of of the of its particular and the particular that through it, however, that on in emergent for new among the emerging the and that with each next of actionable The of information technologies and security at here is In addition to the brief just there are and among other Liljefors, it is to the of it in to this makes the of a The stakes are but to the of the information technologies and so the of information that drive the new security with the in the of international law and there has been of techniques from information and techniques that aim to the of political and These techniques in the of as an of legal and political it is not clear whether techniques the by the emerging world of or whether they are part of the of that I one time to the by Robert Work, on of the that it is a moral imperative for the US to pursue the hypothesis that LAWS constitute a humanitarian As Noll this is not a cases may be to the to this or but the by over and There are no parallel by which to test the world it makes the it will have There is a constant of so by AI at work in particular which are to some and not a moral imperative is in a policy one that the to over the to their this I with a quote from the of War and In order to we to be to the time to on the of our and on the to which they and our can we our into so that we may the as an the for our to mean to to to the of the and to the that its from