An Assessment of the Likely Roles of Artificial Intelligence and Machine Learning Systems in the Near Future

Ali Crawford has an M.A. from the Patterson School of Diplomacy and International Commerce where she focused on diplomacy, intelligence, cyber policy, and cyber warfare.  She tweets at @ali_craw.  Divergent Options’ content does not contain information of an official nature nor does the content represent the official position of any government, any organization, or any group.


Title:  An Assessment of the Likely Roles of Artificial Intelligence and Machine Learning Systems in the Near Future

Date Originally Written:  May 25, 2018.

Date Originally Published:  July 16, 2018.

Summary:  While the U.S. Department of Defense (DoD) continues to experiment with Artificial Intelligence (AI) as part of its Third Offset Strategy, questions regarding levels of human participation, ethics, and legality remain.  Though a battlefield in the future will likely see autonomous decision-making technology as a norm, the transition between modern applications of artificial intelligence and potential applications will focus on incorporating human-machine teaming into existing frameworks.

Text:   In an essay titled Centaur Warfighting: The False Choice of Humans vs. Automation, author Paul Scharre concludes that the best warfighting systems will combine human and machine intelligence to create hybrid cognitive architectures that leverage the advantages of each[1].  There are three potential partnerships.  The first potential partnership pegs humans as essential operators, meaning AI cannot operate without its human counterpart.  The second potential partnership tasks humans as the moral agents who make value-based decisions which prevent or promote the use of AI in combat situations.  Finally, the third potential partnership, in which humans are fail-safes, give more operational authority to AI systems.  The human operator only interferes if the system malfunctions or fails.  Artificial intelligence, specifically autonomous weapons systems, are controversial technologies that have the capacity to greatly improve human efficiency while reducing potential human burdens.  But before the Department of Defense embraces intelligent weapons systems or programs with full autonomy, more human-machine partnerships to test to viability, legality, and ethical implications of artificial intelligence will likely occur.

To better understand why artificial intelligence is controversial, it is necessary to distinguish between the arguments for and against using AI with operational autonomy.  In 2015, prominent artificial intelligence experts, including Steven Hawking and Elon Musk, penned an open letter in which the potential benefits for AI are highlighted, but are not necessarily outweighed by the short-term questions of ethics and the applicability of law[2].  A system with an intelligent, decision-making brain does carry significant consequences.  What if the system targets civilians?  How does international law apply to a machine?  Will an intelligent machine respond to commands?  These are questions with which military and ethical theorists grapple.

For a more practical thought problem, consider the Moral Machine project from the Massachusetts Institute of Technology[3].  You, the judge, are presented with two dilemmas involving intelligent, self-driving cars.  The car encounters break failure and must decide what to do next.  If the car continues straight, it will strike and kill x number of men, women, children, elderly, or animals.  If the car does not swerve, it will crash into a barrier causing immediate deaths of the passengers who are also x number of men or women, children, or elderly.  Although you are the judge in Moral Machine, the simulation is indicative of ethical and moral dilemmas that may arise when employing artificial intelligence in, say, combat.  In these scenarios, the ethical theorist takes issue with the machine having the decision-making capacity to place value on human life, and to potentially make irreversible and damaging decisions.

Assuming autonomous weapons systems do have a place in the future of military operations, what would prelude them?  Realistically, human-machine teaming would be introduced before a fully-autonomous machine.  What exactly is human-machine teaming and why is it important when discussing the future of artificial intelligence?  To gain and maintain superiority in operational domains, both past and present, the United States has ensured that its conventional deterrents are powerful enough to dissuade great powers from going to war with the United States[4].  Thus, an offset strategy focuses on gaining advantages against enemy powers and capabilities.  Historically, the First Offset occurred in the early 1950s upon the introduction of tactical nuclear weapons.  The Second Offset manifested a little later, in the 1970s, with the implementation of precision-guided weapons after the Soviet Union gained nuclear parity with the United States[5].  The Third Offset, a relatively modern strategy, generally focuses on maintaining technological superiority among the world’s great powers.

Human-machine teaming is part of the Department of Defense’s Third Offset strategy, as is deep learning systems and cyber weaponry[6].  Machine learning systems relieve humans from a breadth of burdening tasks or augment operations to decrease potential risks to the lives of human fighters.  For example, in 2017 the DoD began working with an intelligent system called “Project Maven,” which uses deep learning technology to identify objects of interest from drone surveillance footage[7].  Terabytes of footage are collected each day from surveillance drones.  Human analysts spend significant amounts of time sifting through this data to identify objects of interest, and then they begin their analytical processes[8].  Project Maven’s deep-learning algorithm allows human analysts to spend more time practicing their craft to produce intelligence products and less time processing information.  Despite Google’s recent departure from the program, Project Maven will continue to operate[9].  Former Deputy Defense Secretary Bob Work established the Algorithmic Warfare Cross-Functional Team in early 2017 to work on Project Maven.  In the announcement, Work described artificial intelligence as necessary for strategic deterrence, noting “the [DoD] must integrate artificial intelligence and machine learning more effectively across operations to maintain advantages over increasingly capable adversaries and competitors[10].”

This article collectively refers to human-machine teaming as processes in which humans interact in some capacity with artificial intelligence.  However, human-machine teaming can transcend multiple technological fields and is not limited to just prerequisites of autonomous weaponry[11].  Human-robot teaming may begin to appear as in the immediate future given developments in robotics.  Boston Dynamics, a premier engineering and robotics company, is well-known for its videos of human- and animal-like robots completing everyday tasks.  Imagine a machine like BigDog working alongside human soldiers or rescue workers or even navigating inaccessible terrain[12].  These robots are not fully autonomous, yet the unique partnership between human and robot offers a new set of opportunities and challenges[13].

Before fully-autonomous systems or weapons have a place in combat, human-machine teams need to be assessed as successful and sustainable.  These teams have the potential to improve human performance, reduce risks to human counterparts, and expand national power – all goals of the Third Offset Strategy.  However, there are challenges to procuring and incorporating artificial intelligence.  The DoD will need to seek out deeper relationships with technological and engineering firms, not just defense contractors.

Using humans as moral agents and fail-safes allow the problem of ethical and lawful applicability to be tested while opening the debate on future use of autonomous systems.  Autonomous weapons will likely not see combat until these challenges, coupled with ethical and lawful considerations, are thoroughly regulated and tested.


Endnotes:

[1] Paul Scharre, Temp. Int’l & Comp. L.J., “Centaur Warfighting: The False Choice of Humans vs. Automation,” 2016, https://sites.temple.edu/ticlj/files/2017/02/30.1.Scharre-TICLJ.pdf

[2] Daniel Dewey, Stuart Russell, Max Tegmark, “Research Priorities for Robust and Beneficial Artificial Intelligence,” 2015, https://futureoflife.org/data/documents/research_priorities.pdf?x20046

[3] Moral Machine, http://moralmachine.mit.edu/

[4] Cheryl Pellerin, Department of Defense, Defense Media Activity, “Work: Human-Machine Teaming Represents Defense Technology Future,” 8 November 2015, https://www.defense.gov/News/Article/Article/628154/work-human-machine-teaming-represents-defense-technology-future/

[5] Ibid.

[6] Katie Lange, DoDLive, “3rd Offset Strategy 101: What It Is, What the Tech Focuses Are,” 30 March 2016, http://www.dodlive.mil/2016/03/30/3rd-offset-strategy-101-what-it-is-what-the-tech-focuses-are/; and Mackenzie Eaglen, RealClearDefense, “What is the Third Offset Strategy?,” 15 February 2016, https://www.realcleardefense.com/articles/2016/02/16/what_is_the_third_offset_strategy_109034.html

[7] Cheryl Pellerin, Department of Defense News, Defense Media Activity, “Project Maven to Deploy Computer Algorithims to War Zone by Year’s End,” 21 July 2017, https://www.defense.gov/News/Article/Article/1254719/project-maven-to-deploy-computer-algorithms-to-war-zone-by-years-end/

[8] Tajha Chappellet-Lanier, “Pentagon’s Project Maven responds to criticism: ‘There will be those who will partner with us’” 1 May 2018, https://www.fedscoop.com/project-maven-artificial-intelligence-google/

[9] Tom Simonite, Wired, “Pentagon Will Expand AI Project Prompting Protests at Google,” 29 May 2018, https://www.wired.com/story/googles-contentious-pentagon-project-is-likely-to-expand/

[10] Cheryl Pellerin, Department of Defense, Defense Media Activity, “Project Maven to Deploy Computer Algorithims to War Zone by Year’s End,” 21 July 2017, https://www.defense.gov/News/Article/Article/1254719/project-maven-to-deploy-computer-algorithms-to-war-zone-by-years-end/

[11] Maj. Gen. Mick Ryan, Defense One, “How to Plan for the Coming Era of Human-Machine Teaming,” 25 April 2018, https://www.defenseone.com/ideas/2018/04/how-plan-coming-era-human-machine-teaming/147718/

[12] Boston Dynamic Big Dog Overview, March, 2010, https://www.youtube.com/watch?v=cNZPRsrwumQ

[13] Richard Priday, Wired, “What’s really going on in those Bostom Dynamics robot videos?,” 18 February 2018, http://www.wired.co.uk/article/boston-dynamics-robotics-roboticist-how-to-watch

Ali Crawford Alternative Futures Artificial Intelligence & Human-Machine Teaming Capacity / Capability Enhancement United Nations

Assessing How Article 51 of the United Nations Charter Prevents Conflict Escalation

Jared Zimmerman is an M.A. candidate at American University’s School of International Service where he is studying United States Foreign Policy and National Security with a concentration in terrorism and political violence.  He can be found on Twitter @jaredezimmerman.  Divergent Options’ content does not contain information of an official nature nor does the content represent the official position of any government, any organization, or any group.


Title:  Assessing How Article 51 of the United Nations Charter Prevents Conflict Escalation

Date Originally Written:  March 8, 2018

Date Originally Published:  June 4, 2018.

Summary:  Article 51 of the United Nations Charter is sufficiently vague to allow states to assert their right to self-defense without escalating a conflict. While either side in a conflict may see the other as the aggressor acting beyond mere self-defense, Article 51 is vague enough that neither side can prove the other has acted offensively. This vagueness can aid in, if not the de-escalation of conflicts, preventing the rapid escalation of conflicts.

Text:  The first sentence of Article 51 of Chapter 7 of the United Nations Charter reads as follows:

Nothing in the present Charter shall impair the inherent right of individual or collective self-defence if an armed attack occurs against a Member of the United Nations, until the Security Council has taken measures necessary to maintain international peace and security[1].

This sentence is particularly vague on the following points:

  1. It does not define what constitutes an attack. Is the seizure of ships or aircraft an attack? Is the accidental or intentional violation of another country’s airspace an attack? Is industrial espionage an attack? Is a spy satellite taking photographs of military installations an attack?
  2. It does not define what constitutes an armed attack. For example, is a cyber attack an armed attack?
  3. It does not define “collective self-defence.” Does the attacked nation need to request assistance or can other nations preemptively intervene and claim their intervention constitutes collective self-defense? Requiring the attacked nation to request assistance might seem like the most responsible position, but this requires that the United Nations Security Council determine who the original aggressor and defender are. This determination may not be possible or delivered in a timely manner.
  4. The phrase “…until the Security Council has taken measures necessary to maintain international peace and security” begs several questions. What if the Security Council does nothing? What if the Security Council does act, but these actions are not sufficient to resolve the conflict? What constitutes a resolution and who decides whether a resolution is satisfactory?
  5. The phrase: “international peace and security” also begs several questions. What is international peace and security? Was the world at peace during the Cold War? Is the world not at peace when great powers are not in conflict but relatively small regional or civil wars are ongoing? Is the world at peace when there is no open conflict between states but despots murder and oppress their own people?

It is apparent from the questions in the preceding paragraphs that the first sentence of Article 51 is exceedingly vague. Opposed parties in a real-world conflict are certain to interpret portions of the sentence in their own best interest, and these interpretations could be wildly different yet equally valid[2]. But this begs the question, does this vagueness expand and escalate conflicts or limit and de-escalate them?

On the surface it might appear that a more explicit Article 51 is to be desired. If it was clear to states what actions constitute an armed attack and what circumstances allow for collective self-defense, perhaps states would judiciously aim to abide by these rules lest they risk United Nations’ intervention. There are several problems with this approach:

  1. It would be impossible to explicitly account for all types of armed attacks, not simply because of the variety that exists today, but because new types are continually being invented. For example, the authors of the United Nations Charter could hardly have conceived of cyber warfare in 1945.
  2. States are ingenious and will always find new ways to circumvent—or even outright ignore—any explicit rules that are laid out.
  3. If a state realizes it must break one explicit rule to advance its agenda, why not break more? If the United Nations Security Council does not intervene when one rule is broken, will it if two are broken? Three? Four? States will test how far they can push the boundaries because it is advantageous to do so.

Is it possible, however, that having a vague Article 51 is advantageous? The world is not rigid, so would it be beneficial to have a rigid Article 51? Given the reasons above, a rigid Article 51 is certainly not practical. Let us take the Iranian drone shot down by the Israeli Defense Force in February 2018 as an example of the advantages of a vague Article 51.

On February 10th, 2018 an Iranian drone entered Israeli airspace and was shot down by an Israeli helicopter. The Israeli Defense Force followed up by attacking what they believed to be the “drone launch components in Syrian territory[3].” Later, Israeli Air Force (IAF) aircraft attacked 12 targets in Syria, including a mix of Syrian and Iranian military targets. “During the attack, multiple anti-aircraft missiles were fired at IAF aircraft. The two pilots of an F-16 jet ejected from the aircraft as per procedure, one of whom was seriously injured and taken to the hospital for medical treatment[4].”

To summarize, Iranians in Syria used a drone to violate Israeli airspace. The Israelis responded by destroying the drone and the drone’s launch structures in Syria. The Israelis then violated Syrian airspace to attack Syrian and Iranian infrastructure. While doing so, one of their F-16’s was shot down and one of its crew was wounded. All of this has occurred, yet Iran and Israel have not declared war in response.

Incidents like this are so common that it is easy to overlook the miraculous fact that while such incidents are not “peaceful,” the world does not face open war in response to each of them. There are certainly a variety of reasons for this lack of open war that can be unique to each situation such as level-headed leaders on either side, mutually assured destruction, war-weary populations, etc. One compelling reason that many share, however, is that each side can claim it was acting in self-defense while not being able to convince the international community and United Nations Security Council that this is true. In this above example, Israel could claim that it was attacked when the Iranian drone entered its airspace so its response was in self-defense. Iran and Syria could claim that their drone was unarmed and entered Israeli airspace accidentally. Israel then attacked them and they downed an Israeli aircraft in self-defense. This familiar dance occurs in other comparable situations: opposing sides take limited aggressive actions towards each other but generally stop short of open war. Article 51 doesn’t eliminate conflict, but prevents it from escalating or at least escalating quickly.


Endnotes:

[1] United Nations. (n.d.). Charter of the United Nations: Chapter VII. Retrieved February 27, 2018, from http://www.un.org/en/sections/un-charter/chapter-vii/

[2] Glennon, M. (2018, February 13). ILO L201: Public International Law [Class discussion]. The Fletcher School of Law and Diplomacy at Tufts University, Medford, Massachusetts, U.S.

[3] IDF intercepts Iranian UAV. (2018, February 10). Retrieved February 28, 2018, from https://www.idf.il/en/minisites/press-releases/idf-intercepts-iranian-uav/

[4] IDF intercepts Iranian UAV. (2018, February 10). Retrieved February 28, 2018, from https://www.idf.il/en/minisites/press-releases/idf-intercepts-iranian-uav/

Assessment Papers Governing Documents Jared Zimmerman United Nations

Assessment of Opération Turquoise: The Paradoxical French-led Humanitarian Military Intervention During the 1994 Genocide in Rwanda

Ross Conroy is a researcher and program designer for Komaza Kenya, a social enterprise focused on poverty reduction through sustainable timber production.  Ross also serves as Public Relations advisor for Sudan Facts, a start-up which intends to build investigative journalistic capacity in Sudan.  Ross studied Political Science at the University of New Hampshire, and wrote his capstone on the French military intervention during the Rwandan Genocide.  He spent most of his senior year in Rwanda doing field and archival research to supplement this study.  Ross later attained his Master’s degree in African Studies from Stanford University, where he focused on politics in Central Africa and continued his research on French involvement during the 1994 Genocide.  Ross can be found on Twitter @rossconroy or at rconroy7@outlook.com.  Divergent Options’ content does not contain information of an official nature nor does the content represent the official position of any government, any organization, or any group.


Title:  Assessment of Opération Turquoise: The Paradoxical French-led Humanitarian Military Intervention During the 1994 Genocide in Rwanda

Date Originally Written:  February 12, 2018.

Date Originally Published:  May 14, 2018.

Summary:   In response to escalating genocidal violence in Rwanda in April 1994, France launched Opération Turquoise for ostensibly humanitarian purposes.  However, much evidence has implicated this mission, and France, in the genocide and subsequent violence.  By examining archives, interviewing genocide survivors, and compiling testimonies of French soldiers, a more clear, and far more sinister, picture of Opération Turquoise emerges.

Text:  The French-led Opération Turquoise, mobilized by the United Nations (UN) Security Council through Resolution 929, was controversial from its genesis.  The debate leading up to the final vote on the resolution was riddled with arguments about France’s true intentions.  Having been the main sponsors of the Hutu regime that was now organizing and perpetrating genocide against the Tutsi minority, an abrupt change in France’s policy was viewed with suspicion.  Publicly, France argued that violence in Rwanda had escalated to the point that it necessitated international intervention on humanitarian grounds.  The wording of the resolution seemed to confirm this, stating that the mission was “aimed at contributing, in an impartial way, to the security and protection of displaced persons, refugees and civilians at risk in Rwanda[1].”  However, due to concerns over France’s intentions, and the proposed departure from the Chapter VI mandate of the United Nations Assistance Mission for Rwanda (UNAMIR) and UNAMIR II, five countries abstained from the final vote[2].  Rather than supplying and funding the existing UNAMIR mission, France wanted a Chapter VII mandate over which they had near complete jurisdiction.  When the mission was eventually condoned by the Security Council, France mobilized their force and, as some countries had feared, used it to promote their interests.  Rather than protecting Tutsis from the genocidal regime, Opération Turquoise was co-opted to allow the perpetrators to continue their campaign of violence and eventually escape the Rwandan Patriotic Front (RPF) advance by fleeing into neighboring Zaire[3].  The results of this would prove disastrous.

The origins of Opération Turquoise were rooted in the close ties between then French President François Mitterrand and his Rwandan counterpart Juvénal Habyarimana.  The Technical Military Assistance Agreement signed between the two countries after Habyarimana came to power in the 1970s solidified this relationship by formally incorporating Rwanda into the linguistic and cultural sphere of la francophonie and promising economic aid and military protection[4].  Following this agreement, military aid was passed for decades to the Rwandan army and its militias directly through the Quai D’Orsay, the French Ministry of Foreign Affairs.  These same weapons were later put to use exterminating the Tutsi minority.

In October 1990, Rwanda was invaded by Rwandese rebels from Uganda who wanted a return to their country of origin.  Due to xenophobic policies of the Rwandan regime, these Rwandese rebels had long been denied this opportunity.  The French, fearing an incursion from what they saw as ‘the anglophone bloc,’ rushed to the aid of their francophone ally.  The rhetoric of French politicians at the time indicated that the Fashoda Syndrome, the inherent French fear of francophone influence being supplanted by anglophone influence, motivated France to support the Rwandan regime, and solidified ties between the two countries further[5].

After the assassination of President Habyarimana in April 1994, it quickly became clear that the violence engulfing Rwanda was of an unspeakable magnitude; a genocide was unfolding.  Despite this, France declined to act, not wishing to aid the RPF in the fight against their erstwhile ally, the Rwandan regime.  As it became clear that the Rwandan government was failing in its fight against the RPF, France chose to intervene under the guise of a much-needed humanitarian mission.  The UN had little choice, given the dearth of alternatives, and accepted the French offer of assistance.

The French originally conceived the mission as a means to halt the advance of the RPF militarily and assist the government of Rwanda in retaking the capital, Kigali[6].  However, it soon became evident that this position was untenable as the evidence of genocide mounted.  The goal of Opération Turquoise was thus altered to aid the Rwandan government forces in fleeing the country to Zaire, with the intention to preserve the government and its hierarchy intact to pursue future power-sharing agreements[7].  It was a final, frantic attempt to avoid losing influence in the region, and one that would have devastating consequences.

In the process of assisting the genocidal regime, France often neglected to protect the Tutsi whom they were charged with safeguarding.  There is irrefutable evidence that the French demonstrated gross negligence of their mandate by abandoning thousands to die in various locations around the country, most notably at Bisesero[8][9].  Indeed, numerous reports cite French soldiers trading sexual favors for food and medical supplies, raping, and even killing Rwandan citizens[10].  The French further neglected to disarm Rwandan troops and militias whom they escorted to Zaire, and in some cases supplied them with food, weapons, and vehicles[11].  These same Rwandan forces would later profiteer in the Zairean refugee camps, syphoning humanitarian aid intended for victims of genocide.  As the refugee camps were often not the internationally required 50 miles from the border of Rwanda, the ex-Rwandan Armed Forces and militias were able to use the camps as bases and launch a devastating and deadly insurgency back into Rwanda, killing thousands[12].  In response to the insurgency, and renewed killings of Tutsi in Zaire, the new RPF-led Rwandan government invaded Zaire, setting in motion the Congo Wars, the most deadly series of conflicts worldwide since the two World Wars.  Years of suffering, disease, and death can be traced back to the decision made by the French to escort the génocidaires to Zaire and continue to supply and support them in a vain attempt to cling to their influence in the region.

Ultimately, Opération Turquoise failed on two fronts: It failed to maintain the integrity and legitimacy of the former Rwandan regime and also failed to uphold its mandate to protect victims of genocide.  Although it is impossible to establish a direct causal relationship between violence in the Great Lakes Region following Opération Turquoise and the 1994 Genocide against the Tutsis, there is ample evidence that Opération Turquoise exacerbated the humanitarian situation.  Opération Turquoise, conceived as a humanitarian mission, thus paradoxically contributed to one of the worst humanitarian disasters in modern history in the Eastern Democratic Republic of the Congo.


Endnotes:

Note: Some names have been changed to protect the identities of interview subjects. 

[1] United Nations Security Council (SC), Resolution 929. (1994, June 22). Opération Turquoise. Retrieved April 22, 2016 from https://documents-dds-ny.un.org/doc/UNDOC/GEN/N94/260/27/PDF/N9426027.pdf?OpenElement

[2] Schweigman, D. (2001). The Authority of the Security Council under Chapter VII of the UN Charter: Legal Limits and the Role of the International Court of Justice. The Hague: Kluwer Law International.

[3] Twenty Years after Genocide France and Rwanda Give Different Versions of History. (2014, April 11) Retrieved February 28, 2016 from http://www.english.rfi.fr/africa/20140410-twenty-years-after-genocide-france-and-rwanda-give-different-versions-history

[4] Totten, S, and Sherman, M. (2005). Genocide at the Millennium. New Brunswick, NJ: Transaction Publishers.

[5] Simon, M. (1998) Operation Assurance: The Greatest Intervention That Never Happened. The Journal of Humanitarian Assistance. Retrieved April 12, 2016 from http://sites.tufts.edu/jha/archives/123.

[6] Melvern, L. (2009). A People Betrayed: The Role of the West in Rwanda’s Genocide. London: Zed.

[7] Mukasarasi, J. (2016, April 17). Personal Interview.

[8] Assemblée Nationale. (1998, December 15). Rapport d’information de MM. Pierre Brana et Bernard Cazeneuve, déposé en application de l’article 145 du Règlement par la mission d’information de la commission de la Défense, sur les opérations militaires menées par la France, d’autres pays et l’ONU au Rwanda entre 1990 et 1994. Paris: French National Assembly.

[9] De Vulpian, L & Prungnaud, T. Silence Turquoise. Paris: France. Don Quichotte.

[10] Mvuyekure, A. (2016, April 16). Personal Interview.

[11] Des Forges, A. (1999). Leave None to Tell the Story: Genocide in Rwanda. New York: Human Rights Watch.

[12] Gribbin, R. (2005). In the Aftermath of Genocide: The U.S. Role in Rwanda. Lincoln, NE: IUniverse.

Assessment Papers France Mass Killings Ross Conroy Rwanda United Nations