Historical case studies of technology governance and international agreements (compilation – various authors)

By AI Safety Fundamentals Team (Published on August 29, 2023)

The following excerpts summarize historical case studies that are arguably informative for AI governance. The case studies span nuclear arms control, militaries’ adoption of electricity, and environmental agreements. (For ease of reading, we have edited the formatting of the following excerpts and added bolding.)

Case studies of arms control and nonproliferation agreements

“International Control of Powerful Technology: Lessons from the Baruch Plan for Nuclear Weapons” (Zaidi and Dafoe, 2021) [Excerpts]

1. Introduction

[...] This paper looks to the development and attempted governance of nuclear technology for lessons [...]. It focuses on the uncertain early years of this technology, especially 1943 to 1951. [...] This dual-use nature led [stakeholders] to conclude that [...] the world needed to devise international governance mechanisms which would both reduce the risks but also allow the beneficial outcomes to emerge. [4]

[...] In summary, we find that radical schemes for international governance can get widespread support, even from skeptics, but that the support can be tenuous and fleeting. Technical experts can bolster support, but muddled policymaking, secrecy, and concerns over security can undermine it. [...]

Nuclear technology as an analogy

[...] [N]uclear technology stands out as a particularly promising candidate for study of the pressing, but thorny, problem of international control. In particular, we [8] would highlight the following properties which make this case relevant to understanding efforts to control a future powerful technology (such as AI):

  1. Nuclear technology was marked out as a powerful technology when it was first revealed, and policymaking was made within a context that took its potential impact seriously.
  2. Because of the sudden way the atomic bomb was revealed, and its seemingly esoteric nature, there was significant uncertainty about its impact. Consequently, there was a rich public and policymaking debate about the nature of this technology and its impact. As well as strategic and political dimensions, this debate included an ethical dimension.
  3. Many people, including many elites, perceived nuclear technology as an existential risk and so [9] engendered a rich policymaking debate on international governance, known then as “international control.”
  4. Elements of national competition and negotiation, and of a technological arms race, were present during the early history of nuclear weapons.
  5. Nuclear technology rested on complex, rapidly developing science.

Nevertheless, [...] Consider the following disanalogies to AI:

  • Private Sector Involvement: AI is primarily being developed and deployed by the private sector, and the private sector is likely to continue to push forward the science of AI irrespective of what governments do. Nuclear technology in its earliest decades was controlled and funded by states.
  • Secrecy: While nuclear technologies were heavily guarded secrets (though the basic science was broadly known), artificial intelligence technology is more international, broadly held, and public.
  • Impact and Proliferation: AI is already a major economic technology and deployed around the world, whereas the economic value of nuclear technology was unclear and it was deployed in only a few locations in the period 1943–1951. AI is deployed and innovated in a greater number of fields as compared to nuclear technology, and offers greater future economic potential. The barriers to entry for the development and deployment of AI are also lower than in the case of nuclear technology.
  • Discernibility of Risks: It may be easier to understand how nuclear weapons could be dangerous, whereas the accident risks from AI are more subtle, theory dependent, or fantastic seeming.
  • Safety Difficulty: The accident risks from nuclear weapons are likely easier to manage than from AI, because nuclear bombs or power plants are not complex adaptive (intelligent) systems.
  • Verification: It is easier to unilaterally verify nuclear developments (nuclear tests, ICBM deployments), and it appears easier to control the nuclear supply chain with relatively low disruption of industry.
  • Strategic Value: The strategic value of nuclear weapons plateaus once one has secure second-strike capability, whereas from the present vantage point, there is no obvious plateau in AI’s strategic value.

Further, the historical context for the early development of nuclear technology differs in important ways [...] Postwar Context: Atomic development occurred at the end of a war widely seen as catastrophic. This led to a very different social and political context within which nuclear weapons were introduced. Visceral Example of Danger: The world witnessed the use of nuclear weapons to destroy cities and some of the horrors this entailed. Future technology risks may not produce visceral harms in advance of attempts to govern them. [...]

4.1 Serious Radical Proposals


Radical proposals which would normally appear naive or extreme may, in the right circumstances, be seriously proposed, discussed, and even adopted as official policy. Two conditions are conducive to this. First, if the emergent technology is spectacularly disruptive, it can expand the realm of politically feasible policies. Second, a sense of rupture or crisis in international political affairs can make otherwise unrealistic proposals more acceptable and possible.

Historical Case

Proposals for international control of atomic energy were radical for their time. They proposed that states be bound by powerful and wide-ranging multilateral agreements and that powerful international organizations be created with the power to police such agreements. In both these senses, these proposals were much more radical than other serious (that is, taken up at the diplomatic level) discussions on international governance at the time. The United Nations charter, for example, did not create obligations on states as binding, or intrusive on national soil, as some envisaged in the Acheson-Lilienthal Report. Similarly, no other armaments were subject [29] to such proposals in the 1940s. [30]

The Acheson-Lilienthal Plan, for example, proposed that a powerful new U.N. Atomic Development Authority (ADA) would set up large R&D centers and conduct research on peaceful and warlike uses of atomic energy. It would own and operate all mining, refining, and production of fissionable raw materials—including having its own operational reactors—and dispense fissionable raw materials to nations for their nuclear power plants. It would also license and inspect operating civilian nuclear facilities in nation states. [31]

These proposals were taken seriously by many of their proponents and much of the public. [...]

There are several conditions that allowed the Acheson-Lilienthal Report to gain traction. [Formating added.]

  1. The report was accepted because of a growing public and elite perception of the threat that atomic weapons posed. It was widely believed in late 1945 and 1946 that atomic weapons would be used in any future major war, wiping out cities and killing millions. [...]
  2. The recent experience [34] with an awful war increased public and elite receptiveness to radical political proposals. There was a sense that regular politics had failed, and more radical measures were required.
  3. The need for postwar reconstruction and the growth of U.S. influence globally made possible new initiatives in international relations which were not possible before. The formation of the United Nations and other international organizations (e.g., the Bretton Woods system to manage the global economy) were widely welcomed.
  4. Scientists who worked on [35] atomic matters generally threw their weight behind the Acheson-Lilienthal Report and formed powerful organizations that advocated for international control. Their status and newfound prominence gave their message traction in the media and in government. [36]

“Ch. 1: The Nuclear Non-Proliferation Regime,” from Deterring Nuclear Proliferation (Rosenthal and Stern, 2019) [Excerpts]


The most important, and the most difficult, step in manufacturing a nuclear weapon is to acquire the necessary nuclear-weapon-usable nuclear material. The possible materials are called fissile materials and are characterized by their ability to sustain a nuclear chain reaction, which is the mechanism by which the nuclear explosion’s energy is generated. The two fissile materials used in nuclear weapons currently deployed are high-enriched uranium (HEU) and plutonium. [...] For our purposes, we assume that if the production and use of these fissile materials can be controlled, the risk of proliferation will be reduced. [...] As we will see below, many aspects of the nuclear nonproliferation regime have been directed towards this end.

Nuclear Proliferation – A Status Report:

Most developed nation-States have the means to acquire nuclear weapons, and more than a few have considered doing so. [...] Nuclear weapons are within the reach of States whose technical and industrial infrastructures are underdeveloped. The DPRK [North Korea], one of the poorest and least developed countries in the world, first tested a nuclear weapon test in 2006 and has conducted a series of tests since then. [...]

In 1960, President John F. Kennedy worried that, “There are indications because of new inventions, that 10, 15, or 20 nations will have a nuclear capacity, including Red China, by the end of the Presidential office in 1964. This is extremely serious . . . I think the fate not only of our own civilization, but I think the fate of world and the future of the human race is involved in preventing a nuclear war.”

Despite the widespread availability of technology unknown in the 1960s, the number of States today that have nuclear weapons is smaller than President Kennedy thought likely. [As of 2023, only 9 states have nuclear weapons, and only 3 of them acquired nuclear weapons after the Non-Proliferation Treaty entered into force in 1970. Four states—South Africa, and three states that inherited nuclear weapons when the Soviet Union dissolved—even willingly disposed of all their own nuclear weapons.]

In 2004, Libya voluntarily abandoned a nuclear-weapon program that it was pursuing in violation of its NPT safeguards obligations. Other States with nuclear-weapon ambitions have had them thwarted. In accordance with a United Nations Security Council resolution after the first Gulf War in 1991, the elements of Iraq’s nuclear-weapon program were “removed, destroyed, or rendered harmless.” In 2007, Israel destroyed a reactor in Syria from the air. Senior United States officials reported that this reactor, which was being built secretly with assistance from the DPRK, would have been capable of producing plutonium for nuclear weapons. In 2011, it was reported that an effort was made to stop or restrain the Iranian uranium enrichment program using cyber warfare.

[...] Why have so few States proliferated when so many have the capability to do so? [...] The answers influence views about what tools to use to reduce the risk of proliferation. [...] [M]any different tools have been developed and coexist with one another, and the answer might differ from State to State, with the result that in specific instances, more than one of the tools available might be emphasized.

Tools Available to Reduce the Prospect of Proliferation


Addressing capabilities

Secrecy and denial: The U.S. Atomic Energy Act of 1946 created a system to control information related to nuclear weapons. [...] Except for declassified information, it covered all data concerning the manufacture or utilization of atomic weapons, the production of fissionable material, or the use of fissionable material in the production of power. Secrecy was intended to prevent other countries from proliferating, especially the Soviet Union. It failed to do so because the Soviet Union had pierced the veil of U.S. secrecy during the war and had many capable scientists and engineers who could exploit the information. In addition, what had been the greatest secret, that one could make a weapon, was exposed at Hiroshima.

In general, secrecy and denial are waning assets. One reason is that they impede nuclear cooperation. [...] In addition, information or technology that is not readily available at one time may become readily available later as States industrialize, the pace of technology development quickens, and information becomes global. [...]

Export controls: [...] Export controls offer the opportunity to cooperate selectively with partners where the risk of proliferation is perceived to be low and to deny export to countries where the risk is perceived to be too high. Even when cooperation is pursued, criteria for supply may be used to reduce risks even further. The United States insists that specialized nuclear cooperation be allowed only under Agreements for Cooperation. Under such agreements, a recipient country agrees, for example, not to use material supplied by the United States for any nuclear explosive device or for any other military purpose; to accept international verification; and to obtain the approval of the United States before it reprocesses, enriches, or transfers nuclear material subject to the agreement. [...]

The concept of export controls is built into the NPT [Nuclear Non-Proliferation Treaty]. It stipulates that nuclear material and especially designed equipment and material can only be exported when IAEA safeguards are applied in the recipient State. In order for export controls to be effective, all relevant suppliers need to apply the same ground rules. To allow buyers to shop for the weakest non-proliferation condition would undermine the control system. In order to achieve common non-proliferation objectives and create a level playing field, likeminded States have joined together to create multilateral organizations. [...] Unfortunately, illicit trafficking in nuclear equipment has undermined the effectiveness of export controls; some States have resorted to illegal and clandestine procurement practices; and some States have become suppliers that are not scrupulous about non-proliferation requirements.

Multinational facilities: If proliferation decisions depend on the availability of sensitive technologies, especially enrichment and reprocessing facilities, then nuclear fuel cycles that depend on these technologies have a technical risk. A number of means have been proposed for ensuring that nuclear material is available for peaceful purposes without increasing the number of countries that have national enrichment or reprocessing facilities. These include: continued reliance on a robust market that depends on present suppliers; a “fuel bank” that backs up this market by providing an assured supply of nuclear fuel in the event of a supply disruption not based on proliferation (see Section 3.3 below); and the development of multinational facilities where the technology holder does not share the technology with partners. In principle, these steps can enhance supply without spreading technology. [...]

Addressing incentives

Some means of reducing the risk of proliferation do not rely at all on limiting either a State’s technical capabilities or the availability of the ingredients for making a bomb.

Security alliances: If States’ national security interests are satisfied without possession of nuclear weapons, the incentive to acquire them is absent. One means of doing this is through security alliances. For example, the North Atlantic Treaty Organization (NATO) and the U.S. security alliance with Japan provide an environment in which U.S. partners have chosen not to pursue nuclear-weapon acquisition. Security assurances can also be provided to reduce incentives to proliferate even where an alliance is absent. Negative security assurances are guarantees from nuclear-weapon States that they will not use nuclear weapons against non-nuclear-weapon States parties to the NPT. (See “Security Assurances” in Section 4.3.3 below.)

Sanctions: Sanctions – diplomatic, economic, or military – may be employed to deter proliferation by threatening to impose penalties on States. The goal is to deter non-compliance with nuclear nonproliferation norms or obligations. In the area of nuclear non-proliferation, sanctions are typically associated with violations of nuclear non-proliferation agreements, both bilateral agreements and international treaties. For example, U.S. nuclear Agreements for Cooperation contain provisions that cancel such cooperation in the event that a partner violates a safeguards agreement or tests a nuclear weapon. Sanctions may also be imposed by the United Nations Security Council. They can range from travel restrictions on individuals to economic embargoes. The Security Council may also authorize the use of blockades or other use of armed force.

Safeguards: The U.S. drafted Acheson-Lilienthal Report concluded that the fuel cycle should be internationalized, and an international inspection system put in place. [...] While internationalization of the fuel cycle has not taken place, the concept of an inspection system with early warning of diversion became a part of the IAEA safeguards system. It was incorporated into and made explicit in NPT safeguards agreements. [...] IAEA safeguards were also applied prior to entry into force of the NPT. These safeguards were applied to individual facilities, nuclear material and other items that were specified in the agreements. They also covered any nuclear material produced through the use of these items. These safeguards agreements were concluded in connection with exports where the supplier required safeguards as a condition of supply.

Nuclear-Weapon-Free Zone Treaties: Although the NPT is the primary international nuclear non-proliferation agreement, regional or multinational agreements can also enhance security and reduce the risk of proliferation. Nuclear-weapon-free zones (NWFZs) are important examples of regional frameworks for this purpose. Such treaties are in force in Africa, Central Asia, Latin America and the Caribbean, the South Pacific, and Southeast Asia. [...] Several other treaties also establish NWFZs: in Antarctica [...] in outer space [...] and on the seabed and ocean floor [...]

“The Pope and the Crossbow: The Mixed History of Arms Control,” from Army of None (Scharre, 2018) [Excerpts]

In the summer of 2015, a group of prominent AI and robotics researchers signed an open letter calling for a ban on autonomous weapons. “The key question for humanity today,” they wrote, “is whether to start a global AI arms race or to prevent  it  from  starting.  If  any  major  military  power  pushes  ahead  with weapon development, a global arms race is virtually inevitable.” [...]

Why Some Bans Succeed And Others Fail

[...] The number of countries that need to participate for a ban to succeed is [...] a critical factor. Arms control was easier during the Cold War when there were only two great powers. [...]

[...] Legally-binding treaties have been routinely violated and restraint has  existed  in  some  cases  without  any  formal  agreements.  International Agreements,  legally  binding  or  not,  primarily  serve  as  a  focal  point  for coordination. What actually deters countries from violating bans is not a treaty, since by default there are no legal consequences if one wins the war, but rather reciprocity. Countries show restraint when they fear that another country might retaliate in kind. When fighting nations who do not have the ability to retaliate, they have shown less restraint. For example, during World War II Japan used chemical weapons in small amounts against China, who did not have them, and Germany  killed  millions  of  people  in  gas  chambers  during  the  Holocaust. Neither country used poison gas against adversaries who could retaliate in kind.

For  mutual  restraint  to  occur,  there  must  be  a  clear  focal  point  for coordination. In his books Strategy of Conflict and Arms and Influence, Thomas Schelling  explained  that  “the  most  powerful  limitations,  the  most  appealing ones, are those that have a conspicuousness and simplicity, that are qualitative and  not  a  matter  of  degree,  that  provide  recognizable  boundaries.”  Schelling Observed: “Some gas” raises complicated questions of how much, where, under what circumstances: “no gas” is simple and unambiguous. Gas only on military personnel; gas used only by defending forces; gas only when carried by vehicle or projectile; no gas without warning—a variety of limits is conceivable; some may make sense, and many might have been more impartial to the outcome of the war. But there is a simplicity to “no gas” that makes it almost uniquely a focus of agreement when each side can only conjecture at what rules the otherside would propose and when failure to coordinate on the first try may spoil the chances for acquiescence in any limits at all.

This simplicity undoubtedly played a role in making it possible for European Nations to refrain from using poison gas against each other in World War II, in spite of a total war that devastated the continent. [...] Germany  and  the  United  Kingdom  also  attempted  to  mutually  avoid bombing attacks on civilian targets. These failed [...] The chief difference between aerial bombardment and gas, and what made restraint with aerial bombardment so difficult, is that restraint against civilian targets  lacked  the  clarity  and  simplicity  of  the  “no  gas”  rule. [...]

Treaties  that  completely  ban  a  weapon  tend  to  be  more  successful  than complicated  rules  governing  a  weapon’s  use.  Other  attempts  to  regulate  how weapons are used on the battlefield in order to avoid civilian casualties—such as restrictions on submarine warfare, incendiary weapons, and the CCW landmine protocol—have had a similarly poor track record of success. Complete bans on weapons—such  as  those  on  exploding  bullets,  expanding  bullets,  chemical weapons,  biological  weapons,  environmental-modification  weapons,  and blinding lasers—have fared better.

[...] Carving out exceptions can make it easier to get more countries to sign on to a ban, but can be problematic if the technology is still evolving. One lesson from history is that it is very hard to predict the future path of technology. [...] Successful  preemptive  bans focus  on  the  intent  behind  a  technology,  rather  than  specific  restrictions.  For example,  the  ban  on  blinding  lasers  prohibits  lasers  specifically  designed  to cause permanent blindness, rather than limit a certain power level in lasers. The United States takes a similar intent-based interpretation of the ban on expanding bullets, that they are prohibited only to the extent that they are intended to cause unnecessary suffering.

Preemptive bans pose unique challenges and opportunities. Because they are not  yet  in  states’  inventories,  the  military  utility  of  a  new  weapon,  such  as blinding  lasers  or  environmental  modification,  may  be  amorphous.  This  can sometimes make it easier for a ban to succeed. States may not be willing to run the risk of sparking an arms race if the military utility of a new weapon seems uncertain. On the other hand, states often may not fully understand how terrible a weapon is until they see it on the battlefield. [...]


One topic that frequently arises in discussions about autonomous weapons is the role of verification regimes in treaties. Here the track record is mixed. A number of  treaties,  such  as  the Nuclear  Non-Proliferation Treaty, Chemical  Weapons Convention, INF Treaty, START, and New START have formal inspections to verify compliance. Others, such as the Outer Space Treaty’s prohibition against military installations on the moon, have de facto inspection regimes. The landmine and cluster munitions bans do not have inspection regimes, but do require transparency from states on their stockpile elimination.

Not  all  successful  bans  include  verification.  [...] Inspection regimes are not always essential. What is required is transparency. Countries need to know whether other nations are complying or not for mutual restraint to succeed. In some cases, the need for transparency can be met by the simple fact that some weapons are difficult to keep secret. Anti-ballistic missile facilities and ships cannot be easily hidden. Other weapons can be.

Why Ban?

Finally, the motivation behind a ban seems to matter in terms of the likelihood of success. Successful bans fall into a few categories. The first is weapons that are perceived to cause unnecessary suffering. By definition, these are weapons that harm combatants excessive to their military value. Restraint with these weapons is self-reinforcing. Combatants have little incentive to use these weapons and strong incentives not to, since the enemy would almost certainly retaliate.

Bans on weapons that were seen as causing excessive civilian harm have also succeeded,  but  only  when  those  bans  prohibit  possessing  the  weapon  at  all (cluster munitions and the Ottawa landmine ban), not when they permit use in some  circumstances  (air-delivered  weapons,  submarine  warfare,  incendiary weapons, and the CCW landmine protocol). Bans on weapons that are seen as destabilizing  (Seabed  Treaty,  Outer  Space  Treaty,  ABM  Treaty,  INF  Treaty) have  generally  succeeded,  at  least  when  only  a  few  parties  are  needed  for cooperation. Arms limitation has been exceptionally difficult, even when there are only a few parties, but has some record of success. Prohibiting the expansion of  war  into  new  geographic  areas  has  only  worked  when  the  focal  point  for cooperation is clear and there is low military utility in doing so, such as banning weapons on the moon or in Antarctica. Attempts to regulate or restrict warfare from  undersea  or  the  air  failed,  most  likely  because  the  regulations  were  too nuanced.  “No  submarines” or  “no  aircraft”  would  have  been  clearer,  for example.

Ultimately, even in the best of cases, bans aren’t perfect. Even for highly successful  bans,  there  will  be  some  nations  who  don’t  comply.  This  makes military utility a decisive factor. Nations want to know they aren’t giving up a potentially war-winning weapon. This is a profound challenge for those seeking a ban on autonomous weapons.

Case studies of general-purpose technology governance

Engines of Power: Electricity, AI, and General-Purpose Military Transformations (Ding and Dafoe, 2021) [Excerpts]

Abstract: Major theories of military innovation focus on relatively narrow technological developments, such as nuclear weapons or aircraft carriers. Arguably the most profound military implications of technological change, however, come from more fundamental advances arising from “general purpose technologies” (GPTs), such as the steam engine, electricity, and the computer. With few exceptions, political scientists have not theorized about GPTs. Drawing from the economics literature on GPTs, we distill several propositions on how and when GPTs affect military affairs. We call these effects “general-purpose military transformations” (GMTs). In particular, we argue that the impacts of GMTs on military effectiveness are broad, delayed, and shaped by indirect productivity spillovers. Additionally, GMTs differentially advantage those militaries that can draw from a robust industrial base in the GPT. To illustrate the explanatory value of our theory, we conduct a case study of the military consequences of electricity, the prototypical GPT. Finally, we apply our findings to artificial intelligence, which will plausibly cause a profound general-purpose military transformation. [...]

GMT Feature 1: Broad Impact Pathway

The main applications of electricity in the late nineteenth century were in lighting (searchlights), electric power (electric handling of weapon systems), and communications (the telephone and telegraph). [...] The electric firing of guns was another key component of military transformation. [...] Advances in electricity fed into telegraph and telephone communications, which enabled a military system that could coordinate mass armies. [...] Eventually, the electromagnetic spectrum became a new arena for conflict [through radar jamming and radar-based spying].

[...] For Electrical World, the mayor of Kansas City captured the accumulated impact of electricity across a broad range of systems by 1890: “[Electricity] now not only guards the vessel from the inventions of the enemy, but aims and fires the guns, illuminates the sights that the aim may be sure, discharges torpedoes, measures her speed, is the most successful motor for submarine boats, and renders possible a system of visible telegraphy by which communications may be flashed against the clouds and understood at a distance of sixty miles.”

[...] the foreseeability of electricity’s impact on the conduct of warfare should have been very limited. As advances in electricity were emerging, many predicted that the main application in the military realm would come in the form of war-winning weapons. [...]

GMT Feature 2: Industrial Productivity Spillovers

In addition to spurring a variety of military innovations, a GPT should also transform military effectiveness by boosting industrial productivity. In the case of electricity, it is well-documented that the diffusion of electricity across manufacturing industries resulted in a productivity surge. [...] Crucially, electrification enabled mass production, as the adoption of electric unit drive in factories resulted in standardized workflows and plant capacity expansion. [...]

The impact of electrically-boosted production capabilities was revealed in the two great wars of the 20th century. [...] For example, Britain possessed only 154 airplanes at the outbreak of WWI, but British aircraft factories were producing 30,000 planes per year by the end of the war.[97] Access to cheap, plentiful electricity drove these surges. [...]

GMT Feature 3: Delayed Effects

In the economic realm, scholars hold up the diffusion of electricity as an example of the long time lag between the emergence of a GPT and its significant bearings on national productivity. [...] American electrification did not take off until the 1920s.[106] This was a full four decades after major advances like the dynamo and incandescent light bulb emerged.

Do we find a similarly delayed timeline for the electrification of the military? [...] even the early movers did not achieve widespread adoption until right before WWI. Among later adopters, widespread adoption did not take place until the interwar period or after WWII. [...] Indeed, some of the most significant military applications of electricity did not emerge until the 1940s. [...] The delayed diffusion of military electrification was due to the need for significant organizational adaptations and skills upgrading. [...] the U.S. Navy took about fifteen years to fully integrate the radio into its operations, as senior naval officers saw the radio as a direct threat to their authority onboard ships. [...]

IV. Conclusion and Lessons for AI

[...] speculation about how AI will transform military affairs places excessive emphasis on the

narrow effects of weapon systems. [...] In contrast, a GMT approach emphasizes the accumulation of AI-enabled improvements across many military systems. This impact pathway will likely involve significant upgrades to weapons capabilities, as was the case with electricity and centralized fire control. On the whole, though, effects of AI advances in other military domains, including communications, cyberspace operations, intelligence, information and psychological operations, logistics, strategic decision-making, etc. will be more consequential. [150] Moreover, the focus on AI weapons neglects the indirect pathway of influence through AI’s potential to upgrade the productive capabilities of the overall industrial base. In particular, the intelligentization of manufacturing lines (smart manufacturing) could have significant follow-on effects for military readiness.

Second, existing conjectures about the impact of AI on military affairs severely underestimate the timeframe for when substantial effects will occur [...]

We highlight the significance of a state’s industrial capacity to provide AI infrastructure and skilled labor to militaries. Specifically, militaries able to draw from a wide skill base in AI will better exploit the AI-based GMT. [...]

Case studies of environmental agreements

“Of Montreal and Kyoto: A Tale of Two Protocols” (Sunstein, 2007) [Excerpt]

Over the last thirty years, climate change and depletion of the ozone layer have been widely believed to be the world's largest environmental problems. The two problems have many similarities. [...] But an extraordinarily successful agreement, the Montreal Protocol, has served largely to eliminate the production and use of ozone-depleting chemicals, while the Kyoto Protocol has spurred only modest steps toward stabilizing greenhouse gas emissions. What accounts for the dramatic difference between the two protocols? Part of the explanation lies in the radically different self-interested judgments of the United States; part of the explanation lies in the very different payoff structures of the two agreements.

Influenced by the outcome of a purely domestic cost-benefit analysis involving reductions in ozone-depleting chemicals, the United States enthusiastically supported the Montreal Protocol. Influenced by the very different outcome of cost-benefit analyses for reductions in greenhouse gas emissions, the United States aggressively opposed the Kyoto Protocol.

An examination of the two protocols suggests that neither agreement fit the simple structure of a prisoner's dilemma, in which a nation gain from an enforceable agreement, gains even more if it is the only nation not to comply while all others do, and lose most if it, and everyone else, pursue their own national self-interest. For the United States, at least, compliance with the Montreal Protocol would have been justified even if no other country had complied; for the United States, and for several other countries, compliance with the Kyoto Protocol would not have been justified even if all other parties had complied. An understanding of the judgments that surround the two protocols indicates that even though moral considerations require the United States to spend a great deal to protect citizens in other nations, and even though such considerations can influence behavior, the nation is unlikely to act in response solely to those considerations.

A general implication is that any international agreement to control greenhouse gases is unlikely to be effective unless the United States believes that it has more to gain than to lose. An illuminating wrinkle, also suggestive of the role of domestic self-interest, is that some European nations, above all the United Kingdom, initially contended that ozone depletion was a greatly exaggerated problem while later calling for strong controls on greenhouse gases.

We use analytics cookies to improve our website and measure ad performance. Cookie Policy.