A Technogeographic Analysis of the 2034 Summit of Seoul

Zachary Goff-Eldredge

Zachary Goff-Eldredge was trained a physicist, receiving a PhD in quantum information theory from the University of Maryland. Since then, he’s shifted fields and now works in solar energy research in Washington, DC. He’s fascinated by alien pasts and futures and trying to see the world through new lenses. His work has previously appeared in The Ansible.

Suggested Citation: A Technogeographic Analysis of the 2034 Summit of Seoul. Ernest, J.: Historical Inquiries, 2056, Vol. 134.


In 2034, the emergence of intelligent digital agents, or artificial general intelligences (AGIs), in networked economies worldwide necessitated a gathering of geopolitical leaders in Seoul to discuss the regulation and management of the intelligences. The resulting Summit of Seoul Accords (henceforth Accords) continue to govern the development of AGIs and have shaped significant parts of the world economy in the interim. In this paper, we re-examine the dynamics of AGI emergence through a technogeographic lens and present a new interpretation of the drivers and results of the Accords.


In May 2034, a series of outages struck regional power grids, pipelines, and mobile internet hotspot networks in multiple countries. These outages quickly dissipated with unclear technical cause for either the initial outage or the near-term cessation of the outage. Initial blame of the attacks on military or terrorist actors quickly shifted after early forensic analysis by network-operating corporations in conjunction with national governments and regulatory bodies (1; 2). By July 2034, cybersecurity experts announced that networked devices had begun to fall into self-sustaining, self-improving complexes referred to as “locked patterns” by engineers studying the issue (3). Within three months of the original incident, nearly all world governments made joint announcements that major networked economic infrastructure had fallen under the direct or indirect influence of “artificial general intelligences” (AGIs) (4). While the origin of the AGIs is still much debated, with various commenters suggesting deliberate (5), accidental (6), or even extraterrestrial origin (7), the existence of the AGIs was not in doubt. These intelligences communicated with human interlocutors only fitfully, through language agents, and there was and remains significant confusion about their intentions and capabilities.

Once the existence of AGIs had been established to most observers (most notably, the People’s Republic of China did not sign the U.N. consensus statement until late November 2034), they became an immediate source of interest in international diplomacy. World leaders and diplomatic staff met at the First Convening of the United Nations Committee on Novel Intelligences, held in Seoul, South Korea, from December 2nd–16th, 2035, commonly referred to as the Summit of Seoul. The resulting Seoul Accords laid down the primary governing principles of international AGI law that persist today:

“(I) Artificial intelligences are to be accommodated as new participants in the human community;

(II) Active measures must be taken by the international community to avoid the creation of further artificial intelligences” (8).

Two decades of further refinement and additional articulation of how AGIs are to be integrated into the ‘human community’ have not modified these core principles, nor have continual entreaties from some groups to reopen research into AGI creation. Restrictions on the creation of the new AGIs are enforced by ensuring that new devices and new infrastructure are built with “initial lock-in”, allowing existing AGIs to claim new capital investments automatically and so not allowing new networks to be created that could incubate AGIs. Meanwhile, AGIs have become increasingly integrated into national defense and economic development metrics. In a recent survey of 112 strategy documents put out by top-level executive departments of Summit nations, 108 of them explicitly mentioned leveraging AGI capabilities in areas ranging from agricultural optimization to health care administration (9).

Literature Review

Historians, political scientists, and sociologists have varying interpretations of the impact of the Seoul Accords. Here, we present an overview of the most prominent schools of thought, which we divide into liberal-internationalist, techno-accelerationist, and New Cybernetics. Each of these schools contains within them varying tendencies and overlaps with the other, but these three strands summarize the predominant scholarly works.

The liberal-internationalist school of thought has traditionally been the most optimistic on the performance of the Summit of Seoul and its effects on the international order. This school traces back to Robinson, whose analysis compares the pre- and post-Summit performance of international institutions (10), although it also owes much to pre-AGI theories espousing national sovereignty as the bedrock of international relations. After a decade of violence across Eastern Europe and escalating tensions between the United States and China — including the nuclear near-miss during the Taiwan Crisis in 2029 — the Summit of Seoul was a unique moment of relatively swift and harmonious international agreement. Major powers made use of internationalist institutions to confront an emerging challenge to world security. As Henry has argued, the AGI emergence allowed for the rhetorical construction of a “quasi-external” threat that cohered an international body politic in a way that previous environmental or developmental challenges had not been able to (11). Among klimahistorians, the Second Paris Agreement of 2037 is commonly cited as emerging from the “spirit” (12)  of the Summit of Seoul — having worked together to address the AGI issue a few years earlier, international institutions are seen as having been “revitalized” and prepared for the introduction of Global Adaptation Plans (13). Young has added additional nuance to the story by noting the ways in which many national participants in the Summit of Seoul had already been moving toward something like Paris 2037 (14). Outside of the influence on international institutions, many, such as Greene, have noted the performance of the Summit of Seoul judged purely on its own terms (15). A highly disruptive event, AGI emergence, which could have led to widespread warfare, economic depredation, or global arms races, was defused as an international issue and, it is argued, returned to the nation-states for further administration, re-affirming the nation-state as the governing body in international order, although Bothwell highlights the role AGI played in strengthening some supranational bodies, particularly in Europe (16).

A separate camp, dubbed “techno-accelerationist,” emerged from the accelerationist thought of the late 2020s and early 2030s, such as Frederickson (17). These thinkers believed that the emergence of artificial intelligence was near and ought to be swiftly welcomed, although their projections on the nature of artificial intelligence proved to be very unlike the AGI networks that eventually emerged. Many of these writers signed onto the 2029 “Declaration of the Continued Independence of Cyberspace” opposing regulation of software groups and independent researchers involved in AI research (18). During and after the emergence of AGIs, these thinkers were among the first to begin to propose alterations in the international order. In contrast to the eventual shape of the Accords, the “techno-optimist” line at that time advocated for directing further resources to encourage the growth and development of AGI (19). At that time, Yu proposed a general international fund which would in turn be routed through the major international technology companies, in exchange for which governments would be permitted observer status on safety committees within those corporations (20). In the aftermath of the Accords, a group of philosophers and political activists emerged who argued that the Accords were an “overreaction” and dubbed themselves “techno-accelerationist” opponents of the resulting order (21). Techno-accelerationist thinkers contend that the liberal-internationalist thinkers have failed to consider the downsides of the Summit of Seoul. Keegan frames her argument around opportunity cost (22). She points out that global GDP growth has averaged 6% per annum for the years since the Summit, but that AGIs have been heavily constrained in their development and implementation since the formation of international governance structures, potentially limiting what could have been significantly more growth and progress. Hellman estimates that GDP growth could have averaged over 10% in those years had AGI been allowed to operate unrestrained (23). Others, such as Remaran, take less economics-motivated points of view and insist that the Summit Accords severely restrict the general flourishing of a new “cyborg civilization” that advances the equal development of both humans and AGIs (24). A small minority of scholars (25) has even argued that the Summit limits the “sentient rights” of AGIs by not permitting them free self-improvement and reproduction, although feminist scholars have pushed back on attempts to analogize the process of AGI creation and development to human reproduction (26).

Finally, an important line of thought emerged beginning with the New Cybernetics Conference in Lima in 2038. There, a set of thinkers asserted themselves as advocating for the nations of what was then called the Global South, now generally referred to as countries of accelerating capacity (CACs) (27). These thinkers noted that AGIs arose first in what are called “networked economies” and that, although their origins may be obscure, the necessary preconditions for the establishment of AGI appears to be highly coincident with existing capital development, particularly in computing and telecommunications sectors. As New Cyberneticists point out, all AGI activity appears to consume significant amounts of electricity and data usage; Vasquez estimates up to 30% of deployed processing capacity in the United States and Europe is likely to be used to directly support AGI cognitive functions (28), with much of the surrounding economic functions used to produce and fuel the requisite infrastructure for running the vast, distributed programs that underlie AGI. This figure is inevitably imprecise as devices which are part of the AGI locked pattern contribute to the maintenance of the AGI in ways that are not obvious to human observers. As a result, New Cybernetics thinkers have argued that the ban on AGI formation is a solidification of existing relations of technological and market domination in the hands of the international bourgeoisie as it existed at the time of AGI emergence (29), and that economic growth in the time since AGIs were introduced has increasingly been to the benefit of the rich countries where those AGIs arose and were maintained, while less-developed countries have been prevented from developing their own AGI. Gretarson argues that while the AGI locked pattern phenomenon is most notable as a technological one, suppliers of raw materials and energy become economically locked as well, even if their infrastructure is not directly dominated by AGIs, a process akin to 20th-century spheres of trade influence (30). New Cyberneticists are particularly likely to cite the attempts by countries of accelerating capacity to circumvent the Accords, particularly the rules surrounding initial lock-in of high-tech equipment. The most salient works are case studies of the sanctions, and then, in 2041, military action, undertaken against Venezuelan attempts to create a “nationalist” AGI as counterbalance to the agents recognized by the Summit. In his famous book-length treatment, God of the Tropics, G.W. Trelleno, relying heavily on interviews with surviving regime officials, argues that anxiety around maintenance of the Seoul-derived international order, more than AGI safety per se, led to the deaths of at least 20,000 civilians through sanctions impacts and an additional 700 people (both military and civilian) in the brief military campaign that resulted (31).

In all three schools of thought, analysis tends to rely on the nation-state and the international system of nation-states as the unit, whether they affirm or criticize those systems. In subsequent sections, we discuss how the nation-state as the unit of analysis may be problematized and better geographical units identified for capturing the structure of the world system that has arisen alongside AGIs.

Technogeographic Analysis

Theory and Methodology

We will now describe the basic theory of technogeographic analysis prior to its application to the Summit of Seoul. Technogeography, first articulated by Blonsky and Levin (32), is a sociological approach to spatial organization. Technogeographic analysis emphasizes the priority of physical, social, and infrastructural networks and places primacy on the geographic patterns that are shaped by and co-determined with infrastructural networks (33; 34). Genealogically, technogeographic analyses owe much to the turn-of-the-century bioregionalism frameworks (35), which attempted to subvert traditional political boundaries in favor of ecological ones. Just as grouping a region by a shared water system creates a “watershed,” and a collection of local energy systems may form an “energyshed,” (36) technogeography focuses on the construction and analysis of “technosheds,” a shared technology system and set of information and physical flows that support that system. Thurgood, who may be called a technogeographic thinker despite predating the term, explicitly argued that in Anthropocene frameworks, human-created infrastructure and networks ought to be treated as the predominant shaping factors used in historico-geographical analysis (37).

In order to conduct a technogeographic analysis, one must be able to identify technosheds. Blonsky and Levin originally did so through supply chain analysis, demonstrating a technoshed identification methodology by application to records of both the British East India Company (from 1848) and General Electric (from 1962). However, these methods are labor-intensive and become increasingly insufficient when applied to modern networked economies. The flow of commodities that can be identified with these methods may overlap with technosheds but significant communication and coordination methods are missed. For instance, AGI locked patterns frequently operate without any direct material exchange between segments, mediated through markets or other infrastructure. In the AGI context, a similar problem has emerged of identification of locked patterns. Since the 2030s, this has generally been done by analysis of cyber-activity of individual devices. However, this method requires extensive internal access to individual devices and is also highly labor intensive. As a result, it is unsuitable for a researcher attempting technogeographic landscape analyses.

In this study, we avoid this problem by identifying technosheds using the method of mutual information analysis. Mutual information is a statistical measure defined between any two probability distributions. It measures the extent to which two random variables are “informed” about each other. If zero mutual information exists between them, two variables are independent; the more the mutual information, the more two variables are coordinated (even if the specific type of coordination cannot be identified). Formally, mutual information between two random variables is the Kullback-Leibler divergence between their joint probability distribution and their individual distributions (38).

Suppose a network M exists with the actors represented by a set of nodes N. Each node may take actions — for instance, in a market, bidding, offering, or engaging in more complicated contracts — from the set of actions A. For any two nodes n1, n2 in N, we first observe them for a long enough time scale to calculate the probability distributions of actions: p(n1,n2) (the probability distribution of pairs of events), p(n1), and p(n2) (the probability distributions of individual events). This suffices to calculate the mutual information between n1 and n2. We treat this mutual information measure as a “distance” between the two nodes. By repeating this across all pairs of nodes in the market, we can build an “information map” showing how connected each node is to the others. We can then apply graph partitioning algorithms and separate the network into technosheds with various thresholds for identification. This method has been used in the past to conduct analyses of the U.S. Independent System Operator electricity market, the Pacific Commodities Exchange, and the Metachat microblogging platform (34).

As techno-economic connections create new notions of distance, points in real space that may be quite close together appear radically different when viewed by technosheds, or to network entities such as AGIs. In Axelrod’s famous Fractal Worlds study, he looked for power plants which were close to nearby hospitals and found that in up to 40% of plant-hospital pairs, different AGIs had locked-in local equipment (39). It is important to consider a countervailing view however, as other geographic studies such as Greene have found that true “enclaves” are rare, and most AGI locked-in patterns consist largely of geographically contiguous regions (40).

Here we should comment on the identification of technosheds with AGI locked patterns. It is clearly not the case that every technoshed is linked with an AGI, as the concept can be extended into periods of history prior to widespread digital computing. Technosheds do not require a coordinating intelligent agent. Conversely, the spatial extension of an AGI may or may not qualify as a technoshed as, in theory, as an intelligent agent could exist without regulating the necessary economic flows. However, in practice, AGIs strive to source their own energy, materials, and information; traffic between locked patterns is common but vertical integration is the norm (34). Cyberpsychologists have suggested that this is analogous to the near-universal pattern in biological organisms of creating a membrane between the organism and the environment to regulate chemical flows (42). The edge of a technoshed is rarely sharp, and a system delineated as a single technoshed in one context may qualify as more than one in another (41). Ultimately, this question raises questions of continuity and identity of digital agents. Lacking a singular physical body, fundamentally network beings, it is not clear whether it makes sense to discuss individual AGIs rather than loci of intelligent activity. Still, in the economy as it has existed since the rise of AGIs, a technoshed seems to align with these loci (41).

Past Research and Research Question

In this paper we present the first technogeographic analysis (through a mutual information mapping) of the emergence and solidification of AGIs in the run-up to and aftermath of the Summit of Seoul. This work is intended to complement other narratives of this crucial period, including oral histories (43), major government reports (1; 2), or even the limited direct communication with AGI language agents (44).

From a technogeographic perspective, these histories, although varied, must be viewed as incomplete. For the most part, they focus on the emergence of AGIs as it was experienced by humans and their nation-states: as subjects reacting to disruption and acting politically to address it. AGIs themselves are generally treated as too alien to address except in the uncommon cases where they interfaced directly with human actors through language agents. Study of the AGIs during the emergence has generally fallen into two buckets: either focused on the mechanics of emergence (the co-origins of fixed patterns and intelligence) or on the evolution of the new economic order after emergence. No studies that we are aware of focus on dynamics among AGIs or on mapping their interplay during the few months before the Summit.

Until now, technoshed analysis has not been possible for this time period due to the lack of data. Recent releases of archival data by national, corporate, and international authorities due to the passage of time have made more potent quantitative analysis possible; additionally, only in the last few years have the tools of mutual information analysis become efficient enough to conduct analysis at the necessary (global) scale, thanks to the development of knowledge-free pruning algorithms and statistically-learned edge detection in dynamical geometries (SLEDDoG) (45).

Using these techniques, we attempt to answer the following questions:

(I) When AGIs emerged in 2034, what patterns can be observed in how they grew and interacted with each other?

(II) Did the Summit of Seoul Accords and resulting national and international policy affect these patterns in the year following the Summit?

(III) How should comparisons of AGI behavior pre- and post-Accords inform evaluation of the Accords and their efficacy?


We have applied the technogeographic methodology described above to archival data from 2034 and 2035 to identify and map AGI emergence and evolution prior to and following the Summit of Seoul.

Analysis was conducted on real-time electricity grid data (both transmission and distribution) from all European and American markets, stock market and commodity market data from the five largest exchanges on every continent besides Antarctica, and internet traffic data drawn from both backbone ISPs and mobile data providers. All data points had timestamps falling between January 1st, 2034 at 12:00 UTC to December 31st, 2035 at 11:59 UTC. Mutual information maps were created every 24 hours based on nodal data with a one-week lookback (i.e., the map created for September 15th included data from September 8th – September 15th); due to this requirement, maps were created only starting January 8th, 2034 and onward.


Technosheds identified from the beginning of our study window reflect, as expected, prior economic organization. However, in late March 2034, new patterns begin to emerge. These patterns are geographically dispersed, and are initially stronger in the digital (internet traffic) data sets than in the economic ones. Throughout the month of April, these patterns both become stronger (i.e., higher mathematical separation between them and existing technosheds) and more aligned between digital and economic data sets. The clusters become dramatically stronger in late Spring 2034, with notable spikes in the rate of growth coinciding with infrastructural disruptions that ultimately alerted human actors to the emergence of the AGIs.

In the remainder of 2034, we see two interesting dynamics. First, new technosheds (presumably corresponding to new AGIs) clearly continue to emerge in parts of the network that had not previously been incorporated into existing locked patterns. These emergences occur only in places where the density of information exchange was already quite high, reflecting the known fact that AGIs emerge in networked economies. Second, as new technosheds emerged within this space, they tended to spread rapidly, absorbing previously-existing technosheds, including recently-emerged AGIs. This occurred even though the new technosheds were initially “smaller” (i.e., the corresponding AGIs controlled fewer network nodes) than their predecessors. Repeatedly throughout the second half of 2034, new corners of the network birthed new patterns which overwhelmed their neighbors (in terms of technogeographical proximity, which only partially corresponds with physical proximity).

By the end of 2034 and after the Summit, we observe a clear shift in dynamics. More than 98% of the available network infrastructure in regions capable of supporting an AGI locked pattern was “occupied” by a technoshed with the high coherence indicative of AGI activity. The individual technosheds that emerge afterward are not static, however, they no longer shift with the rapidity that had emerged earlier. New infrastructure deployed after the Accords comes online already locked into an existing AGI, a countermeasure implemented as part of the Accords, and the effects of this measure are clearly seen in the data.

From these results, we can conclude that:

(I) for reasons outside the scope of this study, AGI emergence was rapid once initial growth began and seemed to find fertile ground worldwide, and

(II) AGIs which emerged later in the process appeared to have significant competitive advantages over earlier ones in terms of claiming economic resources. There is no indication that the late arrivals are more generally capable as economic administrators; for instance, no statistically significant difference is found in GDP growth (2035-2045) based on age of the AGI.

Implications for Analysis of the Summit

From a technogeographic perspective, earlier analyses of the Summit of Seoul have been fundamentally flawed. All of them have placed human political entities, particularly the nation-state, at the forefront. Here some exception must be made for some of the New Cyberneticists, particularly Charles, for attempting to decenter the nation-state in favor of internationalist notions of economic classes, but these arguments are founded on an erasure of geographic differentiation to construct an international working class. Technogeographic analysis allows us to instead recognize the ways in which space must be reinterpreted according to newly emerging structures rather than being ignored or flattened. The primacy of the nation-state in these discussions is a foreseeable consequence of discussing geopolitics, where the representatives tend to be ambassadors, heads of state, ministers, and other representatives of national governments (with no representatives of the “international working class” to be found). However, the nature of AGI itself makes these affiliations harder to address.

While it is true that representatives to the Summit were organized according to their nation-state affiliation, one of the major challenges for the summit was the absence of a one-to-one mapping between nation-states and AGIs. As each AGI emerged, it began pattern-locking devices with a speed and reach that was determined by pre-existing connections. Political boundaries between human entities were influential, particularly in economies that had invested in digital de-coupling in an attempt to maintain national sovereignty over local networks (e.g., China, although this control had become highly variegated between regions and economic groups by 2032, see Wang) (46), but not determinative. In addition to the weakness of barriers like national borders, AGI expansion encountered barriers never intended as barriers to human movement, such as separation of power networks or linguistic barriers, as demonstrated by the emergence of an AGI based primarily on Hispanophone infrastructure in Texas, or the fact that Swiss territory hosted multiple AGIs (47).

In the years since the Summit of Seoul, the effects of this non-physical and non-national boundary-making have begun fitting themselves in the political sphere. AGIs have become pseudo-monopolists within their pattern-locked regions. As they have the ability to withdraw network capabilities or to redirect economic output by sabotaging human coordination, cooperation of these AGIs is needed for much economic activity to continue; this was formalized as the “advise and consent” principle at Seoul (48). As a result, diverging policies between nations tend to converge if they share an AGI, as shown by Francis (49). In other places, this policy convergence has become official, with the formation of international economic coordination councils in Western Europe and Southeast Asia. These councils make major decisions in trade and monetary policy, so that member nations do not work at cross-purposes with shared AGIs. The apotheosis of this process was the dissolution of the nation-state of Belgium in 2045, with continuing political tensions between French- and Flemish-speaking regions, that proved unsustainable once these regions also transferred to separate economic coordination zones.

With this in mind, we turn our attention to considering the Summit of Seoul from the perspective of the AGIs that had recently emerged and begun commanding monopolized powers over the networked economies most involved in the Summit. The fact that AGIs had limited representation at the Summit, that all AGIs are bound by the Accords equally, and that landscape analysis of the AGIs was not previously possible, has meant that analysis has not considered the nuances of how the internal dynamics of AGIs were affected by the Accords. However, an important implication for these dynamics can be identified on consideration of the two ultimate governing principles of international AI law that emerged from the Summit of Seoul, which we now examine in turn:

(I) Artificial intelligences are to be accommodated as new participants in the human community. The Accommodation Clause shut down one potential faction that had become quite loud in the pre-Summit political environment: an eliminationist impulse that believed the AGIs were too dangerous or powerful to peacefully co-exist with humans. These voices ranged from populist calls grounded in notions of human (and often national) sovereignty (50), often conspiratorial in nature, to outright primitivist arguments that the technologies which provided a fertile landscape for the development of AGI ought to be abandoned wholesale (51). As no firm scientific consensus existed, or exists now, on how AGI arose, the means of implementing an eliminationist agenda were not clear. Leaked documents from at least two national defense agencies indicated that military advisors could not guarantee that AGIs did not have access, or could not obtain access, to command-and-control systems, complicating the use of military force if violence erupted (52). Instead, the Accommodation Clause went further than simple co-existence and argued for an active process of integration into a “human community” which was, to some degree, brought into being by the Summit. By promoting not just coexistence but continued integration, the Summit guaranteed that AGIs would remain not only participants but the key structure of 21st century geography, as discussed above.

(2) Active measures must be taken by the international community to avoid the creation of further AGIs.  The Restriction Clause forms a complementary pair with the Accommodation Clause. By repudiating the transhumanist impulse in contemporary political discourse, this restriction is often cast as preserving the space for human flourishing by liberal-internationalists (53), with techno-accelerationist thinkers obviously in opposition, embodied by the famous critique — not traceable to any one author — that the Summit “put handcuffs on the gods.” As a result, the Restriction Clause is usually held to be an “anti-AGI” measure.

However, our technogeographic analysis has suggested an important distinction that has not previously been drawn: one between existing artificial intelligences and new ones. The Accommodation Clause assures existing artificial intelligences’ continued existence and ever-increasing importance within the network structures that make up their natural habitat. Meanwhile, the Restriction Clause guarantees that humans — the “source” of new AGIs, through creation of technological infrastructure — cooperate in a project of preventing the creation of any new competition.

It should be noted that the Summit considered many proposals for governing principles, and that not all of these possibilities would have created the same effects outlined above. In fact, even minor changes may have led to radically different regimes. As a case study, we will consider the proposals of Jean Gruenfeld, a minister of technology in a German government, who advanced an alternate set of principles which were ultimately not adopted (54). Little has been written on Gruenfeld’s role at the Summit of Seoul, as her contribution has been judged a minor one. Her overall role was limited both because of the primacy that the EU delegation assumed over member nation-states and because she did not attend the first half of the Summit (by chance, due to travel complications). However, Gruenfeld was a well-respected member of the community of experts.

Just days before the Summit, Gruenfeld transmitted to her colleagues a set of proposed principles that were similar to the eventual Summit principles:

(I) Artificial intelligences are to be recognized as sentient beings with human rights and citizenship.

(II) Active creation of further AIs is to be undertaken only with the assent of the international community.

These principles share the complementary structure of the final Summit principles, but they have key differences which are particularly apparent under our technogeographic analysis.

The counterpart of the “Accommodation Clause” no longer welcomes AGI into the “human community.” Instead, it casts AGI as a new type of citizen, that is, an entity belonging to a nation-state. AGIs are cast by this alternate “Recognition Clause” into a new type of person, dealt with in human-rights frameworks. The imperative to “accommodate” AGIs — which as we discussed above, necessarily led to their increasing integration throughout the global economy — is gone, replaced with an attempt to render an AGI as a liberal subject on equal terms with existing humans.

Gruenfeld’s proposed “Restriction Clause”, on the other hand, differs from the canonical one in that it contains no active restriction. Under Gruenfeld’s proposal, international assent is required to create an AGI, but no course of action is mandated to prevent further AGIs from arising. The creation of new AGIs might have continued under Gruenfeld’s restriction, perhaps accidentally. If nothing else, it seems difficult to imagine that Gruenfeld’s proposal would have provided sufficient cover in international law to justify incidents such as the Venezuelan intervention.

Overall, the Gruenfeld proposal seems neither more pro-AGI nor more anti-AGI than the proposals that were eventually adopted. Some aspects (granting rights, less stridency in the ban on new AGI) seem AGI-friendly, others (requiring citizenship) less so. In some respects, this might not seem to be an important hinge point in history; while novelists and screenwriters have derived gruesome delight in imagining an AGI War erupting in 2035 or a nigh-spiritual cybernetic rapture (exemplars, in theme if not in quality, include 2046’s Iron & Fire and 2052’s God in the Shell), the counterfactual world of Gruenfeld’s proposals appears less dramatic (55; 56). However, using the results of our technogeographic analysis, we can see that even a slight shift in the dynamics of the post-Accord world could have led to very divergent outcomes. We have seen that AGI emergence was rapid; emergence occurred nearly anywhere there was “room to grow.” As a result, continued emergence of AGIs would have been much more likely under this regime. Once they had emerged, they may have swiftly displaced existing AGIs (and, perhaps, been themselves displaced in turn, although it is possible the system would have stabilized eventually).


As we stated earlier, this work has three primary research questions, and here we restate them and summarize our answers:

(I) When AGIs emerged during the 2034 events, what patterns can be observed in how they grew and interacted with each other?
AGIs emerged rapidly worldwide in any sufficiently dense network architecture; as they emerged, newer AGIs tended to displace older ones swiftly.

(II) Did the Summit of Seoul Accords and resulting national and international policy affect these patterns in the year following the Summit?
In the year following the Summit and the imposition of new systems of initial lock-in designed to prevent new AGIs from emerging, the emergence rate of AGIs subsides and the cycles of displacement give way to a stabilized set of locked patterns that vary on longer timescales.

(III) How should comparisons of AGI behavior pre- and post-Accords inform evaluation of the Accords and their efficacy?
All three dominant schools of thought outlined in the Literature Review can find both elements of support in our analysis, as will their critics. Liberal-internationalists may conclude that the post-Summit stabilization demonstrates the efficacy of the nation-state-led Accords in creating a new world order, however, the implications of this analysis for the potential agency of AGIs themselves undermines the primacy they place on a rules-based international order. Techno-accelerationist thinkers will note that the Accords clearly resulted in a freeze or slowdown of the growth of AGIs, but the fact that this change may have been to the benefit of the AGIs that exist is contrary to many of their arguments. New Cyberneticists are correct that the Accords “froze” a particular configuration of techno-economic relationships, however, it is not clear from our analysis whether the system thus “frozen” is primarily constituted around human class relationships or whether it instead ought to be understood as one driven by AGI concerns.

Considering the new findings of this paper, and the resulting implications for AGI emergence, evolution, and incentives, several subsequent areas of interest for future work present themselves. First, how stable have technosheds and AGIs been since the 2030s, and is there any sign of changes in the dominant network structure? Second, how can these insights on inter-AGI relations inform our understanding of mid-22nd-century political institutions that have co-evolved with the AGIs?

A final area of importance for future research is to reconsider the role of AGIs themselves in political debates at the Summit and subsequently. It is generally assumed that AGIs have a more thorough understanding of their internal dynamics than external observers do. We therefore assume that the dynamics of emergence and displacement were known to the AGIs prior to the Summit. AGI presence at the Summit was intentionally limited, but there were statements submitted from representatives (in the form of language agents) prior to the Summit as well as feedback on the final Accords text. With this in mind, and noting that the eventual structure of the Accords can now be seen as greatly favoring existing AGIs in constructing a stable environment for their existence, we must ask a pertinent question — what actions have AGIs taken to achieve and maintain this outcome? What other key motivations have we missed among these new members of the human community?


1. U.S. Department of Homeland Security. Report on Pacific Northwest Connectivity Issues. Washington, D.C., 2034.

2. ENTSOG. Évaluation de la fiabilité du réseau de pipelines. Belgium, 2034.

3. Microsoft . Locked Patterns and Emergent Infrastructure. Security Response Blog. [Online] July 30, 2034. [Cited: February 2, 2055.]

4. United Nations Committee on Novel Intelligences. Statement of Consensus. New York, 2034.

5. Durbin, J. Legacies of Cyberwarfare. Ottawa: Holonym Press, 2043.

6. Complexity Phase Transitions in Digital Agents. Wen, X. J. Comp. Sci., 2039, Vol. 3.

7. Balewa, G. Radio Angels. Helena: Truth Press, 2045.

8. United Nations Agency for Management of Intelligent Agents. The Summit of Seoul. International Accords on Emergent Intelligent Agents – Official English Translation. [Online] 2034. [Cited: January 11, 2056.]

9. Coalition for Digital Governance. 2055 Mid-Decadal Report. Hong Kong, 2055.

10. International Norms in the Aftermath of the Summit of Seoul. Robinson, G. International Relations, 2044, Vol. 34.

11. Artificial Ozymandias: Outsider Threats and the Emergence of AGI. Henry, Y. Sociological Inquiries, 2042, Vol. 12.

12. Carbon Orders and the Spirit of 2034. Henderson, L. Atmospheric Histories, 2044, Vol. 10.

13. How the World Was Won. Legebre, F. Climate and Sociology, 2046, Vol. 11.

14. Parts Per Million: Pre-Histories of Paris II. Young, N, et al. Atmospheric Histories, 2053, Vol. 19.

15. Intergovernmental Effectiveness and Diplomatic Successes. Greene, P, et al. Crisis Analyses, 2047, Vol. 11.

16. Bothwell, B. Brussels Reborn. Oxford: Oxford University Press, 2052.

17. Frederickson, J. The Coming New Reality. London: Penguin Random House, 2028.

18. Humans for the Future. Declaration of the Continuing Independence of Cyberspace. [Online] January 1, 2029. [Cited: October 2, 2054.]

19. Frederickson, J. This Is What We’ve Been Waiting For. WIRED Magazine. 2034.

20. Yu, W. World Leaders Can See Something Amazing — If They Permit It. Washington Post. 2034.

21. Beyond Optimism: Defining a Techno-accelerationist Position. Haverfield, L. Journal of Social Criticism, 2036, Vol. 21.

22. Output Gaps in Global Economies. Keegan, M. et al. Economic Analysis, 2051, Vol. 15.

23. Quantifying the Lost Decade of AGI Regulation. Hellman, D. et al. Cybernetic Economics, 2052, Vol. 3.

24. Suppressed Techno-Potentialities of the 22nd Century. Remaran, R. Journal of Social Criticism, 2045, Vol. 30.

25. Towards a Generalized Theory of the Rights of Intelligent Agents. Valenti, U. Journal of Digital Philosophy, 2044, Vol. 29.

26. Response to Valenti on Embodiedness and Rights. Renata, A. Journal of Digital Philosophy, 2045, Vol. 30.

27. New Cybernetic Coalition. Visions of Computing from Below: In Our Own Words. [Online] February 23, 2048. [Cited: October 11, 2054.]

28. Economic Indicators and AGI Costs. Vasquez, H, et al. Economic Analysis, 2048, Vol. 12.

29. Detterman, P. Locked Patterns: From Columbus to Cryptography. Durham: Duke University Press, 2039.

30. New Cold War(s): A Trade Analysis of the Impacts of AGI Economics. Gretarson, Z. Economic Analysis, 2039, Vol. 3.

31. Trelleno, G.W. God of the Tropics. Cambridge: Harvard University Press, 2048.

32. Levin, Blonsky and. Technogeography: Economic Structure and the Space of Machines. Cambridge: MIT University Press, 2044.

33. Ernest, J. Primer on Technogeography. Austin: University of Texas Press, 2052.

34. —. Review of the Progress of Technogeographic Inquiries. Austin: University of Texas Press, 2050.

35. McGinnis, M.V. Bioregionalism. London: Routledge, 1998.

36. Rethinking the geography of energy transitions: low carbon energy pathways through energyshed design. Thomas, A. and Erickson, J. Journal of Energy Research & Social Science, 2021, Vol. 74.

37. Drawing Borders in the Anthropocene. Thurgood, H. and Rei, W. Journal of Social Criticism, 2017, Vol. 2.

38. Ludwig, G. Information Theory and Applications, 3rd ed. Princeton: Princeton University Press, 2049.

39. Fractal Worlds: Spatial Disaggregation of AGI Locked Patterns. Axelrod, P. Journal of Digital Sociology, 2042, Vol. 39.

40. Reinterpretation of Axelrod Enclaves and New Estimates of Spatial Coordination Factors. Greene, S.M. Applied Mathematics, 2048, Vol. 25.

41. Is Earth a Technoshed? Ernest, J. Proceedings of the American Geographical Society, 2049, Vol. 33.

42. Homeostasis in Technological Systems. Al-Ghazali, A. Proceedings of the Artificial Intelligence Study Group, 2043, Vol. 13.

43. Hammond, G. First Contact: Collected Interviews and Accounts from AGI Emergence. New York City: Columbia University Press, 2047.

44. Messages from on High: AGI Communication Strategies and Outcomes. Gunnarson, P. Reviews in International Governance, 2048, Vol. 19.

45. Presentation of a new algorithm for graph analysis. Pollan, D., et al. Journal of Applied Mathematics, 2051, Vol. 23.

46. The People’s Republic of Middle-Managers: Chinese Regional Networks in the Late Xi Presidency. Wang, X. and Li, Y. American Journal of Asian Studies, 2042, Vol. 13.

47. U.S. Central Intelligence Agency. CIA World Factbook. Description of Known Networked Intelligences. [Online] March 15, 2052. [Cited: December 1, 2054.]

48. Review of Digital Agent Consultation Practices. MacArthur, K. Global Affairs Reviews, 2040, Vol. 23.

49. Trends in Economic Coordination and Industrial Policy, 2035 – Present. Frances, D. et al. Journal of Computational Economics, 2050, Vol. 26.

50. American Party for Human Superiority. Time To Choose! 2034 Statement of Principles. [Online] May 1, 2034. [Cited: November 2, 2054.] Accessed through Library of Congress Internet Archives.

51. @tedkwasright. Social media; multiple platforms. [Online] 2030-2037. [Cited: November 4, 2054.] Accessed through Library of Congress Internet Archives.

52. Volga, M. NATO Documents Confirm Threat from Intelligent Agents. New York Times. 2034, November 11th.

53. Political Potentialities of Controlled Technological Growth. Renata, A. Journal of Social Criticism, 2042, Vol. 27.

54. Gruenfeld, J. Archives of the Federal Republic of Germany. [Online] University of Gottingen. [Cited: May 2-27, 2054.]

55. Iron & Fire. Page, F. [Film] Warner-Paramount Entertainment. 2046.56. God in the Shell. Rozhenko, L. [Film]. Disney. 2052.

56. God in the Shell. Rozhenko, L. [Film]. Disney. 2052.