While military armament is often considered the most pressing concern, recent developments over the past few weeks have brought attention to the shifting landscape of political alliances. The conspicuous gestures made by various nations during this period have revealed a notable division of global alliances. On one side, we have North Korea and Russia, along with a few Baltic nations. On the other side, we find a coalition consisting of most European countries, India, the United States, and Pacific Asiatic nations. It's important to note that these are the alliances that have been prominently featured in the news, and this assessment is not meant to exclude any nation.
These events may appear unusual at first glance, but they could be indicative of preparations for a global conflict with distinct sides emerging worldwide
Perhaps the most significant threat resides much closer to the realm of human thought and society. Battle lines are becoming increasingly distinct among nation-states. Here, I attempt to articulate why the widespread distribution of AI technology is urgently needed, especially in light of these ongoing socio-political events.
_________Why We Need Distributed AI Urgently _________________________
To provide a foundation for discussion, I'd like to outline some key concepts that underline my reasoning. In today's digital age, social data is continually collected and linked to individuals rather than machines. This vast pool of data serves as the foundation for creating algorithms capable of analyzing and understanding both individual and collective behaviors. These algorithms can then power AI predictive models, which, in turn, can offer insights into societal structures and potential influences.
The control of this technology and data by nation-states holds the power to shape the information upon which individuals base their present and future decisions. Presently, only a limited number of nation states possess the computational resources required to develop truly disruptive AI models. Therefore, the most efficient path to dominance lies in the hands of those who can establish a more centralized and controlled AI system within a less complex social framework, often characterized by command and control structures. Such a system could be utilized as a potent force of social organization and influence.
Once such a system is established, particularly when consolidated and controlled by a nation-state, it becomes capable of fabricating a version of reality for a society. In this context, humans may find it increasingly difficult to distinguish AI-generated content from human-generated content, given AI's ability to mimic human behavior. Moreover, as our society becomes more intertwined with technology, the impact of AI on social dynamics is amplified.
The current AI arms race is intense, with various nation-states vying for dominance. Within this competition of different social models, nation-states possess significantly greater computational resources than individuals. Consequently, it is conceivable that a less complex social system, empowered by a consolidated and controlled AI infrastructure, could swiftly surpass more complex social systems in the race for influence and dominance.
Next, a few ideologically social principles are associated with nation states, and even at a high-level some social organizational principles are in conflict due to power distribution and decision-making. For the sake of argument, let’s say dictatorship (top-down consolidated power to control masses by a few) and democracy (a system of power distribution to the masses garnered by rules of conduct). The point is, they are human social structures and knowing the data associated with the population is important to run any AI that could distort or create a type of reality for different purposes.
Furthermore, we must consider the ideological and social principles associated with nation-states, as these principles often come into conflict due to disparities in power distribution and decision-making. For the sake of discussion, let's juxtapose dictatorship (a top-down consolidation of power wielded by a select few to control the masses) with democracy (a system of power distribution to the masses governed by established rules of conduct). These are fundamentally human social structures, and understanding the data associated with a population becomes paramount when running AI systems capable of distorting or shaping different versions of reality for various purposes.
The central thesis here is that there exists an urgent need to empower a broad spectrum of individuals with the computational power and access to data required to create AI models because complex social systems are harder to foster well and any AI to foster that complexity much harder to build. This urgency arises from the inherent complexity of modeling human societies, which must compete swiftly with less intricate AI models of social structures before the latter become the predominant reality.
The social organizational systems mentioned earlier stand in opposition primarily due to the locus of power they represent. It's conceivable that data on individuals can be harnessed to influence the distribution of power within these systems. Both dictatorship and democracy rely on social data to exert influence or sway over a population, but the advent of AI technology has magnified their capabilities to construct social structures through influence that mimic human behavior. However, the technology required to challenge these structures is currently beyond the reach of the majority of the population.
This is where the danger of AI emerges, and it becomes a cause for concern if we fail to distribute the technological capabilities and datasets more broadly. We must also take into account the concept of human free will and the potential risks associated with AI manipulation. Distributing AI models, as opposed to permitting the consolidation of a monolithic system along with its associated data, may be a safeguard against undue and false influences that run counter to our best interests as human beings. It also helps mitigate the risk of inducing behaviors that contradict our core beliefs.
When AI is employed to assert control over a population in a highly centralized manner, its potential for influence becomes immense, as it can manipulate the population's perception of reality. A consolidated technology system, armed with AI, stands a greater chance of success in this endeavor than a distributed system comprising many models. This rapid establishment of societal norms underscores the sense of urgency surrounding this issue.
If a multitude of individuals gain the ability to create algorithms and integrate diverse datasets (such as different languages in the case of Large Language Models), even without access to certain proprietary data or not, they may start to wield influence, either positively or negatively but certainly adding complexity and decentralization. Currently, the computational power required for these tasks, including GPUs and access to data, may only be within the reach of nation-states and large organizations. Therefore, it becomes essential to consider either sharing this capability or establishing regulations for its responsible use (though how that could be influential world-wide is very difficult to know and therefore I think we still urgently need the deployment of AI technology fast to a large population). Additionally, the development of accessible tools and platforms that enable individuals to create AI models is crucial.
In embracing this approach, we can cultivate an ecosystem that promotes competition among complex AI models, steering away from a scenario where power consolidation remains the prerogative of large social entities, often at the nation-state level. This transition encourages a more diverse and democratic landscape in AI development and utilization.
The precise manner in which specific models emerge victorious or falter remains to be determined. Nevertheless, it's possible that humans may not yet reach a consensus on unity, but we can draw inspiration from nature. In nature, we often observe that erring on the side of complexity tends to favor continued existence, health and longevity. By fostering a dynamic interplay of complex AI models, we may similarly enhance our chances of adapting and thriving in an ever-evolving technological landscape. Complexity is not easy yet the opposite is rather stultifying to imagine as any slight error brings monolithic structures to a halt while human beings do not fare well marching towards that moment. Maybe we can choose differently.
In light of the global context where lines seem to have been drawn, and considering the very real likelihood that centralized nation-states may not see it in their best interest to collaborate on a worldwide set of rules for AI, it becomes imperative to contemplate what lies ahead. It's conceivable that these nation-states have already run simulations indicating that a less complex system could swiftly gain dominance and control over society. Relinquishing this control to a more intricate power structure might not align with their interests. But what transpires if something disrupts this perceived or actual dominance in the years to come? Does it lead to an escalation, potentially resulting in a hot conflict as an attempt to reassert control? What measures will be in place to discern and manage such a situation?
What I do know is that within a highly rigid command and control system, even though it wields significant centralized control, the weakest fracture can topple the entire structure. Could this fracture represent a sense of human free will and diversity from a complex system? I don't have a definitive answer at this moment, but it is prudent to err on the side of caution, seeking to prevent any such collapse and striving to cultivate a high degree of complexity now. This complexity is essential to ensure the possibility of a future marked by collaboration, one that encompasses various expressions of human and natural diversity, and abundance of differences with a few guiding principles only. However, achieving this complexity takes time, and that is why I find the current situation deeply concerning.
What will happen next?