Anarchic AI Governance
We are currently living in a state of Anarchy when it comes to AI Governance - there are no established institutions and structures to guide or dictate our road towards building the most powerful technologies Sapiens is known to have ever built. How do we, as a civilization, deal with this decisive moment in history?
Understanding Anarchy
In the context of International Relations theories, anarchy is a state of disorder due to the absence of a central authority or governing body that can regulate or dictate the behavior of states and other actors in the international system. This concept is fundamental in the study of International Relations, as it highlights the challenges and opportunities that arise when there is no global government or authority to enforce rules and norms.
Realist theories of International Relations emphasize the role of power and self-help in international politics and argue that in the absence of a supra-state power or arbiter, there are no enforceable rules of conduct, especially for strong states. As a result, the international environment is anarchic both in the strict sense of lacking enforceable international law and in the broader sense of being violently chaotic. The prevalence of this environment in turn requires that the primary goals of individual states be survival and security.
However, the constructivist interpretation of anarchy takes a different approach. According to Alexander Wendt, a prominent constructivist scholar, “Anarchy as such is an empty vessel and has no intrinsic logic. Instead, anarchies only acquire logic as a function of the structures that we put inside them.” In other words, the behavior of states and other actors in the international system is not determined solely by the absence of a global government or authority, but also by the norms, identities, and interests that shape their interactions.
According to liberal theory, anarchy in international relations can be counteracted through the establishment of liberal democratic institutions and norms. By promoting free trade, individual rights, and multilateral diplomacy, states can work together to create a stable and peaceful international order. The emphasis on individual rights and freedoms, rather than state power, allows for the possibility of cooperation and collaboration among states. However, the effectiveness of liberal theory in addressing anarchy in international relations is still a subject of debate among scholars.
The implications of anarchy in International Relations theories are significant. It highlights the need for global governance structures and institutions to regulate state behavior and interactions to prevent violent conflict and promote stability and cooperation. At the same time, it also underscores the challenges of establishing such structures and institutions in a world where states and other actors have different interests and goals.
However, in modern discussion, things become a little more complicated:
First, in addition to states, we now have other actors who have gained immense power in the last few decades and are trying to claim their space in this anarchic reality. These are big multinational companies, a new type of empire without borders and violence, instead characterized by borderless soft power.
Second, society already lives in two different realities - virtual and physical - the consequences of which will become increasingly noticeable on a larger scale over time. Virtual realities are already more difficult to manage and govern, and over time, we will develop different values in virtual worlds characterized by an even more anarchic nature.
Third, we are witnessing social complexities increasing at a faster pace, from historically highly centralized powers to less centralized governance towards democracies, which may be transitioning into more complex and decentralized systems in our times. We are witnessing rising regionalism and various (virtual) self-governing communities. Decentralized Autonomous Organizations (DAOs) also emerge as an alternative structure to conservative organizations with centralized powers. Some social scientists explain this tendency through the lens of the second law of thermodynamics, where entropy is ever-increasing, and in turn, local order arises. However, we do not yet have evidence that social structures obey fundamental laws of physics; so far, it's just an entertaining of thought (you may see a separate blog post from me on this topic!).
Now that we found out that we have additional actors (corporates) in our current state of anarchy, as well as additional dimension (virtual) and additional speed, let’s move forward with understanding how this can apply to AI.
Putting AI into the equation
Putting all these variables together, our AI anarchy is the most anarchic as it can get but this does not mean we should not aim to understand it. Moreover, we are obliged to construct new realities in it. Moreover, AI itself adds some more complexity to the question we need to solve.
In its 315,000 years of existence, Homo Sapiens is not known to have ever created anything comparable to AI. Can you imagine that holding a (smartphone) camera in your pocket documenting daily life is just a decade-old concept/reality? And that all technological achievements, including AI, happened in the last 200 years of Sapiens living on earth. AI can already somewhat accurately mimic at least 2 human senses: vision and hearing, they are the best chess and Go players and they are capable of processing huge datasets. This growth is exponential and it's hard to imagine what it can achieve in the next 10, 100, or 200 years Would you exclude the probability that they’d be able to not only mimic human senses and capabilities but surpass them? You get my point, we are living in special times holding special tools that we will need to very carefully tame. They’ll be self-learning, adaptive, and autonomous requiring us to change how we’ve been approaching governance, management, law-making, or anything else before.
People who are affected most by these technologies do not have the power to participate in their making. Corporate companies are the ones setting the rules whether formally or informally and corporates do not represent people’s voices unless we reiterate that customers can exercise their ‘voice’ by unsubscribing from undesirable services (but guess what, I’ll have to ask you how possible this is in the context of monopolies which all big tech firms are).
Things get even more challenging - not only the companies but also common people, 15-year-old sitting in his/her room may build a powerful algorithm or product that can affect thousands of lives. It is amazing how AI empowers everyone unless it empowers wrong purposes and power gets disconnected from responsibility.
We are witnessing an AI race happening that is comparable to an arms race in certain ways but not completely - the AI race will be more fierce, more subtle, and more dangerous just because it’s subtle and also more widespread, more accessible. And because we have a virtual dimension which is harder to govern. Gini is out of the bottle and we are at a point of no return.
If AIs become autonomous agents, will we need to equip them with fundamental AI rights? This means we may have one more additional actor in our anarchic system who would want to exercise their own interests and seek power.
Now that we’ve seen that AI may be contributing additional variables in our anarchy equation - additional actor (AI itself), additional speed (exponential), and additional complexity to the two dimensions (physical and virtual) - we are left wondering how to solve this equation.
Take a deep breath but don’t panic! We should be able to find a solution. We are humans, capable of amazing things.
So, what’s next?
If we are heading to a rather exponentially chaotic world, is it possible to control (manage, govern) such chaos?
How can we use this challenge or disadvantage to our advantage?
Let’s start from the top down: some scientists have been suggesting, that we need to set up a global/international AI organization that would work out standards and regulations for AI. Is this a good idea? Somewhat yes, unless we deceive ourselves that this is all we need and stop there. For example, in order to address the anarchic international system and avoid future world wars, the United Nations has been formed and while solving many global problems, it has never been the ultimate answer to every global challenge.
Wendt stated that three different cultures of anarchy dominate international systems - Hobbesian, Lockean, and Kantian - that I also like to look at as stages of anarchy respectively. In Hobbesian culture states perceive each other as enemies and deny each other’s existence, they have no self-control or self-restraint. As a result civilizing process is almost non-existent and violence is omnipresent. We seem to have passed this stage of anarchy. In Lockean culture, states perceive each other as rivals and they recognize each others’ existence because they recognize each others’ sovereignty. As a result, the states seek to find a balance between self-control and external societal restraints. Civilizing process is developing and wars also may break out periodically. We seem to be in this stage of anarchy heading to the next stage, Kantian. In Kantian culture states perceive each other as friends, they recognize each others’ existence and respect the rules of non-violence. The civilizing process leads to empathy, mutual aid, and a global cosmopolitan society.
Now let’s try to apply this to our AI anarchy - in the last couple of years, we’ve been in Hobbesian anarchy where there is enmity, a race to the bottom amongst companies to develop the most powerful AI systems. No self-control or self-restraint. As a result, we have not achieved common governance methods. Next, in Lockean anarchic culture, companies start to see each other as competitors but recognize others’ achievements and common threats, as a result trying to balance between self-restraint and societal pressure, working towards common solutions for AI governance and risk mitigation - we seem to be heading there. Finally, in Kantian anarchic culture, companies will choose collaboration/friendship over competition and exercise more self-control. Empathy, responsibility, and respect for common values will emerge as a major underlying governing system. Now, the question is, how long will it take us to get to the Kantian stage of AI governance?
Governments still have their role. Do regulations always stifle innovation? If found the right balance we could even argue the opposite, regulation can serve as a coordinated effort to set the highest standards and focus on quality - and we have all the reasons to set the highest standards/requirements for AI (listed above). At the same time, it will be very difficult to benchmark quality, and especially to make them relevant in the long term. We’ll make many mistakes but we need to learn from them extremely fast. However, existing policy-making processes do not allow this iterative approach. Hence we also need to innovate policy-making processes.
This leads us to the question of whether we need to build completely new types of institutions, and whether AI will prompt us to live life like never before. My answer is yes, AI will fundamentally change not only how we use technologies but also how we function as a society and as individuals. We will need to answer to this entirely new type of challenge in entirely new ways. There are already a couple of interesting projects emerging such as MetaGov, Collective Intelligence Project, pol.is. But we are yet far from achieving common grounds on how we make policy or accurately predicting its implications. In any case, these new types of organizations will prioritize the direct participation of people/citizens/users in AI policymaking as it will affect almost every aspect of everyone’s life - such decisions cannot be made by unelected individuals at tech headquarters.
Each of these points deserves a separate article to describe in more detail how we should achieve the best results or what are even the best results? I’ll continue trying to provide answers but we need collective global intelligence to solve humanity’s greatest challenge - how to treat a competing intelligence.