In the last two years, machine learning systems, more broadly known as artificial intelligence, have entered the realm of wider economic conduct and have partially been made available to the general consumer base. The main difference between artificial intelligence systems and classic computational processing is that those systems can perceive and interpret audio-visual content and recreate content to achieve certain goals. For example, generative artificial intelligence systems can create new visual content based on databases that have been interpreted. From the extracted information in the databases, programmes can then combine common elements and defining properties to create new content that fulfills the criteria of the concept that is to be created while actively avoiding including criteria that would place the created content outside of the definition. If the entered command is to create an image of a dolphin, the artificial intelligence system will most likely create a picture of a dolphin in water and not in the jungle. The same holds true for the processing of texts. On a more complex level, artificial intelligence systems can also solve equations and process and interpret comprehensive mathematical data that informs scientific progress in the natural sciences. The most important aspect of artificial intelligence, and the reason why they are also called machine learning systems, is that the processing and recreating of all sorts of content informs future interpretation and development of knowledge. In contrast to conventional computing systems, machine learning can hypothesise based on all available structures, rules, dynamics and data. With those properties, machine learning systems will become an integral part of our daily lives, as the impactful entry into the general economy has already shown. Many processes and structures have immediately been affected by artificial intelligence due to its rapid and continuous functioning. This means that it will also affect politics once it is a more consolidated part of our globalised and digitalised conduct. This essay discusses how artificial intelligence can inform politics and what potential problems machine learning will cause and encounter in the context of politics.
A Devletist Tool
One of the major strengths of machine learning is that subjective data, such as images, opinions and behavioural patterns, is quantified by translating it into quantitative data points. Those data points serve as the fundament for the system to define rules and regularities about humans and our societies. Based on the massive amount of data that can be processed simultaneously, the system can also process irregularities within those defined rules and explain how they come into existence. The existence of those irregularities either falls outside of the scope of those rules and regularities, usually because they are caused by other complexly covered factors, or the rules and regularities are not sufficiently examined or accurately defined. In the political context, the major advantage that artificial intelligence systems introduce is that social, economic and societal structures and their rules can be processed simultaneously. Processing then includes examining whether dependencies and causalities exist and whether there are conditionalities between various phenomena in the realms of society and economy that determine outcomes at the political level. The reason why it is very important to process comprehensive datasets with complex information simultaneously is that it helps to prevent universalism. One of the main reasons why political parties come into existence is because they are often based on a limited set of assumptions, observations and theories that form the concept of an ideology. That is not to say that those theories are in themselves faulty, but they do not depict the truth fully, and often not even partially. Simplistically, ideologies emerge when someone stumbles over a pragmatic rule of thumb that seems to address a variety of problems of a certain group. Naturally, politics is being viewed through the lens of this ideology and reality is skewed to fit the frame of the ideology where possible; where impossible, reality is ignored. In political systems with different ideological groups, political power (in the sense of capacity) is dispersed into different camps that approach political questions from their ideological perspective. Artificial intelligence systems, however, have the potential to discover dynamics in society, economy and politics much more accurately due to access to a larger database. Further, it can directly identify which dynamic, rule or regularity is affected by others. Moreover, this also helps us understand the hierarchy of dynamics in our world better. Some dynamics are more influential on other factors of life than the other way around. Additionally, the importance of factors varies over time and is tied to other varying conditionalities. Accordingly, we can also utilise machine learning to understand better which factor is influencing which factor in what way.
To illustrate this thought with an example, we can think about consumer preferences. Among economic scholars, there has been a long debate between proponents who argue that consumption is driven by demand, while others defend that supply drives demand. On other levels, sociologists identify other factors that influence consumption preferences, such as social status, education, emotional development, and demographics. Yet other approaches emphasise that tax regimes and labour conditions affect consumption preferences to a much greater extent than other factors. All arguments are embedded into different, sometimes even shared, ideological frameworks. Those ideological frameworks inform the formation and operation of political parties. Their understanding and reactive policymaking, in turn, affect our reality in ineffective and inefficient ways as their organisational interests will make them tend towards reproducing desired outcomes within their ideological frameworks to validate their assessment of a fraction of our reality from which they derived a rule of thumb that they try to apply universally.
Artificial intelligence is, so far, emotionally independent. While humans tend to seek superiority over their peers caused by different factors, such as bad education, lack of confidence or knowledge, trauma responses, revenge and many others, machine learning systems, currently, do not include such dimensions in their assessment of things. They can identify why human behaviour is informed by those motives, but, and this is significantly more important, how such behaviour can be prevented through structures and preventive counteraction. Through thoroughly examining many complex dynamics at once and within their interactive context, the policy options such an artificial intelligence machine would suggest will inevitably exceed any human output in effectiveness, efficiency and feasibility. It has the potential to be used as a policymaking tool, especially for devletist political systems. The reason for that is that devletism is a technocratic form of statecraft that redirects state conduct to advance society through the production of genuine knowledge primarily. This means that policies are crafted through analysis processes that try to account for as many influencing factors as possible always to implement the best possible policy tailored to the respective situation. Other than contemporary democratic systems with political parties, where the goals of parties, and accordingly the government, change instead of the approaches, the sole and ultimate goal of genuine knowledge production is a constant in devletist systems, and the ways to achieve this goal naturally vary with the changes of our circumstances. Artificial intelligence will be an important key to further support the logical underpinnings of devletism and then also produce a potential course of policy action to advance our societies within the realm of this naturally given goal that devletism places at the centre of statecraft.
Problems in Contemporary Political Systems
Unfortunately, societies have not yet entered the developmental stage of devletist policymaking. Currently, they are structured around the goals of power and comfort. As those goals are relational and hence require points of reference, actions taken to further one’s own interest naturally mean the disadvantage of others. Then, the further one achieves relative success within the defined goals, the more acknowledgment this person receives within society. Therefore, human behaviour is strongly incentivised not only to work for own success within this framework of the two goals of power and wealth but also to prevent the success of others to maintain the relative distance of success to them, and with it, favourable social status. Within a fully free market, there are opportunities to work both ways. Certainly, there is the classical way of working for the realisation of the own power and wealth. Contemporary economic behaviour, however, displays increased leniency towards working against the power and wealth accumulation efforts of others, as well as the exploitation of consumers. Mainly, the contemporary approach to do that is to alter consumer preferences through crafting a social environment of artificial depth. Primarily through media, economic actors can shape the world view of consumers, who then shape their consumption preferences in accordance with the altered view of norms and values. We can look at music, art, fashion or any other industries, but also at more general content formats, such as cinema and television. The content composition determines how we perceive the world and what is desirable within. In an attempt to keep consumption rates high, the common consumer is driven towards more superficial lifestyles and less valuable content, activities and products. Within such a normatively impaired economic environment, artificial intelligence will be very dangerous.
Currently, artificial intelligence is mostly used within the common realm as a tool to mass-produce superficial content. For once, this is done to increase profits through higher exposure as algorithmic programmes of other profit-maximising platform providers value high-output, low-value content to be more attractive for other companies to sell products to an undemanding consumer base. Artificial intelligence is also very effective and efficient in identifying approaches to the goals of its users within the current system. Within the context of what is being valued in our contemporary systems, namely power and wealth, artificial intelligence systems are inevitably used by profit-driven economic actors. Being directed towards those goals, the artificial intelligence system will identify the best way to increase profits with the given opportunities and the environmental circumstances of our time. Just as it is a yet unemotional system in the context of useful policymaking, it is an unemotional system within other contexts, too. Given the rules and dynamics of our time, it can and will identify the most effective and efficient ways to generate profits for the user of the programme, regardless of its normative implications. Because of the rules and normative goals within devletist political systems, the results of machine learning systems will inevitably be different and, as a result of my analyses, significantly more beneficial to societal progress. However, since current political systems do not account for more advanced normativity and proper goal definition, neither the user of an artificial intelligence system nor the programme itself will produce outcomes favourable to society because those objectively desirable results are subjectively worthless within the current societal and economic setup.
This is true for the mass consumption of media, where artificial intelligence systems are expected to further develop manipulative measures to alter our social normativity in ways that highly favour consumption. However, this also holds true for the use of artificial intelligence systems within the power political realm. Here, the main field of application of those systems is that of defence and security. Information processing, forecast modeling, cyber offensives, digital infrastructure attacks and data theft will be among the most utilised features of machine learning systems. Again, as a result of policy goals structured around the idea of power, the application of this system will inevitably result in the prevention of societal progress in other nations. Other than defensive military strategy, which is passive and reactive upon breach from other nations, machine learning systems are inherently offensive in the military realm. As their main difference from traditional computing is their self-sufficient development through data collection and automated processing, they immediately produce new policy outcomes within this policy field utilising the same advantages examined in the first part. However, since the power political goals of our time are relational, meaning that the lessened power of others equals the amount of power gained in comparison to the said actors, the artificial intelligence system will find ways to achieve this. Doing so will reduce the global societal development capacity. A purely defensively oriented self-sufficient computing machine within the policy of defence and military would be one that produces policy outputs with the overarching goals of genuine knowledge production. Such defence policies would then mainly comprise risk management and improvement of the survival chances of a nation. In the most offensive scenario, policy production by artificial intelligence systems would centre around intervention strategies in nations that are on the brink of falling into systemic failure. This would fall within the eighth stage of devletist foreign policymaking, which predicts that states, after having established high functionality within the first seven stages of foreign policymaking, would take care to prevent failing states from moving too far away from modes of genuine knowledge production in situations such as tyranny or fallbacks into modes of relational goal setting.
Final Remarks
Artificial intelligence systems, so we need to conclude, are merely as good as the goals they work towards. If the goals are of high value and in line with our purpose of existence, namely genuine knowledge production, then those computing systems will bring us very far very quickly. If the goals that we feed those computing systems with, however, are of low value, such as power and wealth, then the systems will bring us very far very quickly within the frameworks of those goals as well. The main difference is that with low-value goals, at the end of those goals, there is always societal collapse. If we do not correct our societal and political goals, then those systems will surely mean the quicker end of highly developed civilisations as we know them today. Then, we will also see whether the machines have developed so much that they become a dominant and conscious species. Hence, the question marks and fears around the development systems are not so much based on the nature of machine learning but rather on the nature of the goals with which we equip those systems. If the measurement of success and purpose are rectified and brought into alignment with the natural and objective truth, namely genuine knowledge production, then the artificial intelligence systems will also develop along those lines and enhance our societal progress.
