Conclusion
Artificial Intelligence technologies are one of the critical elements of change in the short-medium term. The advent of a new GPT (which could very well be considered GMI46) will deeply change the dynamics of the economy, politics, society, and international relations. The multidimensional character of technology influences the power dynamics between those who will be able to exploit the altered scenario, and those who will not.
However, it would be fundamentally wrong to adopt a perspective in which we are merely victims of this transition. While this process of change cannot be stopped, individual and collective actors can guide and shape its direction, by establishing norms and institutions that minimize the damage and maximize the benefits. While science and technology can be used to pursue political goals, in their essence, their basic principles are politically agnostic, and only through awareness of the scope of the impact of these new technologies policymakers can aim to maximize the social welfare of the AI revolution.
Unfortunately, today, one can observe an increase in the competition between the major countries in the AI industry: the US and China (Benzell et al. 2019), which have launched in a technological race. While this could be expected, since dominance in AI technologies is strictly linked to economic and political supremacy, blind competition is particularly problematic since there is the concrete risk that it would bring to the overseeing of keystone aspects of AI development, such as its ethical use. As it was presented in chapter 1, although AI systems represents a fundamental improvement in prediction technologies, they cannot be considered intelligent in the same way humans are, and they are marked by flaws and dangers that are very difficult to forecast or detect. They can safely be used only in situations where the cost of error is minimal, while for more complex situations they need to be under human oversight. With the decrease of prediction cost, the value of judgment increases, especially in complex situations, where the negative consequences of AI mistakes increase, and human intervention is needed both for security and monitoring purposes.
In chapter 2, some of the main normative constraints that shape the development of new technologies were presented: Intellectual Property Rights. The first part evaluated the main economic arguments put forward to justify their existence and concluded that the only valid descend from an utilitarian perspective, which does not presume the introduction of temporary monopoly rights without taking into account the negative consequences they generate into the economy. Then, the three IPRs most commonly applied to AI technologies were presented, with a specific focus on the economics of patents. Finally, economic analysis was applied to AI technologies, with the objective to verify whether the claims made by legal scholars on the need to introduce additional IPRs applicable to AI are justified or not. However, I found that current evidence does not justify this approach, since no additional incentives are needed to promote innovation in AI, and since there is a strong tendency towards self-regulation using patentleft or copyleft models, this calls towards a reduction in IPRs, rather than the opposite.
Chapter 3 focused on the effects of AI technologies on innovation and the market. The first part centered on some technology categorization, namely General Purpose Technologies, Inventions of a Method of Inventing, and General Methods of Invention. Particular attention was given to GPTs, since they represent the focus of chapter 4 and there is a rich literature on their effects on the economy. A crucial point regards the phenomenon of innovation complementarities, which describes why GPTs represent a game-changer in terms of economic growth. Managing the virtuous cycles of innovation by minimizing the negative impact on employment will represent a crucial challenge for XXI century policymakers. The second part of the chapter focused on determining whether AI technologies possess some characteristics of these technological classes. Using a qualitative perspective, it is safe to affirm that AI presents features of both GPTs and IMIs, a combination that the literature could only find in digital computing. AI is configuring itself as a rapidly evolving, widely applicable, and complementary technology while at the same time capable of expanding the knowledge frontier of science and increasing the rate of scientific discovery. The AI revolution will invest a large fraction of economic sectors, incentivizing investments in AI research and development, with effects that will bounce from one sector to another. Although the invention of multi-function closed-loop Autonomous Discovery Systems is far from being a reality, AI technologies are likely to cause an unprecedented increase in productivity in academia, opening up entire new paradigms of science. Finally, the chapter concludes by presenting some of the consequences that AI is having on market structure. In particular, the presence of economies of scope is leading to a process of market concentrations between a small number of firms that compete with each other in innovation, conquering new markets and having higher and higher barriers of entry, represented either by extreme network effects or the exclusive possession of fundamental assets such as users’ data. Considering that the AI market structure leaves little space for new entrants, the competition in innovation is likely to increase, especially now that the COVID-19 pandemic has prompted world economy in a deep economic crisis, opening opportunities for a redistribution of global economic power. At the same time, state and regional actors are trying to manage the digitization of the economy, with the European Union at the forefront. In particular, two pieces of legislation are in the process of being evaluated by the European Parliament: the Digital Market Act (DMA) and the Digital Services Act (DSA), while a third piece of legislation specifically targeted to AI technologies is expected in the first quarter of 2021. The DMA aims to reduce the market power of gatekeepers, multi-products platforms that have extreme prescriptive and indiscriminate power on what it can and cannot be done in their platforms, while at the same time potentially competing with actors using their platforms. The DSA is instead focused on establishing obligations and accountability for intermediaries’ platforms that provide digital services, and enhance consumer protection, whether the services are located in or outside the EU. These sets of norms are characterized by having extraterritorial character, since their effectiveness is tied to the individual, and not to the legal site of the firm, and it is conforming to the trend of EU law launched by the GDPR. Differently from the imposition of additional IPRs, these sets of norms aim to improve competition and consumer protection, possibly minimizing the negative consequences of a widespread introduction of AI technologies.
Finally, chapter 4 objective’s was to provide quantitative evidence to the claims of chapter 3: whether AI is a General Purpose technology, and whether we are assisting to a market concentration process of the AI market. The data used to perform the analysis was formed by a subset of PCT patents issued from 1995 to 2017. Most of the chapter focused on using traditional indicators, such as the generality index, and developing new indexes based on the evolution of the technology network formed by AI-related patents. The empirical evidence shows that AI is indeed a GPT, and that the number of AI-related patents is increasing at high rates, pointing towards the fact that it is rapidly pervading our economy in a variety of different application sectors. However, patent data is intrinsically limited, and the results of the analysis on market concentration must be treated carefully, as it can only be deducted that the patent portfolios of strong patentees is increasing when compared to those of weak patentees, thus suggesting that some companies are heavily investing in AI technologies.
Being at the dawn of the economic explosion of AI, treating it as a GPT opens up new prescriptions for policy intervention. From a purely economic standpoint, we are likely to assist in increases in the innovation rate in the short term, causing disruptions in the labor market. Some jobs will cease to exist because human labor will be substituted by artificial labor (Abrardi, Cambini, and Rondi 2019), while new professional figures capable of building, managing, or operating AI systems will be needed, creating a mismatch in the labor market. If policymakers will not draft effective redistribution policies, there is a substantial risk of deepening existing inequalities between social groups, leading to unrest and discontent among the “losers” of the AI revolution. Acosta, Peña, and Saalfeld (2019) studied the relationship between income inequality and political polarization, identifying how economic factors have a deep impact on voting behavior. In particular, they confirmed the assumption that governments parties get penalized in periods of poor economic performance, and proved that income inequality has become a main driver for the success of far-right political parties, while structural unemployment favors far-left parties. More interestingly, they also affirmed that political polarization was reduced in contexts where income inequality was countered by specific policies, either in the form of progressive taxation or by addressing the job market mismatch. Technological advancement can be considered one of the factors at the root of political polarization but, at the same time, while its direction and pace can be guided and influenced, it cannot be stopped. The Luddite fallacy is an ever-present threat when dealing with technological innovation, however economic and political analysis show us that states where short-sighted emergency policies have not solved structural problems, and we assist to a rise of political extremism, while at the same time they lag behind other countries in terms of economic performance, up to a point where the resources to make structural reforms lack. At the same time, excessive regulation in the digital arena may lead to a reduction in the economic competitiveness in the international scenario. The extent to which a country will be successful in dealing with the AI revolution thus will be largely depend on its ability to balance incentives for technological innovation and, the implementation of labour policies that prepare the workforce to switch occupation towards the new opportunities offered by AI technologies, and ensuring competitiveness in the digital market, that has now reached a prominent place in the economy.
AI technologies are configuring as a key factor of change in the international and national economy. They are expanding the borders of what is technically possible and conceivable by human societies, but with great power comes great responsibility. If not properly guided, mismanaged innovation can have devastating effects on social cohesion, potentially paving the way for increasing inequality and authoritarian regimes. Ethical use of AI technologies is key to ensure a future that is sustainable, rich, and fair. While regulation of what can or cannot be produced may be seen as a cost in the short term, if properly designed it may lead to enhanced social welfare in the medium-long term.
References
Abrardi, Laura, Carlo Cambini, and Laura Rondi. 2019. “The Economics of Artificial Intelligence: A Survey.” Vol. 58. Robert Schuman Centre for Advanced Studies Research Paper No. RSCAS.
Acosta, Christian Proaño, Juan Carlos Morales Peña, and Thomas Saalfeld. 2019. “Inequality, Macroeconomic Performance and Political Polarization: An Empirical Analysis.” Bamberg University, Bamberg Economic Research Group.
Benzell, Seth, Nick Bostrom, Erik Brynjolfsson, Yoon Chae, Frank Chen, Myriam Côté, Boi Faltings, Kay Firth-Butterfield, John Flaim, and Dario Floreano. 2019. “Technology Trends 2019: Artificial Intelligence.” WIPO.
A technology widely applicable in a variety of fields, with strong innovation complementarities and that can potentially be used during the R&D process, shifting entire scientific paradigms and drastically increasing the innovation rate.↩︎