Fifteen Eighty Four

Contemporary Artificial Intelligence (AI) is a learning technology that has come into prominence at the same time that humans are learning more about the nature of intelligence. These learning processes mean that what AI will eventually evolve into—and when—is unknown. This uncertainty is one reason why there is much hype and hysteria surrounding the technology.  

In light of this, our book “Artificial Intelligence: Economic Perspectives and Models” aims to help reduce the uncertainty about AI’s contemporary and future economic implications. It proceeds from the basis that economics can and should do more to model AI and improve its models through learning from AI. Better modelling of AI will offer gains in understanding how, why, and when AI affects key economic outcomes such as economic growth, inequality, productivity growth, poverty, innovation and investment rates, wages, and consumption.

Our book illustrates how this can be done. It approaches the modelling of AI from four perspectives: 1) the micro-economics of AI, 2) the macro-economics of AI, 3) the political economy and strategy of AI and finally, 4) the philosophy, ethics  -and science fiction – of AI from an economics perspective.

We are arguing in our book for an “economics of AI” by illustrating the value of economics for AI and vice versa. 

What do we mean by an economics of AI? Let us illustrate this by providing some very brief extracts from our book, which will hopefully motivate the reader to read the hole work!

We mean that economic tools – models – can and should be used more frequently to draw out the consequences of the development and use of AI. One area where economists have been notably silent is the longer-term consequences of continued development in AI. For instance, where will further innovation in AI ultimately lead to? Will narrow AI make way for artificial general intelligence (AGI)? Will this accelerate innovation that will result in “Singularity”? Is superexponential, explosive economic growth possible? Will a future AGI intentionally or unintentionally destroy humanity or, perhaps more likely, be misused by humanity? These are all questions where economists have been relatively silent, leaving the debate to be dominated by philosophers and computer scientists.

With an economics of AI, we also mean that economic models should be updated to reflect how the presence of AI affects their core assumptions. Like software developers issue new updates or versions of their operating systems or programs, economists need to update their models to deal with bugs or new security threats. AI, based on digital technologies and a disruptor of how information is used in economic decision-making, holds radical implications for economists’ assumptions about costs, prices, competition, and distribution.

With an economics of AI, we also refer to economics that draws on AI to better model human decision-making. There are several areas where economists may learn from AI scientists. One such area is procedural rationality. Economic theory tends to ignore the reasoning process by which agents make rational decisions, that is, how agents find the optimum of their expected utility functions – economics has preferred substantive rationality over procedural rationality. Learning from AI about procedural rationality is potentially important for economists because human decision-making is subject to mistakes – even beyond those due to our limited computational ability. 

Economics could also benefit from AI research in the Reinforcement Learning subfield, which may reduce the computational difficulties economists face in understanding reasoning under bounded rationality. AI models could also possibly help to bring more realism into economic models of decision-making by helping to economist to gain better insight into decision-making in disequilibrium situations. “Nonrational quirks” in human judgment may turn out to have its analogy in AI learning that is  subject to bugs.

In conclusion, without an economics of AI, we are likely to obtain less benefit from AI and see more examples of “Awful AI” and unrealistic anxiety about AGI– with the possible unfortunate outcome that AI progress is regulated to a standstill. Most of the chapters in our book illustrate this point using existing and modified economic models to analyze the impacts of AI on the economy, the impact of policies on AI, and the economics of AI in the long run. They show that AI will neither take all our jobs nor lead to the extinction of humanity. However, without adequate consideration of the economics of AI, governments are likely to get AI governance wrong.