Acemoglu, one of the Nobel laureates in economics, argues that if AI continues on its current trajectory and is unregulated, it could harm competition, consumer privacy and choice, over-automate work, exacerbate inequality, unduly depress wages, and fail to increase productivity.
Text: Fan Shuo, researcher of Caijing
Edit | Guo Liqin
On October 14, local time, the Royal Academy of Sciences of Sweden announced that it would award the 2024 Nobel Prize in Economics to three economists from United States universities: Darong· Acemoglu, Simon ·Johnson and James · Robinson in recognition of their contributions to "studying how institutions shape and affect prosperity". The three economists will split the 11 million Sweden kronor prize money equally.
According to the official website of the Nobel Prize, Daron · Acemoglu was born in Istanbul, Turkey in 1967, received his doctorate from the London School of Economics and Political Science, and is currently a professor in the Department of Economics at the Massachusetts Institute of Technology in United States. Simon · Johnson was born in Sheffield, United Kingdom in 1963, received his Ph.D. from the Massachusetts Institute of Technology and currently holds a position at the MIT Sloan School of Management. James · Robinson, born in 1960, earned a Ph.D. in economics from Harvard University and currently teaches at the Harris School of Public Policy at the University of Chicago.
Over the years, the three economists have collaborated on a number of books. Acemoglu co-authored Power and Progress: Our Millennial Struggle for Technology and Prosperity with Johnson, and co-authored Economic Analysis of Political Development: The Economic Origins of Authoritarianism and Democracy and Why Do States Fail? The latter has aroused great academic and public opinion influence.
Among them, Acemoglu is the favorite candidate for the Nobel Prize in Economics predictions in recent years. In the 2011 United States Economist Survey, Acemoglu ranked third on the list of "Most Popular Living Economists Under 60," behind Paul · Krugman and Greg · Mankiw. In 2015, he was named the most cited economist of the last decade, according to data from Research Papers in Economics (RePEc).
Quasimoglu was on the balcony of a hotel in Athens when he received a call from the Nobel Prize Organizing Committee for the Nobel Prize in Economics, just before giving an academic lecture.
It is worth mentioning that Acemoglu is very concerned about the latest developments in AI (artificial intelligence) and is cautious and worried about AI. He told Nobel Prize staff that the overall development of the social system was the main determinant when faced with the disparity between poor and rich countries. At the same time, Acemoglu also expressed his concerns about technology: "AI has a lot of potential, and if we don't use it properly, it will be the main reason for further exacerbating inequality." Some AI industry players collect and manipulate data, further weakening democracy. And it's going to really lead to the emergence of a bipolar society, and I think we're already starting to suffer from that. ”
Previously, on October 8, 2024 Nobel Prize winner in physics Gajeffrey · Hinton, although known as the "godfather of AI", said in an interview with the media that winning the Nobel Prize means that the public will pay more attention to his "AI threat theory" view. Hinton resigned from Google in 2023 because he believed that any further development of AI would definitely threaten humanity.
It proves the importance of institutions to the prosperity of a country
Commenting on the reason for the award, Jacob ·Svensson, chair of the Nobel Prize Committee in Economic Sciences, said: "Closing the huge income gap between countries is one of the greatest challenges of our time. The winners demonstrate the importance of social institutions in achieving this goal. ”
Nie Huihua, a professor at the School of Economics of the Chinese People's University, wrote that in 1993, United States economist Douglas · North won the Nobel Prize in economics. One of his major contributions was to study the rise of Netherlands and United Kingdom in the 16th and 18th centuries, and to find an important conclusion: efficient systems for protecting property rights were a key factor in promoting economic growth. Thereafter, the question left for economists to prove is: how do you prove that institutions are the cause of economic growth, rather than that economic growth in turn shapes good institutions?
In their 2001 article "The Colonial Origins of Development Differences" in the United States Economic Review (AER), Acemoglu, Johnson, and Robinson used "mortality among local colonizers" as an instrumental variable (IV) to demonstrate the causal relationship between institutions and economic growth. According to Nie Huihua, the research of the three economists has solved the problem of identifying the two-way causal relationship between institutions and economic growth.
The Nobel Prize Committee in Economics retraced the colonial movement that the three economists focused on in their research: when Europeans were colonizing the globe, colonizers in some places aimed to exploit indigenous populations for profit. In other places, colonizers established inclusive political and economic systems for the long-term benefit of European settlers. The laureates' research shows that one reason for the different levels of prosperity in countries is the different social systems introduced during colonization. Over time, people in countries that have introduced inclusive institutions tend to become generally wealthy. Some countries are mired in predatory social systems and low economic growth. The introduction of an inclusive system would bring long-term benefits to all, while a predatory system would only bring short-term benefits to those in power.
The Nobel Prize Committee in Economics noted that the correlation between institutions and prosperity does not necessarily imply a two-way causal relationship. Rich and poor countries differ in many ways, and there may be other reasons for differences in development, apart from institutions. Acemoglu, Johnson, and Robinson found that an important explanation for the current difference in prosperity is the political and economic system that colonizers introduced from the 16th century onwards, or chose to keep. Using data such as the mortality rate of colonizers, they found a causal relationship: the higher the mortality rate of colonizers, the lower the GDP per capita (gross domestic product) today. This also means that colonizer mortality means how "dangerous" it is to colonize an area, which in turn affects the type of institutions established by colonizers.
In the city of Nogales, on the border between United States and Mexico, for example, Nogales is bisected by a fence to the north and is United States Arizona, where residents are relatively affluent, have long life expectancy, most children complete high school, and people have relatively safe property and the right to free elections. The southern part is Nogales, Sonora, Mexico, and although this is a relatively wealthy part of Mexico, the inhabitants there are generally much poorer than those on the north side of the fence. Organized crime makes starting and operating a company risky. The three economists argue that the decisive factor that leads to differences is not geography or culture, but institutions, and that Nogales is part of the root pattern of the colonial era.
Be skeptical of AI
In the 2024 Nobel Prize selection, the public saw the "conquering of the city" of AI, and many awards were awarded to AI scientists, among which Jeffrey · Hinton is known as the "Godfather of AI" and pushes the "AI threat theory". The Physics Prize was awarded to John · Hopfield of Princeton University in United States and Jeffrey · Hinton of the University of Toronto Canada for their "fundamental discoveries and inventions that have advanced the use of artificial neural networks for machine learning"; The Chemistry Prize was awarded in part to Demis ·Hassabis and John · Jiangpo, scientists at Google's DeepMind, for their achievements in protein structure prediction of the AI model "Alpha Fold2".
Acemoglu has also been skeptical of AI.
In his 2021 article for the Oxford AI Governance Handbook, The Dangers of AI, Acemoglu argued that if AI continues on its current trajectory and is unregulated, it could harm competition, consumer privacy and choice, overautomate work, exacerbate inequality, unduly depress wages, and fail to increase productivity. In addition, AI has the potential to distort political discourse and undermine the foundations of democracy.
For example, AI's massive demand for data can lead to privacy violations, unfair competition, and behavioral manipulation. In data sharing, an individual's data has an impact not only on themselves but also on others, which can lead to externalities in the data, which can be both positive and negative. In some cases, data marketplaces can be inefficient, resulting in data transactions at prices close to zero, while users may actually be willing to pay a higher price to protect their privacy. After the introduction of AI technology, consumer surplus is likely to decrease, as businesses can use AI technology to better predict consumer behavior and adjust prices accordingly.
Acemoglu also suggested that appropriate regulation is needed to limit the potential harms of AI as market solutions may not be sufficient to solve the problem. The development of AI is closely related to international competition, so global regulatory measures are needed. Given that vetting is likely to become more difficult with the widespread use of AI technology, the principle of precautionary regulation should be adopted to slow down the use of AI technology.
In a study released this year, Acemoglu argues that AI's productivity boost will be limited, contributing less than 1% to United States economic output over the next decade. The debate is now about whether AI can effectively automate complex tasks and thereby spur significant economic growth.
According to Acemoglu, AI can currently only automate a small fraction of routine tasks and is not effective in assisting employees with more complex problems. For AI to significantly increase productivity, it needs to be able to automate about 40% of work tasks.
This view contrasts sharply with Goldman Sachs' predictions, which predicted that generative AI could boost global GDP by as much as 7%, equivalent to nearly $7 trillion.
Acemoglu used Total Factor Productivity (TFP) to assess the productivity gains of AI. He argues that AI is particularly inefficient when faced with "hard-to-learn tasks", such as decision-making tasks that rely heavily on context.
Acemoglu's calculations show that in the next ten years, AI will not drive TFP by more than 0.66%, which is equivalent to an average annual growth of about 0.064%; When it comes to identifying and tackling difficult tasks, the overall growth of TFP is capped at 0.53%.
Acemoglu believes that the impact of AI on GDP growth will be greater than that of TFP, as automation and task complementarity (partial automation makes work easier) will lead to more investment. But he estimates that the growth will be similarly limited: if AI brings in limited, modest investment, overall GDP growth over the next decade will be between 0.93% and 1.16%; If AI triggers a massive investment boom, overall GDP growth over the next decade will be between 1.4% and 1.56%.
In addition, he pointed out that even though the new tasks and products brought about by AI will boost GDP growth, not all of the contributions will be positive. Because this technology may generate some new jobs that are manipulative, nasty, although there will also be other new jobs that will compete with it. For example, generative AI has great potential and promise, but only if it can be used primarily to provide better and more reliable information to humans. The current returns of generative AI remain elusive unless there is a fundamental reorientation across the industry, including significant changes to the most ubiquitous models, such as large language models (LLMs). The industry change he proposes is to focus on enabling AI to provide more reliable information to improve the marginal productivity of different types of jobs (e.g., educators, healthcare professionals, plumbers, etc.), rather than prioritizing the development of generic human-like dialogue tools.