As artificial intelligence goes mainstream, a growing number of international organizations are urging economic policy makers around the world to prepare for disruption while figuring out how to take advantage of the transformative technology.
The Bank for International Settlements, often called the central bank for central banks, joined the chorus on Tuesday with a chapter of its annual report dedicated to AI.
“The rapid and widespread adoption of AI implies that there is an urgent need for central banks to raise their game,” the BIS said, noting that large language models and generative AI could impact everything from productivity to inflation to financial stability.
“Central banks need to upgrade their capabilities both as informed observers of the effects of technological advancements as well as users of the technology itself.”
The chapter on AI was released Tuesday in advance of the full BIS annual report June 30.
Last week, the International Monetary Fund published a report urging fiscal policy makers – the officials who determine government taxation and spending – to be ready for the possibility of widespread job losses and growing inequality, as industries are transformed by computers capable of doing cognitively complex tasks.
The IMF report said countries should consider broadening unemployment insurance and doubling down on worker retraining programs. At the same time, it suggested AI tools could be used to deliver a range of public services more efficiently.
“Policy should bring about conditions that steer innovation and deployment in ways that harness the benefits of gen AI and serve collective human interests, and it should be ready to cushion the transition costs for workers and households and prevent rising inequality,” the report said.
The disruptive potential of AI burst into the public consciousness with the launch of OpenAI’s ChatGPT in late 2022. The “chatbot” can respond to prompts in a manner that is eerily human, and conduct a range of complex tasks, from summarizing documents to generating images. Excitement about AI has sparked a stock market frenzy for companies with exposure to the technology.
Hyun Song Shin, economic adviser and head of research at the BIS, said AI will “undoubtedly” transform the economy, although how that will play out across different macroeconomic variables – inflation, GDP growth, unemployment – remains to be seen.
Take labour markets, for example. Workers in white-collar professions previously immune to automation could find themselves out of work. But AI could also improve productivity in many industries and lead to whole new categories of jobs.
“Whenever you have this big structural change in the economy, completely new tasks, completely new jobs, completely new professions will spring up,” Mr. Shin said in an interview.
When it comes to inflation, the key concern for central banks, AI could cut both ways. Rising productivity tends to be disinflationary, while faster economic growth tends to be inflationary.
On balance, Mr. Shin said the widespread adoption of AI could be slightly inflationary as it feeds economic growth. But that does not necessarily mean a higher inflation rate if central banks control things with higher interest rates.
“In the standard macro models, we tend to see that higher growth rates tend to go hand-in-hand with higher equilibrium interest rates. That will be consistent with, if you like, central banks having to lean against the wind,” he said.
These are theoretical debates central bankers and other policy makers will need to wrap their heads around in the coming years. In the meantime, they should be looking for ways to harness AI to do their jobs better, according to the BIS.
For central banks, that means using AI tools to take the temperature of the economy and improve forecasting. AI and other machine learning tools can interpret large amounts of non-conventional data, which could give central bankers insight that goes beyond the numbers provided by statistical agencies such as Statistics Canada.
“It doesn’t have to be statistical series that you feed into these models. It can be anything: you can feed in satellite photos, social media text, you can feed in anything and they will take that context when they do the forecasting,” Mr. Shin said.
“The big proviso here is that you need the timely data. [AI models] are very good at finding the needle in the haystack. But you need the haystack and you need a needle.”
The Bank of Canada is already using machine learning tools for “nowcasting” – understanding the current state of the economy. These tools can scrape media articles or earnings calls to provide “sentiment analysis,” or interpret huge quantities of data tied to electronic payment systems. The BoC is not yet using machine learning tools for actual forecasting, and has yet to introduce generative AI tools.