Singapore's central bank warns AI isn't ready to handle monetary policy

LLMs show promise, but are 'not yet capable of providing credible explanations for their own predictions'

A senior policymaker at Singapore's central bank, the Monetary Authority, has suggested that artificial intelligence technology is not yet suitable to inform its policy development work.

"Over the past year, central banks have had to answer difficult questions about our collective failure to foresee the persistence of inflation after the pandemic, which has in turn called into question the usefulness of our models," deputy managing director for economic policy and chief economist at the Monetary Authority, Edward S Robinson, declared in a Monday speech delivered at the 2024 Advanced Workshop for Central Bankers organised by National University of Singapore.

"Consequently, we may ask if economists should be paying more attention to recent advances in data analytics and artificial intelligence (AI) technologies to improve our forecasts and models," he added.

Edwards's answer to that question is yes – and no.

His "yes" came from the fact that AI has already demonstrably helped policymakers.

"AI/ML techniques have been used to identify anomalous financial transactions, help supervisors sift through large volumes of text data submitted by financial institutions to identify vulnerable areas, and generate dynamic measures of inflation expectations using social media posts," he explained.

Edwards then praised AI/ML modelling approaches for "their ability to let the data flexibly determine the functional form of the model," as that capability "potentially allows AI/ML models to capture non-linearities in economic dynamics in a way that mimics expert (human) judgement."

Generative AI goes even further: "State-of-the-art large language models trained on vast amounts of data can generate alternate scenarios, specify and simulate basic economic models and beat experts at forecasting inflation," he enthused.

But LLMs have limitations, so Edwards also offered a "no."

"The flexibility of this class of models is also a drawback: AI/ML models can be 'fragile' in that their output is often highly sensitive to the choice of model parameters or prompts provided," Edwards observed. Combined with opacity of output, "this flaw makes it difficult to parse the underlying drivers of the process being modelled."

He also noted that current LLMs "struggle with logic puzzles and mathematical operations, suggesting that they are not yet capable of providing credible explanations for their own predictions."

Edwards therefore suggested the best role for current LLMs in central bank modelling toolkits "is to use them in satellite models that complement core structural models."

"Beyond using AI techniques independently for forecasting tasks, this could extend to 'semi-structural' approaches connecting AI methods to economic theory." He also sees "promising applications" for deep learning as a tool to "estimate economic relationships, such as the Phillips Curve, that underpin standard macroeconomic models."

Edwards observed that MAS's current models were built by "rigorously incorporating the most relevant new developments, while retaining their core theoretical foundations." As the central bank builds its understanding of AI, he feels "we could begin to bring them into our workhorse models in a similar way."

Doing so, he suggested, could improve policymakers' efficiency – as long as organizations like MAS intervene to ensure the technology delivers.

Overall, he suggested AI is a welcome development for MAS.

"The path ahead for economic modelling is an exciting one," he concluded. "Ongoing shifts in the global economy are throwing up new questions for our models, and the techniques we can bring to bear to answer them grow ever richer." ®

More about


Send us news

Other stories you might like