Artificial Intelligence & Machine Learning , Finance & Banking , Geo Focus: Asia

Singapore to Banks: Don't Rely on AI to Forecast Inflation

Monetary Authority Says: Stick to Accepted Financial Models But Experimenting Is OK
Singapore to Banks: Don't Rely on AI to Forecast Inflation
Image: Shutterstock

Singapore has urged banks to be cautious when using artificial intelligence tools to forecast inflation and to stick to time-tested structural models.

See Also: Strengthen Cybersecurity with Zero Trust Principles

The Monetary Authority of Singapore warned Monday that the use of AI tools and techniques by banks to forecast inflation or build economic models could backfire, considering that AI tools lack "clarity of structure" and have not been tested rigorously enough to replace existing structural models.

Edward Robinson, the authority's chief economist and deputy managing director for economic policy, said in a workshop organized by the National University of Singapore that banks and financial institutions are currently adopting AI techniques for various use cases, but they should bear in mind how unpredictable the economy can be, as evidenced by the historic surge in inflation during and after the pandemic. Headline inflation more than doubled to 4.1% in 2022, and consumer prices rose by 6.1%.

MAS said it responded to rising inflation by tightening monetary policy five times and releasing two unscheduled monetary policy statements. No one predicted the widespread impact on interest rates and banks all over the world.

Robinson said AI has made major inroads into the banking sector in recent years and has demonstrated superior accuracy in understanding complex supply chain systems, identifying anomalous financial transactions and conducting financial supervision and macroeconomic surveillance.

Recent advancements in generative AI also have enabled advanced large-language models to capture non-linearity in economic dynamics, generate alternate scenarios, specify and simulate basic economic models, and beat experts at forecasting inflation. But the growing flexibility of LLMs can quickly turn into a major risk for banks.

"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," Robinson said. "Together, with their opacity, this flaw makes it difficult to parse the underlying drivers of the process being modeled."

He added that modern LLMs "struggle with logic puzzles and mathematical operations" and are incapable of providing credible explanations for their own predictions, making it difficult for banks and financial institutions to rely on them entirely to forecast future macroeconomic trends.

"Without the ability to articulate a vision of how the economy works or discriminate between competing narratives, AI models cannot yet replace structural models at central banks," he said.

Considering recent advancements in AI/ML technologies, Robinson said, banks should consider incorporating AI techniques in satellite models or smaller projects that complement core structural models, such as using deep learning models to estimate economic relationships.

"Our current models have been built up by rigorously incorporating the most relevant new developments, while retaining their core theoretical foundations. As we improve our understanding of the mechanics underlying AI techniques, we could begin to bring them into our workhorse models in a similar way," he added.

Concerns over bias and inaccuracy in AI tools and large language models have grown in recent years, forcing regulators to incorporate new rules to monitor AI usage and test their effectiveness for overcoming bias and ensuring data privacy, security and transparency.

The Monetary Authority of Singapore worked with leading Singaporean banks in November to develop a risk framework for the use of generative AI for the financial sector.

MAS said the initiative, named Project MindForge, focuses on developing a clear and concise framework for the responsible use of generative AI in the financial industry and on promoting AI innovation to solve common industrywide challenges and enhance risk management.

The framework covers accountability and governance, transparency and explainability, fairness and bias, ethics and impact, monitoring and stability, cyber and data security, and legal and regulatory compliance.

"As the financial industry continues to explore the potential of generative AI technology, it is crucial that we develop a clear and concise framework for its responsible application," said MAS Chief Fintech Officer Sopnendu Mohanty. "MindForge aims to address common challenges and catalyze AI-powered innovation in the financial industry, while ensuring that this technology is used in a responsible and sustainable manner."

About the Author

Jayant Chakravarti

Jayant Chakravarti

Senior Editor, APAC

Chakravarti covers cybersecurity developments in the Asia-Pacific region. He has been writing about technology since 2014, including for Ziff Davis.

Around the Network

Our website uses cookies. Cookies enable us to provide the best experience possible and help us understand how visitors use our website. By browsing, you agree to our use of cookies.