BSP taps AI to improve economic data-gathering

The Bangko Sentral ng Pilipinas (BSP) has ramped up its exploration of artificial intelligence (AI) since 2018 to improve economic data collection, BSP-Department of Economic Statistics senior director Redentor Paolo Alegre Jr. said over the weekend.

During a seminar hosted by the Economic Journalists Association of the Philippines in Baguio, Alegre said the move comes in response to the growing need for efficient data collection, especially after the pandemic hampered traditional methods.

"AI helped when the pandemic hit and data gathering was hampered by movement restrictions," Alegre said.

"(But) only when the economy restarted to reopen that we started to really invest in AI," Alegre added.

According to the BSP official, more budget has been allocated for AI usage to help commensurate the work they do in gathering information on the economy. However, he declined to give any figures on the funding.

Last March, BSP said it was researching how generative AI could affect the banking sector in the Philippines to reduce possible risks.

In an interview with reporters, BSP Governor Eli Remolona Jr. said the central bank is examining the risks of generative AI as banks have unavoidably started investigating machine learning and other programs like ChatGPT.

Using generative models, generative AI could produce new and realistic material including text, pictures, videos, or other data. These models produce new outputs depending on the data they have learned from after being trained on vast volumes of data.

“The issue here is what guardrails we can introduce to the use of generative AI. In other words, what safety features we can introduce,” Remolona told reporters.

“If you use it for different analyses, there’s a tendency for what we call ‘herding.’ They tend to give the same answers for some of the different questions. Herding is something that is a possible source of systemic risk,” he said.

Hence, he said generative AI and machine learning algorithms for data analysis and decision-making still need ongoing human supervision.