The European Central Bank is experimenting with generative artificial intelligence across its operations to speed up basic activities, from drafting briefings and summarising banking data to writing software code and translating documents.
The move comes as central banks cautiously explore how to harness the latest advances in AI, including large-language models such as ChatGPT, though they stress it is a long way before it could be trusted to help set interest rates and other monetary policies.
An ECB unit called the “infinity team” has launched nine trials of generative AI – adapting off-the-shelf systems that can create sophisticated text, images and computer code – after canvassing staff for ideas of where it could be most effective. Myriam Moufakkir, the ECB’s new chief services officer, will announce the move on Thursday in a blog seen by the Financial Times.
“From playing chess to piloting drones – machines have become much smarter in many areas,” Moufakkir wrote. “So why not use artificial intelligence for central banking?”
Trained on large data sets of unlabelled text, generative AI is capable of having a humanlike conversation and producing unique content. While many private sector companies have already seized on the new technology to gain a competitive edge since its emergence last year, central banks have been more cautious due to concerns about reliability, legal risks and transparency.
The areas chosen by the ECB to test generative AI include producing first drafts of briefings, summarising meetings, drafting code for software, improving the language of official communications, translating documents and producing newsletters.
Moufakkir said the ECB was “cautious about the use of AI and the risks it entails”, adding it was “looking at key questions in the fields of data privacy, legal constraints and ethical considerations (such as fairness, transparency and accountability) “.
However, ECB officials say they are working closely with other major central banks to explore AI, including the US Federal Reserve, Bank of England and Monetary Authority of Singapore.
One concern for central bank officials and banking supervisors is if AI is seen as a “black box” that is not transparent on how it produces its conclusions, which would make decisions based on its research hard to defend if they were later subject to a legal challenge.
“An AI black box with no insight into the decision-making process is of limited value,” Lisa Cook, a member of the Fed’s board of governors, said in a speech last week. “As a policymaker, I look upon model-generated forecasts with a sceptical eye, if they are not coupled with a plausible explanation for the driving factors behind them.”
The ECB is already using AI in several areas, including structuring the real-time pricing information it tracks on thousands of products and classifying the data it collects from millions of sources. These are time-consuming and relatively mundane tasks that can be done faster and more efficiently by a computer.
The central bank is also using natural-language processing in Athena, its banking supervision IT system, to scan vast amounts of documents – including news articles, bank documents and supervisory assessments – so it can answer officials’ questions on them more quickly.
“Supervisors can now collate these kinds of enriched texts within seconds, so they can more quickly understand the relevant information – instead of spending time searching for it,” said Moufakkir.
Carlos Bowles, vice-president of the Ipso union that represents ECB staff, welcomed the move to harness generative AI, downplaying fears it could replace humans.
“One needs to mind the associated risks and this will also mean a change of our own jobs as central bankers,” he said. “But my own inclination – as a staff rep – is more about embracing such changes and leveraging them rather than trying to resist something which is anyway unavoidable.”
“With the adequate tools,” he added, “President [Christine] Lagarde could even ask what former president [Mario] Draghi would have done in her situation.” – Copyright The Financial Times Limited 2023
