Bank of England Governor Andrew Bailey warned that there are “clear financial stability risks” from AI – even before taking into account mounting cyberrisks.
That hasn’t, of course, stopped the central bank from baking more AI into its own workflows. The central bank’s Chief Data Officer Ben Martin recently described one example that the bank calls “GRACE.” This layers an email user interface on top of a RAG-based workflow so that staff can interrogate/interact with the central bank’s data using AI – sending their queries via email and getting an asynchronous reply.
Martin added: “We are part way through a multi-year investment programme to build solid data foundations. This includes investing in new systems and technology, but also continuing to invest in the ‘bread and butter’ of good data management…”
At the macro layer, Bailey has concerns however.
In a short speech at Zurich University, published on November 9, Bailey also called for more “extensive model testing, pre and post release [and] better incident reporting.”
Bailey’s talk was billed as one on central bank independence, but AI was a major theme. “We are seeing growing signs of the beginnings of a positive supply shock coming from AI and the investment required for that,” he said – it wasn’t all klaxons.
See also: The Big Interview: Bank of England CIO Nathan Monk
“In the UK, economic activity is turning out to be more resilient than we had expected, and some of that is due to AI investment, particularly in the business services sector,” he said – adding that “we are living through difficult times,” however.
Commenting on a potential asset price bubble in AI, he spelled out four risks.
- AI returns are good, but markets still overvalue them.
- Returns are good, but take longer to realise than predicted.
- Right now “everyone is priced to be a winner”, but “remember Netscape.”
- Public trust, broadly, risks being eroded by AI and this could hit valuations.
Bailey’s comments come after Bill Dudley, a former president of the Federal Reserve Bank of New York, warned in a Bloomberg think piece that “given the rapid depreciation rate of AI data centers, it would likely require $3 trillion or more of revenue each year to generate an adequate return on [current AI] investment…”
Goldman Sachs’s Rich Privorotsky, who is head of European equities trading, commented on October 26 (arguably a lifetime ago in today’s fast-moving markets) that “if you can't get the power to actually energize a data center, will you actually be able to consume the chips you've already forward ordered? So those are concerns that are floating around the market. And then on top of that, you have more competition at the model layer and all of that is just compounded…”
