Stay informed with free updates
Simply sign up to the Artificial intelligence myFT Digest — delivered directly to your inbox.
Why are so many people in America (and elsewhere) so negative on AI? There are many reasons: horror at tech companies’ excess power; fears about impending job losses; alarm about AI’s impact on mental health, energy and the environment.
Then there is a practical issue. While AI can theoretically deliver miracles — in life sciences, say — few people have yet seen much tangible improvement in their own lives. Win-wins are proving elusive.
Techies and governments are now trying to address this public relations problem. In the UK, for instance, Doug Gurr, former chair of the Alan Turing Institute, tells me that the Met Office is revolutionising flood forecasts with AI, and that the Natural History Museum can now identify dinosaur bones 30,000 times faster. Jurassic Park is optimised.
But finance might soon produce another win-win too. As Michael Hsu, formerly one of America’s top bank regulators, told the Cambridge Digital Innovation and Regulation Initiative (which I co-convene), AI could also slash regulatory red tape. If so, that would not just help harassed regulators and bankers but consumers of financial services as well.
“AI is a force multiplier,” says Witit Synsatayakul of the Asian Development Bank, which has published a potent recent paper on the topic.
The key issue is what Cass Sunstein, the behavioural economist, calls “sludge” — the fact that 21st-century life is beset with red tape, generating “excessive or unjustified frictions, such as paperwork burdens, that cost time or money”. In 2015 alone, Americans spent 9.78bn hours on federal paperwork.
That makes “life difficult to navigate” for citizens because sludge is “frustrating, stigmatising, or humiliating”, he notes, calling for “firms, universities, and government agencies [to] regularly conduct Sludge Audits” — and cut it.
That sounds sensible. And groups like the Financial Conduct Authority are trying: the FCA’s Helen Packard said on Thursday that the regulator will save £100mn for firms by streamlining rules — even without AI.
But this “sludge-busting” can be a nightmare too. Hsu says that when the Reserve Bank of India tried to turn 9,445 circulars into a more rational 244 “master directions”, it took almost 40 staff an entire year. And when Congress introduced a law in 1996 requiring US regulators to audit — and cut — rules every 10 years, it proved so “extraordinarily labour intensive” that “the mere planning for the [review] process” took over a year.
The problem is multi-faceted. There is “vertical sludge”, where top-down laws become increasingly complex when turned into practical rules. And there is also “horizontal slush”, where adjacent regulators create overlapping rules, and “time slush”, when new rules are piled on top of old ones.
And now we have “geopolitical slush”: conflict and rising protectionism mean that national regulators are less eager to co-ordinate laws. Just look at what Brexit has done or how US regulators reject new Basel banking rules.
But this is precisely why AI could help, given its ability to scan swelling mountains of text, then synthesise and check it for inconsistencies, gaps and overlaps.
Stanford University has already partnered with San Francisco city leaders to pilot this idea: a team recently studied municipal rules (totalling 16mn words), then focused on the legislatively mandated reports to find duplications. The legislature then cut and consolidated 36 per cent of them. Yes, really. “Because of the length of our code . . . it’s likely a project we would never have undertaken [without AI],” notes David Chiu, city attorney.
Now regulators want to replicate that in finance. And it seems a good moment, given that institutions like the European Central Bank and the UK Treasury face mounting political pressure to slash red tape. Moreover, policymakers face an “arms race” in dealing with bankers’ own AI innovations, a top UK regulator says.
So can this sludge-busting work? It won’t be easy, alas. One problem is that large language models are better at tackling conversational or literary texts than regulators’ ghastly legalese. Another is the limited AI skills of regulators, while a third is the danger the machines might go mad without a human in the loop, or impose excessive standardisation.
But the technical problems are now being solved by new innovations. And Hsu says new legal tools such as Harvey, Irys, CoCounsel, Norm AI and Kimi are outperforming mainstream platforms like Claude and ChatGPT for regulatory tasks. Regulators are hiring staff with AI skills and pledging to maintain all-important human oversight.
Will that cut financial rules by a third, as happened in San Francisco? It seems an ambitious dream. But any reduction would help. So three cheers for red-tape-reducing robots — they might yet make markets more efficient and perhaps encourage moneymaking humans to be a bit more optimistic about AI as well.
[email protected]
This article has been amended to reflect that the FCA will save firms £100mn
