This semi-monthly column by highlights news, government documents, NGO/IGO papers and speeches, industry white papers and reports, academic papers and speeches on the subject of AI’s fast paced impact on the banking and finance sectors. The chronological links provided are to the primary sources, and as available, indicate links to alternate free versions.

NEWS:

TechRadar. August 14, 2024. How Conversational and Generative AI is shaking up the banking industry Banks are more likely to benefit from generative AI (GenAI) than any other industry, according to analysis from Accenture, with a potential productivity boost of up to 30%. This is no surprise when you consider that to take advantage of AI, organizations require stacks of good data – and for the banking industry, data is plentiful. With physical branches closing almost daily, the use of AI to enhance our digital banking experience is on the rise – from improving the customer experience through more efficient service, personalized offerings and greater security.But how are conversational AI (CAI) and GenAI?

Wired, August 10, 2024 – Security researcher Bill Demirkapi found more than 15,000 hardcoded secrets and 66,000 vulnerable websites—all by searching overlooked data sources [unpaywalled]: “If you know where to look, plenty of secrets can be found online. Since the fall of 2021, independent security researcher Bill Demirkapi has been building ways to tap into huge data sources, which are often overlooked by researchers, to find masses of security problems. This includes automatically finding developer secrets—such as passwords, API keys, and authentication tokens—that could give cybercriminals access to company systems and the ability to steal data. Today, at the Defcon security conference in Las Vegas, Demirkapi is unveiling the results of this work, detailing a massive trove of leaked secrets and wider website vulnerabilities. Among at least 15,000 developer secrets hard-coded into software, he found hundreds of username and password details linked to Nebraska’s Supreme Court and its IT systems; the details needed to access Stanford University’s Slack channels; and more than a thousand API keys belonging to OpenAI customers. A major smartphone manufacturer, customers of a fintech company, and a multibillion-dollar cybersecurity company are counted among the thousands of organizations that inadvertently exposed secrets. As part of his efforts to stem the tide, Demirkapi hacked together a way to automatically get the details revoked, making them useless to any hackers. In a second strand to the research, Demirkapi also scanned data sources to find 66,000 websites with dangling subdomain issues, making them vulnerable to various attacks including hijacking. Some of the world’s biggest websites, including a development domain owned by The New York Times, had the weaknesses. While the two security issues he looked into are well-known among researchers, Demirkapi says that turning to unconventional datasets, which are usually reserved for other purposes, allowed thousands of issues to be identified en masse and, if expanded, offers the potential to help protect the web at large. “The goal has been to find ways to discover trivial vulnerability classes at scale,” Demirkapi tells WIRED. “I think that there’s a gap for creative solutions.”

WSJ, August 8, 2024. Mainframes Find New Life in AI Era – Banks, insurance providers and airlines still find uses, including artificial intelligence, for the large computers that have been around for decades. The industrial-strength mainframe computer, invented decades ago for heavy-duty data processing, is proving its staying power even as next-generation artificial intelligence takes center stage. The two are interconnected: Banks, insurance providers and airlines are a few of the big industries that still rely on the mainframe for high-speed data processing. And now, some are looking to apply AI to their transaction data at the hardware source, rather than in the cloud. For banks, analyzing potentially fraudulent transactions must be accomplished in milliseconds. “You can’t go off to the cloud, go do a search, go do a generative AI query, because it will just time out,” said Steven Dickens, vice president and practice leader for cloud at Futurum Group, a research and advisory firm. That means performing a query, looking it up and analyzing it can only happen in real time, on the mainframe, he said. “Everyone’s kind of realizing that it’s better to bring your AI to where the data is, than the data to the AI,” said Ross Mauri, general manager of International Business Machines’ mainframe system. IBM reported 6% growth in its mainframe business during the quarter that ended June 30. Part of the mainframe’s staying power is that it can support huge numbers of transactions, as many as 30,000 a second, at super fast speeds. The latest version of IBM’s zSystem mainframe can hold up to 40 terabytes of data, weigh up to 1,800 pounds and cost upward of $1 million. IBM is by far the biggest player in mainframes, with over 96% of the market, followed by NEC, Fujitsu and Hitachi, according to International Data Corp. Forty-five of the top 50 banks and four of the top five airlines still use mainframes as their core platform, IBM said. Globally, the mainframe market was valued at $3.05 billion in 2023, but new mainframe sales are expected to decline through 2028, IDC said. Of existing mainframes, however, 54% of enterprise leaders in a 2023 Forrester survey said they would increase their usage over the next two years…Banks, telecom keep mainframes in the mix. New York-based Ponce Bank uses a mainframe for its loans business, Chief Executive Carlos Naudon said, which involves a “batch system” running overnight to create a set of files that get uploaded all at once. The bank said it is using the mainframe along with its cloud-based digital and mobile banking platforms. Some businesses still prefer managing their own mainframes and servers. “If you’re in the cloud, you’re relying on Amazon or whoever else is running that cloud,” Naudon added. Telecom giant Verizon Communications similarly uses a combination of cloud with its own data centers and mainframes, said Kamran Ziaee, senior vice president and chief information officer for technology strategy and global infrastructure. Still, efforts to do away with the mainframe have been going for years, especially as the number of developers conversant in Cobol—one of the primary programming languages used in mainframes—quickly dwindles…

The Business Times, August 7, 2024. Detecting anomalies with AI: How financial companies can spot risks faster and reduce downtime. So intertwined and interdependent are today’s complex IT systems spanning the globe that a web server error, a programming glitch, a network issue or, worse, a cyber attack could result in a chain reaction that disrupts not just a website, but also a firm’s core operations. The stakes are extremely high. These days, a bank service that goes down for hours will draw the ire of customers and cause them to lose their trust. Beyond reputational damage, it could lead to a steep loss of business and financial penalties from government regulators. In today’s fast-paced and hyper-connected world, businesses and organisations must know how to respond quickly to these issues and prevent them from happening again. To do so, they need to be able to monitor their IT systems in real time. But how can they do that? The key lies in the need for real-time diagnostics. Consider the warning lights on a car’s dashboard which help pinpoint specific problems in the machine. The driver would be able to get them resolved immediately without opening the bonnet and taking apart the engine to investigate, allowing him to quickly return to the road.

FedScoop, August 7, 2024. Financial agencies’ AI tests could get reprieve from enforcement. A bipartisan, bicameral bill would allow designated financial regulatory staffers to test the technology’s tools without the threat of such actions. Designated officials within federal financial regulatory agencies would be able to experiment with artificial intelligence tools without fear of enforcement actions under a bipartisan, bicameral bill introduced this week. The Unleashing AI Innovation in Financial Services Act from Sens. Mike Rounds, R-S.D., and Martin Heinrich, D-N.M., and Reps. French Hill, R-Ark., and Ritchie Torres, D-N.Y., would institute “regulatory sandboxes” for AI test projects at the Federal Reserve, the Securities and Exchange Commission, the Consumer Financial Protection Bureau, the Office of the Comptroller of the Currency, the Federal Deposit Insurance Corp., the Federal Housing Finance Agency and the National Credit Union Administration. Experimentation within those sandboxes could be done “without unnecessary or unduly burdensome regulation or expectation of retroactive enforcement actions,” the legislation states.

PAPERS:

NBER – Deep Learning for Economists. Melissa Dell. Working Paper 32768. DOI 10.3386/w32768. Issue Date August 2024. Deep learning provides powerful methods to impute structured information from large-scale, unstructured text and image datasets. For example, economists might wish to detect the presence of economic activity in satellite images, or to measure the topics or entities mentioned in social media, the congressional record, or firm filings. This review introduces deep neural networks, covering methods such as classifiers, regression models, generative AI, and embedding models. Applications include classification, document digitization, record linkage, and methods for data exploration in massive scale text and image corpora. When suitable methods are used, deep learning models can be cheap to tune and can scale affordably to problems involving millions or billions of data points.. The review is accompanied by a companion website, EconDL, with user-friendly demo notebooks, software resources, and a knowledge base that provides technical details and additional applications.

IMF Working Papers Enhancing IMF Economics Training: AI-Powered Analysis of Qualitative Learner Feedback. August 2, 2024. Komaromi, Andras ; Wu, Xiaomin ; Pan, Ran ; Liu, Yang ; Cisneros, Pablo ; Manocha, Anchal ; El Oirghi, Hiba. The International Monetary Fund (IMF) has expanded its online learning program, offering over 100 Massive Open Online Courses (MOOCs) to support economic and financial policymaking worldwide. This paper explores the application of Artificial Intelligence (AI), specifically Large Language Models (LLMs), to analyze qualitative feedback from participants in these courses. By fine-tuning a pre-trained LLM on expert-annotated text data, we develop models that efficiently classify open-ended survey responses with accuracy comparable to human coders. The models’ robust performance across multiple languages, including English, French, and Spanish, demonstrates its versatility. Key insights from the analysis include a preference for shorter, modular content, with variations across genders, and the significant impact of language barriers on learning outcomes. These and other findings from unstructured learner feedback inform the continuous improvement of the IMF’s online courses, aligning with its capacity development goals to enhance economic and financial expertise globally. Download PDF

SPEECHES:

Crisis Amplifier? How to Prevent AI from Worsening the Next Economic Downturn. IMF First Deputy Managing Director, Gita Gopinath. AI for Good Global Summit, Geneva, Switzerland. May 30, 2024.Let me describe how AI could worsen the next downturn, starting with labor markets. The experience with previous waves of automation offers a warning here. During good times, firms are often flush with profits. They can afford to invest in automation and hold on to workers, even if the value-added of those workers declines. However, in a downturn, these firms simply let go of workers to cut costs. Therefore, the extent to which automation could replace humans only becomes fully visible during or immediately after a downturn.

NGO/IGOs:

AI Preparedness Index June 25, 2024 – The IMF’s AI Preparedness Index assesses the level of AI readiness across 174 countries. Explore interactive maps and charts of key areas of preparedness.

Generative artificial intelligence and cyber security in central banking BIS Papers |  No 145  |  23 May 2024 by Iñaki Aldasoro, Sebastian Doerr, Leonardo Gambacorta, Sukhvir Notra, Tommaso Oliviero and David Whyte – Generative artificial intelligence (gen AI) introduces novel opportunities to strengthen central banks’ cyber security but also presents new risks. We use data from a unique survey among cyber security experts at major central banks to shed light on these issues. Responses reveal that most central banks have already adopted or plan to adopt gen AI tools in the context of cyber security, as perceived benefits outweigh risks. Experts foresee that AI tools will improve cyber threat detection and reduce response time to cyber attacks. Yet gen AI also increases the risks of social engineering attacks and unauthorised data disclosure. To mitigate these risks and harness the benefits of gen AI, central banks anticipate a need for substantial investments in human capital, especially in staff with expertise in both cyber security and AI programming. Finally, while respondents expect gen AI to automate various tasks, they also expect it to support human experts in other roles, such as oversight of AI models.

IMF, June 25, 2024. Mapping the World’s Readiness for Artificial Intelligence Shows Prospects Diverge – Artificial intelligence can increase productivity, boost economic growth, and lift incomes. However, it could also wipe out millions of jobs and widen inequality. Our research has already shown how AI is poised to reshape the global economy. It could endanger 33 percent of jobs in advanced economies, 24 percent in emerging economies, and 18 percent in low-income countries. But, on the brighter side, it also brings enormous potential to enhance the productivity of existing jobs for which AI can be a complementary tool and to create new jobs and even new industries. Most emerging market economies and low-income countries have smaller shares of high-skilled jobs than advanced economies, and so will likely be less affected and face fewer immediate disruptions from AI. At the same time, many of these countries lack the infrastructure or skilled workforces needed to harness AI’s benefits, which could worsen inequality among nations. As the Chart of the Week shows, wealthier economies tend to be better equipped for AI adoption than low-income countries. The data draw from the IMF’s new AI Preparedness Index Dashboard for 174 economies, based on their readiness in four areas: digital infrastructure, human capital and labor market policies, innovation and economic integration, and regulation.