AI sweeps Wall Street: Hedge funds deploy ChatGPT, banks start an "arms race"

Source: The Paper

Reporter: Wu Tianyi Intern: Liu Zitan

· Wall Street is exploring the latest popular tools, mainly ChatGPT, hoping that by providing sufficient financial information, the machine can reach the level of ability to reasonably price options, build investment portfolios or analyze company news.

About 40% of job vacancies at the most AI-enthusiastic banks were AI-related hires, such as data engineers, quants, and governance roles.

Image source: Generated by Unbounded AI tool

Many, including billionaire investor Warren Buffett, see the enthusiasm for adopting complex AI systems as a harbinger of risks to come.

The AI revolution is underway in the financial world.

Hedge funds are deploying ChatGPT to do the heavy lifting, Deutsche Bank uses artificial intelligence (AI) to scan client portfolios, and ING uses AI to screen potential defaulters. Morgan Stanley said it is experimenting with AI technology in a safe and controlled environment. At the same time, JPMorgan Chase is absorbing AI talent far and wide, offering more related job openings than any rival.

** "Hard work, tiring work" are handed over to AI**

In fact, AI has long been used in various tasks on Wall Street, such as machine learning algorithms that calculate credit risk. At this stage, Wall Street is exploring the latest popular tools based on ChatGPT, hoping that by providing sufficient financial information, the machine can reach the level of ability to reasonably price options, build investment portfolios or analyze company news. Financial practitioners need to process a large amount of financial text data, such as news, reports, ratings, etc., and even deeper program code writing, and AI is gradually taking over.

According to a Bloomberg report on May 31, some hedge fund companies said that generative AI has been used to handle common tasks such as market research review, basic code writing and fund performance summary. All tasks will be digested and completed by AI.

Kevin Cole, CEO of systematic quantitative hedge fund Campbell & Co, revealed that the firm's quants use big language models to summarize internal research reports. Kevin said that AI is very powerful in code completion, editing, and error checking. However, human employees will still intervene. Current generative intelligence tools have not yet reached the point where they can change the way humans invest on a daily basis.

According to a Bloomberg report on June 1, Eigen Technologies, a company that helps companies such as Goldman Sachs Group and ING Group to launch AI business, said that inquiries from banks increased five-fold in the first quarter of 2023 compared with the same period last year.

According to Alexandra Mousavizadeh, CEO and co-founder of consultancy Evident, the launch of ChatGPT in November 2022 “will allow everyone — the board, the CEO and the bank’s Leadership -- more awareness that this is a game changer." She described the situation as an "AI arms race".

Deutsche Bank is deploying deep learning technology to analyze whether international private banking clients are overinvested in a particular asset and match individual clients with suitable funds, bonds or stocks. Subject to regulatory compliance, human advisors deliver AI-generated recommendations to customers.

“I really like this way of combining artificial intelligence with human intelligence,” said Kirsten-Anne Bremke, global head of data solutions at Deutsche Private Bank International.

JPMorgan has a similar plan. The company in May filed a patent for a ChatGPT-like service that helps investors choose specific stocks, according to people familiar with the matter, and the project is still in its infancy.

BNP Paribas is using chatbots to answer customer questions and AI to detect and prevent fraud and money laundering. Société Générale is using the computing power of AI to scan for potential misconduct in capital markets.

In April of this year, Morgan Stanley said it had applied for a patent for a model that aims to detect the direction of monetary policy by using AI technology to classify the Fed's information as hawkish or dovish.

On March 21, Marco Argenti, the chief information officer of Wall Street investment bank Goldman Sachs, said that internal developers have begun to use generative AI to program. "It's still early days for this technology. We're not going to hand over all important work to AI right away, but the priority is to really try and understand the potential of AI." Argenti sighed, "I've been working in technology for almost In 40 years, it's one of the biggest disruptions I've seen -- probably on par with the internet, apps, cloud computing."

Risks and Security Issues

But this trend has raised concerns about the transparency and effectiveness of financial AI. Many, including billionaire investor Warren Buffett, see the enthusiasm for adopting sophisticated AI systems as a harbinger of risks to come.

Bankers have a fiduciary duty not to trade on unreliable information. That's a problem as AI applications expand, said Anne Beaumont, a partner at law firm Friedman Kaplan Seiler Adelman & Robbins in New York. "How do you prove to investors and regulators that you've done your job when you're using AI to answer the question without knowing what it was?"

In April this year, South Korea's Samsung Electronics was exposed to three incidents in which employees used ChatGPT to leak confidential data. The company's employees uploaded the company's system program code to ChatGPT, asked AI to help fix errors and improve the program code, and input meeting minutes into ChatGPT, instructing AI to help with key sorting, resulting in confidential data such as factory performance and output becoming GPT model training part of the data.

Wang Pengbo, chief analyst of Broadcom Consulting Research Institute, once pointed out that the use of ChatGPT by financial practitioners, even if the user is unconscious, is likely to cause the leakage of personal information and data.

In addition, whether it is GPT-4 or other large language models, there is a problem of AI "hallucination (Hallucination)", which occasionally creates information out of nothing and fabricates information. If an analyst uses ChatGPT to generate a research report, the content seems to be quite credible, but if the report sends out false information and mistakenly becomes bad or bullish news for a listed company, it will cause very serious consequences.

In addition, the AI model grabs information training data directly from the Internet, which may contain some copyrighted works. For a bank or a news agency that provides financial information, the move is considered copyright infringement and could damage the company's reputation.

"The financial industry is a strictly regulated industry, and it is particularly sensitive to personal information and related business data." Wang Pengbo said that when financial institutions use ChatGPT products, they must first evaluate the controllable range and take preventive measures.

Carlo Giovine, a partner at consulting firm McKinsey, said that when working with banks and insurers, they are redesigning their risk frameworks to account for intellectual property considerations, an uncertain regulatory environment and AI " Hallucinations" and other risks.

Cole of Campbell & Co said that the company is experimenting with an open source model internally. Although the processing power is not as good as ChatGPT, the advantage is that the entire system can be deployed and run locally. "We have to be very careful about the risk of a breach with such tools," Kerr said.

High development and running costs

The banking industry has been no stranger to using technology to its advantage in recent years, recruiting data scientists, machine learning experts and even astrophysicists. Today, those investments are starting to bear fruit.

Bank of America Chief Executive Brian Moynihan said in April that AI could bring "tremendous benefits" and help reduce headcount, but caution should also be exercised.

Barclays is still in the "serious research" phase of financial AI. It will take "many years" to implement any AI tool across the company, CEO CS Venkatakrishnan told a conference.

Lewis Liu, CEO of Eigen, a research-based AI company, pointed out that the development and operation costs of AI are high, and the processing of complex financial documents involves a lot of cloud computing costs. "These large language models are really unwieldy. You need to be more specific, and probably use smaller fine-tuned models that are better suited to your use case."

Regarding the current state of the industry, McKinsey's Giovanni said: "We are in the hype cycle now, and you can see how quickly the industry has developed. Some banks have begun to realize what it takes to really achieve this goal, but many banks are still there. Trying to understand.” At the same time, he noted, companies need to identify areas where AI can really help and develop a roadmap with senior management, while training employees and hiring more experts.

New changes in human resources in the industry

According to the latest data from consulting firm Evident, around 40% of job vacancies at banks most enthusiastic about AI are AI-related hires, such as data engineers, quant analysts and governance roles. JPMorgan is leading the hunt for talent, with data showing the bank posted 3,651 AI-related job openings globally from February to April, nearly as many as rivals Citigroup and Twice that of Deutsche Bank.

Greg Bond, chief executive of Man Numeric, a subsidiary of Man Group in Boston, sees the “digital workforce” as a “digital workforce” for employees who lack technical expertise but are creative and able to ask the right questions. Posts can be an opportunity. "They'll multiply your existing research and technical staff," Bond said, "but ultimately, if we can automate the innovation process itself, that would be a good thing."