Good morning,
Finance and artificial intelligence aren’t like oil and water. There are areas where the two mix, like expense reporting. But when it comes to generative-A.I. applications such as OpenAI's ChatGPT, a financial institution is taking a pass.
This week, there have been reports that JPMorgan Chase & Co. is restricting staff from using the ChatGPT chatbot. The firm’s mandate wasn’t made in response to a certain event but part of standard controls for third-party software usage, the Telegraph first reported. JPMorgan didn’t immediately respond to my request for comment.
Launched in November by OpenAI, ChatGPT is a chatbot that can answer questions and can generate content on any topic you can think of, and even write articles. It's trained to follow language and thought patterns like humans. (Read more about OpenAI founder Sam Altman here.)
To discuss ChatGPT in the workplace, I had a chat with Vikram R. Bhargava, assistant professor of strategic management and public policy at the George Washington University School of Business, who conducts research on A.I. and the future of work.
“I think that a lot of us, including people working in finance, were sort of stunned by the performance of ChatGPT when we first started playing around with it,” Bhargava says. “A number of employees and even banks might be tempted to use these tools to make their life a little easier,” he says. For example, asking it to come up with a relevant Excel formula for a modeling task that an analyst or an associate might do, he explains. But not fully knowing how the technology operates, "does create a little bit of discomfort in heavily relying on it," he says.
“The thing with banking, of course, is that it's a very heavily regulated industry, and this technology is also new to regulators,” Bhargava says. Along those lines, Mira Murati, chief technology officer at OpenAI, told Time in a recent interview that regulators will need to get involved with ChatGPT and govern the use of A.I. in a way that’s “aligned with human values.”
“I don't know the specifics of the rationale behind JPMorgan's decision, but it does strike me as prudent,” Bhargava says. “This technology is rapidly evolving. One of the difficulties is—what might be true of ChatGPT as it stands, might not be true in three months.”
JPMorgan isn’t a novice when it comes to A.I. The bank recently ranked No. 1 in data intelligence startup Evident’s A.I. Index, the first public benchmark of the major banks on their artificial intelligence maturity. The index covers the largest 23 banks in North America and Europe. JPMorgan spends $14 billion in technology annually, of which approximately half is dedicated to investments, the firm said in an announcement.
“Leading in A.I. and knowing how to use A.I. responsibly, sometimes might require the firm to abstain from using the given technology,” Bhargava says.
Michael Schrage, a research fellow at the MIT Sloan School Initiative on the Digital Economy, spoke with finance chiefs at Fortune’s CFO Collaborative event in January about the possibilities of generative A.I. in finance. I asked him his thoughts on JPMorgan’s reported restriction.
Schrage says he’s not certain how OpenAI currently manages, collects, and analyzes “prompts” (how you get ChatGPT to do what you want). But he suggests prompts may be an issue for a bank concerned about privacy rules, compliance, and proprietary processes. Prompts that are too detailed may inadvertently reveal information that the bank or its clients would prefer not to be shared, Schrage says.
"In the same way that Google and Bing know what topics, themes, and names are being searched, it’s similarly probable that OpenAI is tracking the level of detail and specificity of prompts," he says.
Again, Schrage is not sure of how OpenAI handles and tracks prompts, but says: “It’s easy to imagine and enact ways where prompts can be anonymized, aggregated, masked, and shielded to minimize revealing sensitive information while still getting good ‘generative advice’ and insight.” I reached out to OpenAI to ask about prompts, but haven’t received a response.
Many CFOs are already cautious and experimenting with A.I. And, it will be some time before they'd feel comfortable incorporating ChatGPT, Alexander Bant, chief of research for CFOs at Gartner, recently told me.
What would make financial institutions more open to ChatGPT? "They need a little bit more security in knowing how the use of this technology interacts with the current regulatory environment," Bhargava says. But are there perhaps some tasks where a company can experiment without being reprimanded by the Securities and Exchange Commission?
“Let’s say there’s an entry-level employee on your team who might not write the clearest, most concise emails,” Bhargava explains. “So, using ChatGPT might facilitate clearer communication.”
The jury's still out on applying ChatGPT in finance, but generative A.I. isn't going anywhere.
Have a good weekend. See you on Monday.
Sheryl Estrada
sheryl.estrada@fortune.com
