Wells Fargo Exec: Banks Struggle To Feed Hungry AI

While banks are exploring ways to use artificial intelligence—and, specifically, machine learning—they are struggling with how best to feed these data-hungry models. 

Ravi Radhakrishnan, head of wholesale, wealth and investment management, and innovation technology at Wells Fargo, spoke to WatersTechnology as the bank embarks on research into AI and quantum computing as part of a new partnership with tech giant IBM.

AI models consume a lot of data in their training and in producing predictive outcomes, Radhakrishnan says, and it is reasonable to assume that, as usage of these models increases, the more breadth and depth of data will be needed to make them useful.

“So it is a relevant problem to be thinking about: Is there scientific research or are there methodologies that can use less data than they otherwise would, but still produce near similar types of outcomes? It isn’t any one model of today that has this issue, just an expectation based on trends of data usage,” Radhakrishnan says. 

It is these kinds of questions that Wells Fargo wants to investigate as it has become a member of the MIT-IBM Watson AI Lab and joined the IBM Q Network, groups of academic institutions, large corporates, startups and research labs that are looking into AI and quantum computing, respectively.

David Cox, director at the MIT-IBM Watson AI Lab, says: “One of the Achilles heels of deep learning is that it requires huge amounts of data, and huge amounts of labeled data to work. It is very powerful when you are in that situation, but it makes it hard to apply to many real-world problems.”

Financial institutions do have access to vast quantities of data, Cox adds, “but it is that annotation and curation and labeling that is often the thing that prevents AI from reaching its full potential. So we are very interested in methods that let you do more with less data and less labeling. That is an area where there is a lot of potential in the financial sector.”

Quantum Computing

With regard to quantum computing, Radhakrishnan says, Wells Fargo wants to leverage the partnership with IBM to explore use cases in information security and portfolio pricing and risk calculation, and generally deliver better products and services. “We think of these as technologies with immense potential that we need to be working with, learning from experiments and participating in the advancements of, to ultimately better serve our customers.” Radhakrishnan says. 

The MIT-IBM Watson AI Lab was set up in September 2017, when IBM announced it would commit $240 million over 10 years to fund the lab in partnership with MIT. The lab is intended to deliver research on AI to the healthcare, financial and security industries.

Cox says IBM recently launched a member program that its strategic partners and customers may join if they want to get involved in the lab. Current members include data provider Refinitiv. 

“We don’t intend for the program to be a huge number of companies, but we want to have deep, close, collaborative relationships with a handful of visionary companies that recognize that AI is going to change—and perhaps even disrupt—their industry and want to be on the bleeding edge of that,” Cox says.