Finance industry forced into AI but will maintain human touch
Artwork: Unsplash via Prisma
Finance industry forced into AI but will maintain human touch
TIM LEEMASTER
Financial institutions, forced by competitive pressures, are hard at work bringing artificial intelligence (AI) to bear on the industry, but banking will maintain its long-standing human touch even after the issues the technology clearly has are ironed out, financiers say.
Financial institutions are fundamentally creating a new digital ecosystem - based on data, AI and delivery – that, at this point in time, makes it hard to understand just what it will look like even in a few years, according to Cyril Cottu, global head of ecommerce and digital at BNP Paribas.
Price suggestion is where AI is currently most firmly rooted in the financial sector, according to Sandeep Arora, chief operating officer at Citi Markets and Securities Services. “Traders just make faster and better decisions,” Arora said as banks crunch content and data across all the different areas they do business.
That can be problematic, though, as machine learning always finds a pattern even if there really isn’t one there, according to Marcos Lopez de Prado, head of machine learning at AQR Capital Management.
The hedge fund, one of the largest in the world with over USD 200bn under management, instead is trying to look at broader themes, or what it calls “theories”, that underpin market developments or movements instead of specific price and trading options. That can allow for a broader approach to risk management and the construction of more diversified investment portfolios, Lopez de Prado said.
Going forward a key AI application financiers are actively working to leverage are the data streams from chat applications and email. Both are considered archaic communication forms by AI practitioners but when tracked with natural language processing they can be used to generate predictive analysis on trading desks and elsewhere.
“The sort of thing you can imagine is the streaming information of all the chats between the traders and clients and then what is the word cloud of what people are talking about and then is that going to suggest trade ideas and is there a good way of getting that to clients,” said Charles Elkan, global head of machine learning at Goldman Sachs.
Mixing in, or separately drilling down into, the masses of daily and archival news flows can also provide direction both internally at financial firms and for their clients, who rely on them for information and advice. “What is it that people are fearful and greedy,” and then internal systems can benchmark between the “rational” machine and the “irrational” human, Lopez de Prado said.
The group pointed out that datasets today are fundamentally different than before, where they can make up anything from the constant stream of words in chats, satellite imagery, temperature readings etc. and AI is needed to digitize all of it.
Financial firms will have the capability to create a better path to action, according to BNP’s Cottu. Overall there will be more quantitative or data-driven decision making as opposed to decisions made from the top, Arora added.
Some AI applications will be less directly related to overtly financial or trading issues. Making the wide swath of contracts a bank has accumulated over time, along with the myriad clauses that can go with them, “understandable and at our fingertips particularly when the people that have negotiated them might have moved on” is one attractive goal, according to Elkan, who previously worked at Amazon’s cloud computing unit.
Citigroup, for example, is currently using AI to analyze the implications of the UK exit from the European Union, or Brexit, which recommends changes to the business contracts most likely to be affected, Arora said.
Despite all the automation humans will remain very important in terms of testing and making sure the AI works appropriately, Arora said. And the client relationships that banking and finance have long relied on will continue, Cottu added.
Other issues, besides the technical, confronting financial institutions as they move deeper into the world of AI are the limitations that compliance, particularly privacy concerns, pose and the costs associated with the rollout of an all-encompassing new technology. Two difficult to navigate concerns but Arora pointed out the industry has no choice. “It’s a necessity…companies have to invest in this space,” he said.
Arora added there are off the shelf offerings that can provide an AI template over which banks and others can seed with their own proprietary data to reduce development time and costs.
The group was speaking on a panel at a recent technology forum in New York at the International Swaps and Derivatives Association.