AI-HPC Convergence in Financial Services: Getting Real about Real Time
Intra-day rebalancing of risk portfolios initiated by voice command while at home, central bank-backed cryptocurrencies, transaction clearing and settlement cut from three days to a few hours – these and other capabilities could become realities in the financial services industry (FSI) as exascale-class HPC converges with evolving AI, blockchain, IoT and 5G technologies.
Discussed at Tabor Communications’ HPC & AI on Wall Street conference last week in New York, one of the most interesting ideas raised during the “2020 Prediction in Financial Services” session came from panel moderator Addison Snell, CEO of industry analyst firm Intersect360 Research, who said, “I think risk management is one of the most fascinating areas in finance, and it actually drives more investment in HPC than any other.”
As of now, he said, the “best of breed” in this area is rebalance risk portfolios on an overnight basis, “which means I know I’m in the right spot at 8 in the morning.” But Snell foresees a day when the high performance compute and AI infrastructure is in place for “getting to some kind of intra-day check on my risk portfolio, or toward the idea of real-time risk management in response to changing market conditions. That’s exascale-level stuff, and I think it’s going to require predictive analytics and require an awful lot of HPC.”
Shahin Kahn, senior partner at technology consultancy Orion X, said “I’ll double down on that, and point out the emerging area of risk accounting – there are a few academic projects trying to quantify risk in a way that can account for it. So just like accounting for your financial balance sheet, you’d have accounting for your risk balance sheet. And if you were to do that, that’s another reason why you need all this exascale-HPC-AI stuff.”
Eric Simone, CEO of edge software company Clearblade, added to Snell’s point by quoting his company's CTO: “‘If you’ve seen it on Star Trek, it’s probably going to happen.’ Seriously, I imagine a future where you’re in your house and you want to rebalance your portfolio, you’re going to say out loud, ‘Hey I want to rebalance my portfolio,’ and you’re going to get an answer within 250 milliseconds. Because that’s real time, it’s going to be ubiquitous edge computing, that’s how you get to real time.”
He added that 5G alone “is not going to solve this problem, as much as AT&T and others want to tell you that, 5G is part of an ecosystem. But it’s going to be real-time compute on premises. Once we get done with Google and Amazon and everybody else fighting for your home or anywhere else, there’s got to be some standardization here, but it’s going to come.”
Nodding in assent, Ryan Quick, principal at tech consultant Providentia Worldwide, added: “I’m ready for real real-time. We hear the traders say it would be nice to ingest opinion, as opposed to observing something that you should have done. I think it’s coming, it’s going to take a little while, and it’s going to take a lot of effort.”
Snell emphasized that “real time occurs in two different contexts in this environment. Most people, when they say real time, they mean fast enough that it feels interactive. But then there’s an engineering definition of real time where you can’t have an answer that’s both late and correct – the timeliness is part of the actual validity of the answer, that means ‘hard real time.’”
The foundation for achieving this, of course, is comprehensive digital transformation, which Kahn defined as creating an environment in which “everything that you do lends itself to digitization, it’s a digital view of life, you need to digitize all the interactions that take place in all the processes that take place within your purview… You need to harness your data so it’s not dark, you need to turn it into real time streaming information, you need to have a data supply chain and have a determination to extract value from everything that you’ve now digitized.”
Part of that data supply chain, Snell said, could combine AI with high performance storage to ensure the right data is in the right place (tier) at the right time.
“One thing that fascinates me is all the tiers of storage we have now…, it used to be tiered storage was either disk or you had tape. Now you see five, six, seven tiers, between archive and cloud and all that. I think we’re going to see an evolution of high performance storage to be (about) … how do I manage the data between all these different tiers? I think in the medium term this is an excellent use of AI, a predictive storage environment that monitors your usage patterns and tries to promote data to the hot tier before you request it. That’s something I’d like to see – not only in finance but in a lot of other industries.”
As the FSI increasingly adopts AI, data – AI’s driver – will become increasingly valuable. Dan Simerman, head of financial relations for distributed ledge organization the IOTA Foundation, anticipates the formation of exchanges where data is traded like other commodities.
“OK, so this is a couple of years off,” he said, “but I’d hands down go with data marketplaces. I think once we are able to capture this data and then create data exchanges, and people can speculate off of data, I think that’s going to be a whole new vertical in the world of finance – but there’s a lot of steps to get there.”
Optimized data usage combined with blockchain could add up to faster completion of financial transactions.
“OK, it’s not the most exciting idea,” said Asha Dakshinamoorthy, head of product for security token and cryptocurrency software provider AlphaPoint, “but I think it’s going to come down to faster clearing and settlement. The concept of settling in two to three days is going to go away.”
With broadening acceptance of blockchain and digital assets, Dakshinamoorthy also foresees – possibly as soon as next year – a time when nation states, rather than bitcoin or companies like Facebook (with Libra), issuing their own cryptocurrencies.
“In 2020, what I’d like to see is the first central bank issue currency in a major market,” she said. “I think that will really unlock the security token and other investment space because no one wants to buy real estate with bitcoin when the price is so volatile. We need a government-backed, stable coin from a central bank, and I think we could see that next year.”