As the calendar turns to a New Year, Bank Innovation turns to a deeper consideration of what might be the underlying “futuretrends” in fintech. This is not the “mobile is the next big thing” type of list you see all over. (Some of those lists are just nausea-inducing.) No, we are trying to go beyond the obvious to uncover ideas that are not always apparent as we deal with day-to-day work.
To be sure, we may be well ahead of reality here. Some of these six futuretrends might simply not matter in the here and now, or maybe even over the next few years. We’ll take criticism for that, as long as it makes you think.
One more note before getting into the heart of it: a thanks to Patrick Collison of Stripe for tweeting a link to this wonderful collection of articles on the great accomplishments in technology and science in 2015.
1. AI Comes to Finserv.
Let us say that you are not feeling well, maybe feverish what with all the cold weather (man, it is brrr in New York today). You ask your wife what medication you have in the cabinet — she tells you Dristan. Is Dristan advertising to you? What if you ask not your wife, but an artificial intelligence application what medication can help you with your fever? If that app answers, “Dristan,” is it advertising? Or, better yet, do you care? It is this sort of evolution in AI that is coming, and has profound implications for financial services. PFM is over. PFM without AI is like a winter coat without a zipper: it just doesn’t do enough.
Here’s how Roger Schank, a psychologist and computer scientist (how’s that for a combination), explains it:
Very soon AI programs will be good enough (not because they analyze key words or do “deep learning”) but because they can model situations and can match situations to what people have said about those situations. Imagine a video data base of hundreds of thousands of experts. Well “how would I search through all those stories?” is the natural question. We ask that question because searching is an everyday activity now and it has taught us to believe in search and every one selling AI espouses the usefulness of key words.
But it is not key words that will cause this breakthrough. There is too much information to search through and often what we need isn’t there in the first place. But this is not actually a search problem. It is a problem not unlike the getting the right ad to the right person at the right time problem. It is a question getting computers to have a model of what you are doing, what your goals are, and matching that to what help they might have to offer.
Finserv clearly has such models. Automated underwriting is built on such models. As unappealing as this might sound to the gatekeepers of bank fee income, such models are going to have to be opened to consumers. In the end, I believe that it will be financially beneficial to FIs, but no matter: the AI revolution will mandate it.
2. Biology Will Mesh with Financial Services.
Prudential Insurance, you know, the one with the rock in the logo, offers this explanation of financial planning:
The passage of time alters the nature of the risks, and requires you to reconsider how you are managing them and whether you should reprioritize them at different times of your life. Life-cycle financial planning helps you understand the dynamic nature of the financial risks presented and develop a plan that evolves over time to meet those changing needs.
But is “life-cycle financial planning” really on point? Here, I will share a personal story. My father has been a judge for a long time, essentially since 1980. He is now 74. Last year, his health took a turn and he had to retire. Is this a life-cycle event or a health-cycle event? According to Prudential, “after work ends” takes place from “ages 65 and on” (“As your working years come to an end, your priorities will shift from saving to living off your savings,” is how Prudential frames it). But for my Dad, “after work ends” was at “age 74 and on,” and for others, “after work ends” no doubt started at 60 or 50 or, God forbid, at 40. There is no integration of biological realities with financial services today — but there will be. Actually, there must be.
First, we have the algorithms for it; we just need better datasets. In 2011, IBM’s Watson became the world “Jeopardy!” champion using a variant of the mixture-of-experts algorithm published 20 years earlier. But to do so, it utilized a dataset of 8.6 million documents from Wikipedia, Wiktionary, Wikiquote, and Project Gutenberg updated one year prior.
From Alexander Wissner-Gross, a scientist:
[T]he average elapsed time between key algorithm proposals and corresponding advances was about 18 years, whereas the average elapsed time between key dataset availabilities and corresponding advances was less than three years, or about six times faster, suggesting that datasets might have been limiting factors in the advances. …
[W]e might already possess the algorithms and hardware that will enable machines in a few years to author human-level long-form creative compositions, complete standardized human examinations, or even pass the Turing Test, if only we trained them with the right writing, examination, and conversational datasets. Additionally, the nascent problem of ensuring AI friendliness might be addressed by focusing on dataset rather than algorithmic friendliness—a potentially simpler approach.
Although new algorithms receive much of the public credit for ending the last AI winter, the real news might be that prioritizing the cultivation of new datasets and research communities around them could be essential to extending the present AI summer.
This syncs well with the revolution of “deep learning” networks that go beyond data analysis by using “back propagation” of algorithms. Apparently — and this is well beyond my mathematical capabilities — this is done by using the “chain rule,” a calculus trick that “enables complex families of potential solutions to be autonomously improved with vector calculus.” OK, so that’s pretty out there, but what we need to know is that this “deep learning” allows for computers to solve problems that are not programmed, and could take the notion of AI even beyond what we assume is possible today — which, again, makes the meshing of biological information and financial services a possibility.
3. The Weaponization of Financial Services Will Amplify.
Even a cursory review of the US-Iran geopolitical nightmare makes plain the truth that financial services will increasingly become “weaponized” in 2016 and beyond. At the heart of the US-Iran tete-a-tete are new sanctions against Iran for its ballistic missile program. And when the White House talks about “sanctions,” in a large part this refers to financial sanctions via the Treasury Department, which runs around 30 sanctions programs today. It seems fair to say that as financial services become only more pervasive, whether as a result of the NFC payments revolution or by way of more dynamic digit finserv, the effects of the “financial services weapon” can only be expected to be more profound. What happens when a nation is no longer able to facilitate payments by credit cards? What are the implications for a nation’s economy when it cannot utilize SWIFT? And would anyone suggest that those implications will become less pronounced in 2016 and beyond?