Artificial Intelligence on the Horizon
The Future of Banking
(First in a Series)
Humanity has been on the road for a very long time—from the beginning, when each individual had to collect sufficient food to survive every single day—to the point where we invented agriculture. At that point, we moved from 99% survival and 1% reproduction to a brand new model.
Growing food marked the introduction of leisure. Since then, every step in our evolution has proceeded along the lines of doing more and more with less and less. You might recall the 1899 story of Charles H. Duell, Commissioner of the U.S. Patent Office, lobbying President McKinley for its closure, claiming that “everything had already been invented.” Nonsense, of course (or at least highly apocryphal), but demonstrative of how some people can appear to be short-sighted.
On a timescale where the entire universe was contained within a single calendar year, it would only be in the first hours of December 31st of that year where apes and monkeys had their evolutionary split. At 8:00 PM of that day humans and chimpanzees experienced their evolutionary split. Around 9:30 that night, humans began to walk upright for the first time.
If it was the very last second of that imagined December 31st right now, it was only an hour and a half ago, at 10:30 PM, that our brains began to grow from 1 pound all the way up to 3 pounds—and that was the Dawn of Intelligence. It was only 8 minutes ago that the first Modern Humans evolved. And it was only between 1 and 4 minutes ago that we migrated to all the continents of the world (except Antarctica).
The bulk of human evolution occurred in just the last 8 minutes, but what we identify as our history all happened in the last minute. With only 50 seconds to go before midnight our cave-dwelling ancestors were sketching prehistoric animals on the wall, the most recent Ice Age was coming to an end, and sea level was 400 feet lower than it is today.
About 30 seconds ago, the Mediterranean culture arose giving us agriculture, which led to the first permanent settlements. Just 15 seconds ago marked the beginning of Dynastic China. About 6 seconds ago was the time of the Old Testament and Buddha, with Christ and Mohammed born in the last 3 or 4 seconds. One second ago, Columbus set foot in North America.
The 100-year lifespan of a human on this scale is less than one-quarter of a second. It is only within the last 0.125 seconds that we gained computers, the Internet, microcomputers, cell phones, and the Totality of Human Knowledge doubling every 18 months.
Weliterally have gained twice the knowledge that the Human Race accumulated since we first evolved, in just the last 18 months. In another year and a half, it will have doubled again. And the process is speeding up. Consider this YouTube video made by IBM, showing the world’s smallest movie. It’s a brief (93 seconds) animation called A Boy, and his Atom made entirely with single atoms, manipulated at the atomic level, and magnified 100 million times so we can see it.
We’re learning to do so much, and so fast, that the information is accumulating faster than we can use it. This is a problem. Artificial Intelligence is the answer to the problems encountered in data science use cases in finance, and many other fields besides.
Machine intelligence is the last invention that humanity will ever need to make—Nick Bostrom, TED speaker, Vancouver, B.C.
Movies like Terminator tend to paint pictures about the downfall of humanity because an AI thinks humans are a threat. The movie’s Skynet is probably not something we need to worry about, however.
We should make sure we teach the AI what humans would approve of, and what would be a bad idea. A nebulous goal of “eliminate human unhappiness” might end up with it wiping out the race—no people, no unhappiness, right? We had better have a good, well-described goal for it to work on before we flip the switch. Keep the stories of The Monkey’s Paw and King Midas in mind, and we should be fine. Just don’t expect it much before 2040 or 2050.
Our current AI systems are now quite efficient at interpreting natural language. You can speak to your phone assistant with high accuracy and get the results you want such as weather reports, paying your bills, scheduling of appointments, or finding information online. This is precisely what we need in machine learning use cases in banking.
AI is now achieving >80% accuracy interpreting human facial expressions, the emotional content of sentences; it performs actions millions of times to work out strategies that are successful. It also categorizes those who fail to get the desired results. This amounts to learning, and although it may be primitive, it is still surprisingly good.
People assume things will happen seamlessly, as they have come to experience it on Google and Facebook. Those companies have been employing AI systems and collecting information about us for so many years that it is getting easier to predict what sort of news article, advertisement, or video that we want to see next. Customers want their banking to be the same way, offering precisely what they want, at the time they most likely need it. If you are not using AI, you’ll be seen as archaic, and clients will migrate to banks that fulfill their needs before they need to ask.
Uses of this information for banks mean that they can offer services as appropriate. If a customer buys or leases a new car every five years, it would make good sense to send them a rate-guarantee at 55 or 56 months into their contract, to encourage them to arrange their financing with that particular bank.
If their debt seems to be climbing to an unmanageable level, something a human being might not notice, sending an offer of a consolidation loan at a lower percentage, or arranging a low rate line of credit based on the equity of their home might allow them to keep on top of their finances. Such AI use cases in finance could help assure that the banking institution increases the likelihood of keeping a good customer, and reduces the chance of significant loss due to bankruptcy.
One of the ways to achieve that is through Chat Bot Expert Systems. Collecting all the empirical knowledge of the 100 best heart surgeons in the world and programming it into a robotic surgeon would allow flawless heart surgery (one day, in the far future). Having all that combined knowledge with light-speed calculations and analysis would mean there would always be an answer to a complication. If a human had dealt with the problem before, the AI could solve the problem; if not, it could combine the accumulated knowledge and work out a new solution based on all that experience.
With customers, AI use cases in banking, such as an expert financial system, could take all the information about the client (age, past investment strategies, goals, preferences), and create financial advice. Chat Bots have become so sophisticated that sometimes it becomes impossible to differentiate them from human beings. Those workers formerly answering fairly conventional questions over and over again could then be redeployed to handle more complex issues.
AIs can spot relationships in bulk information that would elude a human being. Companies right now are faced with a massive wave of retiring Baby Boomers. Very few saw this coming despite very clear signs. Now they are scrambling to acquire replacements for their best people.
If AI Technology had been sophisticated enough to be employed by HR, we could have had years of warning. That would have allowed for hiring college graduates and exposing them to the existing experts before they retired, allowing them to mentor these neophytes. All that knowledge would not be lost.
Crooks are getting smarter about fooling the casual observer with their financial actions. It might take a forensic accountant to identify instances of illegal money laundering.
This is not so when you combine machine learning use cases in finance with artificial intelligence. AI, armed with the knowledge of hundreds of forensic accountants, could quickly spot telltale activity. It makes the Federal Reserve, the FBI, and in some cases, the CIA happy; it increases the bank’s reputation; it increases the likelihood of appropriate taxation for the IRS; and, more than likely, it puts a significant dent in crime.
An AI will be able to recognize a customer more reliably than relying on mere encryption and a password. It will know the tempo that a person types, where they hesitate or dwell, their physical appearance on video, and will be able to distinguish between a photo and a blinking, breathing person, with characteristic eye-movements.
It also works for identifying employees for access to restricted areas, or the ability to perform specific actions. It can even identify a pre-actions characteristic of a robbery before it happens, and alert staff, security, and the police before it occurs.
AI can recognize problems much sooner than at Quarterly Report time, allowing corrective actions to be taken immediately, potentially saving millions of dollars per year. This makes it increasingly accurate and eliminates the extraordinary workload of period-end efforts. The financial status, reliability of investments, and all the associated information would constantly be on-hand, and accurate up to that very moment when the inquiry is made.
Asset allocation and forecasting for the bank itself or its customers could allow for much wiser investment decisions. AIs could easily outperform Day Traders once they had been programmed with expert knowledge about market trends, meaning they could make trades to take advantage of the tiniest fluctuations, getting in and out of stock faster than a human could even decide to buy in the first place.
AIs can recognize data patterns and deviations from the norm in real time. If three pensioner’s accounts suddenly transfer their balances to a new account owned by none of them, the AI could freeze that account until a human can investigate. That is only a gross example; AIs could be sensitive to much more subtle clues than that.
Since banks are responsible for these losses once a customer has informed them, it will certainly pay dividends to implement something like this early. Aside from the financial loss, there could also be significant penalties from regulatory agencies if the fraud was determined to be possible because of a lack of proper oversight.
Understanding customers’ tendencies, and how they use the resources of the bank, would be labor intensive for an employee, but child’s play for an Artificial Intelligence. Clients seldom know about all of the offerings of a particular financial institution, whereas an AI will know every single aspect of those services and the tendencies of the individual customer.
It can customize a package for that client that suits their needs perfectly. That creates a happy client that isn’t going to move to another financial institution. You have now achieved customer retention and built loyalty which will be reflected in recommendations to new customers.
Artificial Intelligence is going to have an impact on data science use cases in banking. The same can be said for Finance, Health Care, or Big Pharma. AI will have a significant influence.
Most businesses have recognized the significance of Big Data. The need for Data Scientists has grown exponentially to the point that they’re getting rather hard to come by in this economy. AI is going to help with that.
Once we have some algorithms to teach an AI to analyze the data we can free up a lot of humans to do more critical work. In its current state, this is what AI is best at—the drudge work of data analysis. Whether it’s answering common questions for customers as a ChatBot, or finding fundamental flaws in investment strategies, AI is the grease for our economic engine. It will make everything run better, faster, and at a reduced cost.