In the end of last month, Amazon Go opened its doors in Seattle, as the first digital grocery store without any cashiers. Your shopping bag automatically identifies and adds up everything you gather, so you can just walk in and out and everything is simply charged to your credit card. To get started, visitors simply download a special app to use the digital shopping cart and check-out without waiting in line.
As this store format shows, machines can do many interesting things that make our lives easier. Machine learning is one area banks are focused on to gain a potential lift in the future.
Definitionally speaking, machine learning technology gives computers the ability to learn without explicit programming. IBM coined the term in 1959 while playing around with pattern recognition and data. Forbes reports machine learning "is a current application of artificial intelligence" around the concept that machines are used to access data and "learn for themselves." This is not to be confused with artificial intelligence (AI), which is "the broader concept of machines being able to carry out tasks in a way that we would consider 'smart'."
Hang in there as we go a bit deeper to try and clear out this technical fog. Machine learning has many amazing applications and some very important ones for bankers. It shows great promise for catching money launderers and other financial fraudsters for example. That said, the technology is perhaps 5Ys from major adoption here, because worries around regulatory approval and data have kept some financial institutions on the sidelines awaiting further guidance.
Machine learning's ability to create neural networks may be spooky to some, but that very complexity and strength may also be helpful to banks. Of course, it may also be one of the greatest barriers to more widespread adoption even as a tool against money laundering.
Eager to harness the new technology while also staying on the right side of regulators, some large banks employ data analytics that use some elements of machine learning. These usually also still let people direct the software's pattern search to ensure the bank is following regulations, law enforcement advisories, investigation results, etc. Doing so allows these banks to comb through mountains of customer data seeking patterns across multiple transactions quickly, while ensuring a human is also involved to steer things along.
Precision of any system of course goes back to the old saying of garbage-in-garbage-out. That means predictions depend a whole lot on having accurate input or data. Unfortunately, that isn't simple to supply for most banks.
Antifraud efforts can sometimes be enhanced with machines. For instance, PayPal says its internal algorithms have led to a 50% decline in false positive fraud alerts, big banks are actively using it and the SEC uses it to scan documents for fraud too.
Machine learning and AI continue to be promising technologies, so community banks should start at least thinking about how to use such technologies to boost opportunity and reduce costs and time over the next 5Y to 7Ys.
