I just returned from Money 2020 where behavioral biometrics and authentication solutions had a major presence and where AI and machine learning were sprinkled over existing marketing messages in an attempt to make what was old, new again. It was telling however that for the most part nobody on the booth could provide any machine learning details. Was the machine learning unsupervised or supervised? Was the solution deterministic or non-deterministic? In most cases all I could illicit were blank looks.
This suggests that anyone planning to market a solution as having an AI/machine learning component should be prepared to indicate how many models are implemented and where and how they are used. For example, we could describe Mercator’s GrayOwl solution as having three machine learning models; one to identify if the target object was already in our repository; the second to identify if the object was associated with payments or not; and the third to categorize and rank the document into one of several payment categories. Each model required a unique training approach and a weakness in any of the first two would have negative consequences on the last. Where possible suppliers should consider delivering a similar checklist.
We learned that First Data has a Token Service Provider solution in pilot with MasterCard and an unnamed bank. The power of this is primarily associated with First Data’s ability to link the token issuance with the token acceptance, establishing a closed loop environment that, at minimum, should make it easier to operate incentive and loyalty programs for both the issuer and the merchant. Given the number of individual manufacturers and multiple products those manufacturers produce, becoming a TSP will require significant resources and is a huge commitment. That said, I’m not sure we have thought through all the benefits for a company like First Data that can tie the issuing side of the house directly to the acquiring side – this will need some additional analysis.
I also recently attended BankAI where I presented a Business Manager’s Machine Learning 101 class. If anyone is interested in the slide deck I used for that 1 hour presentation, just let me know! I expect machine learning will fundamentally change our lives and how we conduct business and the impact is already being felt. While the primary impact today is primarily associated with advanced data analytics, as utilized in recommendation and fraud detection engines as well as natural language processing, it will ultimately change the software industry and enable new solutions for business process automation. Add to this specialized machine learning solutions that will drive our cars, power robots, and perform other complex tasks on our behalf and it becomes difficult to predict what this world will look like in 10 – 20 years.
As such, it is no surprise that machine learning has been central to several recent Mercator publications, including a report that explains machine learning titled “Bringing AI into the Enterprise: A Machine Learning Primer” “. Machine learning is also discussed in the publications “Behavioral Biometrics Will Restructure the Authentication Landscape in the Next 5–8 Years” and “Biometrics: A New Wrinkle Changes the Authentication Landscape” because machine learning analyzes the behavioral biometric signals collected from end users as they hold the mobile device and traverse web sites.
If interested in a short intro to machine learning I’d suggest these two short videos. The first shows how a machine learning solution is set up to learn how to play Super Mario Brother and tracks its learning progress through several generations; from idiot to savant. The other video presents a Carnegie Mellon project that implements a relatively dumb sensing device, but then harnesses the signals that device receives from its environment to recognize an unbelievable range of activities that occur in that environment. In my opinion, a few of these $100 devices placed in branch locations would enable Financial institutions to identify everything happening in the branch, from when the doors are opened and closed to the entire range of human activity within.
I have commented on many recent events in Payment Journal, some of the more interesting articles are listed below:
Authoritarian Governments Announce Their Own Cryptocurrencies
Target Joins Walmart on Google Home; But This May Be More About a V2 Shopping Experience
Behavioral Biometrics: ECommerce Sites Validate ID Before You Even Logon
A Great Idea, But How Quickly Can the W3C Payment Request API Be Adopted?
The ICO Bubble, Separating Nerds From Their Money
Machine Learning ForeSight
Identify the long term impact ML will have on Payments, software development and business.
Emerging Technology Outlook
A look at the Emerging Technology issues we predict will have a major impact in 2018 and beyond.
APIs & Cloud Computing: Changing Payments and Enabling Platforms and Markets
From Fintech & Networks to Major Banks, API’s have become the new method for delivering and consuming services.
A Review of the Market for Tokenization
With new payment devices becoming more common, including conversational commerce devices and browsers that support the W3C standard, it is time for a new look at the market for tokens. A Primer on the Pays!
I’d appreciate learning what’s on your mind regarding the exciting world of payments innovation and how you feel I should steer my research, so please drop me a line!
All the best!
Vice President Payments Innovation
Direct Line 781.419.1712 | Fax 781.419.1701