Posted by Bethany Frank on 14 Sep 2016
Humans have long been fascinated by the concept of artificial intelligence (AI). As the technology becomes better and more accessible, financial institutions and others are beginning to utilize AI and machine learning to improve risk management, operational efficiencies, and productivity.
AI’s potential was originally limited due to the high cost and relatively low computing power of technology available at the time. Research in the field was largely confined to the exclusive labs of elite universities until computers became cheap enough for others to explore AI for practical use.
Classic AI eventually emerged, first designed to solve simple problems through a specific knowledge base. These kinds of applications can perform tasks based on a given set of rules fed into the system. While its potential for innovation is limited compared to today's technology, classic AI has become common throughout everyday life – predictive text applications, virtual assistants like Amazon’s Echo, and programmable thermostats are examples.
From an academic perspective, there still isn’t anything on the market that can be considered true artificial intelligence, despite all the strides made in the field. At its purest, AI should be human-like in its ability to think independently and interact with its external environment. (Think Skynet, which doesn’t exist...yet.)
In contrast, what we see today are applications of machine learning, or the process of self-learning from data without human intervention. Spam detection, for example, is an example of basic machine learning — the system improves its ability to recognize junk email over time.
Today’s developers are now attempting to build systems that mimic the human brain. This subset of machine learning, known as deep learning, equips a system to think for itself by design. Deep learning significantly broadens the potential applications of the technology as a whole by eliminating the need for human instruction and bringing it closer to true artificial intelligence.
Many consider machine learning and AI to be one and the same, although that isn’t entirely accurate. Machine learning is more like a subset or the next phase of artificial intelligence, and it could possibly help enhance AI over time. Blogger Humphrey Sheil provides a good model for this:
While its full potential is still unknown, the financial services world is already embracing various forms of machine learning for payments, fraud and risk, and customer engagement. Be sure to check out part two of this series as we explore the various use cases, and visit our Resource Center for more on payments innovation.
17 May 2017 Blog Mobile Payments: The Ball Is in the Merchant's Court Mobile payments adoption could increase in 2017 as merchant acceptance increases and consumers become more familiar with the technology.