- Move over Big Data. There are two new buzz words in town: ‘artficial intelligence’ and ‘machine learning’. The claims being made for these technologies are just as significant, if not more so than those made for Big Data three to five years ago. So, how widespread is the use of artificial intelligence today? And what are the implications for the payments industry?
The rise of powerful artificial intelligence will either be the best or the worst thing ever to happen to humanity, according to Professor Stephen Hawking. Meanwhile, the chief executives of Microsoft and Google have said that the technology will change not only computing but every industry and business process. Artificial intelligence is trending high in the hype cycle. It seems as though every company has a product claiming to have artificial intelligence built in, but that is not quite how it works.
Beyond the black box
Users and suppliers of machine learning technology will have to grapple with various questions, some of which are not new. One of these is causation. Humans like to make sense of the world through causal links. We like to believe that every effect has a cause and vice versa. So, how do we explain the workings of machines to humans?
There are techniques in artificial intelligence and machine learning that are more prone to be understood by humans, explains Dr. Ulrich Dorndorf from INFORM. For example, with rule-based systems where experts represent their knowledge in the form of rules, system operators can look up which rules led to a conclusion. Conversely, there are techniques that will essentially remain a ‘black box’, at least for the moment.
Neural networks are an example of a ‘black box’ approach. People and scientists recognise that they work quite well in certain areas, but even from a scientific point-of-view the question is still open why this is so. Naturally, it is possible to layer the various technologies. “We find that our customers, as users of the technology that we deliver, prefer approaches where the logic does not remain a completely ‘black box’.
says Dr. Ulrich Dorndorf
Machine learning techniques enable companies to analyze huge amounts of data — and to do so quickly. “To pinpoint fraud in real-time, you have to adapt to constantly new patterns. That’s the main complexity: too much data has to be combined that it is close to impossible for humans to react in the time period needed,” explains Konrad Hochmuth, Strategic Business Developer at INFORM. Whereas in the past a bank may have had two-three days to cross-check a payment, real-time payments are typically on the beneficiary’s account within ten seconds. Real- time payments are driving real-time decisions.
Hochmuth makes an important distinction between humans in the discussion about causation. “The human specialists in the bank are very much interested in understanding why transactions have been declined. This is their number one resource to optimize the system and see future trends that may not even be in the data. These humans are hired o want to know why.” However, the ultimate customer in front of the ATM or POS device does not necessarily want to know why the transaction failed. While declines are annoying, typically customers just take out another payment method and do not ask their bank for an explanation.
A Smarter Future?
Just as machines will get smarter, so will humans. They will understand the technology better and incorporate it more into day-to- day business processes. With regard to the payments industry in particular, as Hochmuth from INFORM maintains,
There is a “huge opportunity to combine conventional decision- making strategies, human expertise and machine learning to improve processes.” Machine learning will create a ripple effect as with all innovation. It will be a technology that enables a new raft of business cases and industries, including some we cannot possibly think of at present.
says Konrad Hochmuth.
Read the full cover article here at Payments Cards and Mobile Magazine (Jan/Feb 2017 Edition).