07 July 2017

I’m not going to start talking about HAL9000, about playing and winning at SKYNET in a GO match. Let’s try to be serious, if possible, when it comes to talking about Artificial Intelligence (AI). The first thing that we can say about AI is that most contemporaneous literature about AI is not exactly related to AI. Since we are talking about banking in here, we are going to focus on this basic point, the fact is that I wanted to warn you that I’ll try to do not fall into common topics.

The AI banking application is fortunately not as contaminated as in other sectors. The megalomaniacs computers from the movie industry since to be more interested by a total war with human than by mastering global finance, which shows us that they are not so intelligent after all. Thus, it’s easy to see the use of connectionist techniques, such as machine learning and neural networks, in the fight against money blocking or detection of patterns of behavior in customers without perceiving it as a threat. But all of this is of great interest for bank back office. How can AI improve the user experience?

Maybe conversational systems, or chatbots, will be the most obvious answer: interact with our banking application only by voice. The challenge is to cover up the high expectations of regular customers. When it comes to language, we face an interesting paradox: we use it constantly but we are never truly aware of its complexity.  For instance, it’s very common to use synecdoche when speaking, but very few persons are able to identify them. Let’s just say that we basically speak, but we are not really aware of how we do it. And because we speak in this semi-instinctive way we tend to think that it’s an easy task, and that’s a big mistake. Speaking is extraordinarily complex, that much, that we even understand each other when we do not do it properly o completely. And that’s what we expect from AI, we want it to be able to understand us when we speak improperly or incompletely. It’s like saying that we want it all or we don’t want nothing at all: whether I can talk to you as I want to, without any limitation from a command list, and if it’s not possible I’d better not speak at all and user the classical options menu. When we think about it, it’s totally fair.

Without any doubt, we’ll assist to the raise of the capacity of conversation of the computer of the future, even though talking about specific dates seems to be risky. In the past, predictions went badly. Alan Turing foresaw back in 1950 that there will be machines able to talk to human before 2000., a prediction that was confirmed by IBM around 1970. Let’s just hope that it will come soon, this progress will not only revolutionize baking industry, but all our human user experience.

Click here to read the next article of this serie: Virtual Assistant

Blog Search