Kount LogoBLog Against Fraud

What Does Artificial Intelligence Actually Mean?

posted on: Wed Sep 27 2017

By, Josh Johnston, Manager of AI Science at Kount

It seems Artificial Intelligence is everywhere today. And by that, I literally mean the words: “Artificial Intelligence”. They’re everywhere. Most every startup with a cluster in the cloud adds .ai to their name and promises data-driven optimizations for your business.

As an AI Scientist, people ask me all the time what this term actually means. Fortunately, it isn’t complicated.

AIArtificial Intelligence is simply a machine exercising something we recognize as human judgment. Familiar examples are digital personal assistants communicating with natural language, automated image captioning, and self-driving cars. Each of these are things that we expect people to do, but don’t think computers are good at. 

This definition of AI means that the field is ever-shrinking. As computers do more and more, we change our estimation of what constitutes AI. Not long ago, playing games like chess and Go was seen as requiring a uniquely human intelligence. Now that computers outplay the top grandmasters at these games, we are more apt to consider them brute force calculations. This paradox, sometimes called the AI Effect, means AI scientists are in many ways destined to never solve the problems we work on.

Alan Turing, the godfather of computer science, proposed a test for comprehensive, or so-called General AI. In its current form, we say that if you can’t tell whether you are talking to a person or a computer, the computer has achieved AI. This can be narrowed from General AI to more specific applications. For example: an AI car is indistinguishable from a human-driven one.

With this understanding, we see that AI is nothing more than automating solutions to hard problems that otherwise require human judgement. I’m often asked if Kount uses AI in our products.  This is kind of a silly question. Identifying fraud at scale means automating and augmenting human judgement. So yes, fighting fraud is an AI problem and we use AI solutions. 

For two of the reasons fraud is a hard problem to automate, check out these recent blog posts:

The term Artificial Intelligence is being used so much today because it is a fancy way of saying “make computers solve hard problems”.  The proof is in the pudding, however, and we’re happy to tell you how we make ours. Discover more in our eBook "Two Heads Are Better Than One: Artificial Intelligence + Human Insight".

Artificial Intelligence + Human Insight