Any look back at analytics in 2017 makes it clear that machine learning and artificial intelligence appear to be the ‘next big things’ that can solve just about any problem, from writing new hit songs to curing disease. Not one to buy into the hype, I became curious as to why these topics have become the new darlings of the analytics world. Perhaps it’s because nowadays everything has analytics, from niche solutions to management consultants and cloud service providers, analytics seem to be available everywhere in everything. What isn’t so obvious, however, is how not new these types of analytics actually are. For all these newcomers to the analytics party, I’m sure it’s new to them, but here at SAS, our founder wrote his first machine learning model in 1976. And believe it or not, in my research for this piece I discovered plenty of references to AI that pre-date computers! Let that sink in. (I had to!)
Here at SAS, we’ve used machine learning in our market-leading enterprise fraud management solution since its inception. Primarily, this is done through a “feedback loop” that takes outcomes of the alert triage process (the fraud detection results) and uses machine learning techniques to improve the analytic model performance. If our analytic model is detecting fraud and raising alerts that are not effective, we would want the system to know and learn from this, which is one way we help to reduce false positives within the environment.
Similarly, we often use machine learning algorithms that are also classified as semi-supervised predictive models, to help detect emerging fraud schemes. Unsupervised analytic models can be a pathway to AI, helping to detect the unknown unknowns -- threats we don’t even know exist yet. By employing a variety of analytic techniques (not just machine learning or AI), we can identify new fraud schemes and networks surface over time as they take shape.
However, for the same reasons that self-driving cars are not dominating our roads, I don’t see AI for the hype the market would have you believe. AI and machine learning can be valuable enablers but ultimately, humans will need to remain in the loop. Check out the recent Wall Street Journal article, Without Humans, Artificial Intelligence is Still Pretty Stupid, which kicks off with an…uh…interesting example.
This need for human checkpoints is especially important within the realm of security intelligence. So long as these techniques can create valuable insights that can be conveyed and understood by the decision maker (which is not a computer, natch, but typically a fraud examiner, analyst, investigator, business manager, attorney or jury) then they can be valuable in fighting fraud. If these new-not-so-new and over-marketed techniques cannot be understood by the decision makers, then perhaps they might be more useful in my Roomba to detect when the dog might be out for a walk, or the kids were playing in the kinetic sand on the carpet, again. See, I still have hope for AI, despite all the hype.