Is Your Technology 'True' AI Or A Flock Of Pigeons? It Doesn't Matter Posted on : Dec 11 - 2018

There’s no hype like the hype around artificial intelligence (AI). It's been touted as the next big thing for years and is frequently glorified as a solution for our most challenging problems — one that will change the very fabric of our society.

It's little wonder, then, that companies and investors are throwing their money at AI and anything else that falls under the umbrella of “machines working smarter than we can.” TechRepublic reported that 61% of businesses employed AI in 2017, according to a Narrative Science study. There is solid interest from VCs (according to KPMG, global AI investment nearly doubled in 2017), and the IDC Spending Guide forecasted that AI spending is on track to reach $57.6 billion by 2021.

But all of that money has also come with increased scrutiny. Companies and their solutions are being assessed on their AI-ness. And often, they may be falling short. As I explained in a previous article, many “AI” solutions are simply machine learning (ML) technologies. Some are just tech that’s really good at data analysis or statistics. And, somewhat confusingly, The Guardian reported that others still are humans masquerading as computers.

Shocking, isn’t it? But look: The thing is, whether your solution is actually AI or just a variation on the theme doesn’t matter in the least.

If you’re implementing a solution just because it’s AI, or turning away a solution because it’s not, then I believe you’re doing it all wrong. The exact name and nature of the process doesn’t matter. It’s the outcome that does.

So don’t waste resources fretting about whether your AI solution is truly AI. It doesn’t matter whether you’re working with a basic machine learning model or nothing more than a flock of well-trained pigeons. What does matter is that your solution behaves the way you need it to.

Here’s what I believe you should be asking when vetting a machine-driven solution to a business problem, based on my experience in the text analytics space:

1. Can it adapt?

If your solution is a fixed set of rules, you'll probably need to change your solution each time your needs change. If you’re using AI, you can rely on it to adapt its models as it receives new data. This isn’t to say that non-AI solutions aren’t adaptable or that all machine-learning-based solutions are adaptable. Consider a model built to predict whether a customer will churn or not. If you subsequently launch a new, lower-cost service, that model must be retrained, as the likelihood of churn is directly related to your product offering. How much adaptability matters to you will depend on your business case. View More