The more I read about artificial intelligence (AI), the less clear it becomes what people mean when they talk about AI. And while semantic arguments can be a waste of everyone’s time, a loose definition of Al presents an opportunity for vendors to package software into something that it isn’t. Blockchain and AI vie for position as the most overhyped term in technology today. But blockchain at least benefits from a common understanding of the underlying technology, even if its potential uses are severely overstated. Any buyer of AI-powered software should beware of strangers bearing gifts: always make sure that AI products are capable of what you want them to do, not what the vendor claims.
AI Beats Expectations to Be Simultaneously Everything and Nothing
Everyone has a definition of AI. Your perspective, knowledge of analytics, and desperation to sell some technology have a great bearing on what constitutes AI. I have seen definitions that range from “any code that includes an IF statement” to “nothing currently in existence.”
Technology marketers can never be accused of reticence when it comes to latching on to the latest industry buzzword. AI’s loose definition means it can be applied to virtually any piece of technology. An IF statement in a computer code means that a computer will make a decision based on some form of data input. At a very basic level, this is an intelligent decision.
Many disagree and claim the boundary for what constitutes AI lies in more sophisticated analytics processes. Essentially, people look for examples of a computer mimicking human thought. The big question is: How close to human thought does a computer have to get before it is considered capable of “thought?” Indeed, each time a new development happens in the field of advanced analytics, such as AlphaGo beating a human at the notoriously complex game of Go, someone will always say that it’s not true AI. This trend is summed up in Larry Tesler’s theorem that AI is “whatever hasn’t been done yet.”
The Paradoxical Definition of AI
I have no idea what AI is supposed to be, and I believe that this uncertainty stems from the blurred boundaries of its definition. The sorites paradox explains this well: Starting with one grain of wheat, how many more grains must be added before one has a heap? Similarly, there is no clear answer to the questions around what constitutes intelligence, when AI takes over from basic tech-based processing, or how close to human thought tech processing must get before it can be deemed even artificially intelligent. Some may propose specific demarcation; others immediately disagree.
What Matters Most Is That None of This Matters
We often start reports on technology with a definition to help the reader better understand the boundaries of what we write about. I have spent far too long trying to come up with a definition of AI with which I am comfortable. I can’t come up with a sensible boundary for what does and doesn’t constitute AI, and even if I did, more people will disagree than agree. But who cares? AI really means the latest and shiniest analytics product to hit the market.
What buyers should bear in mind is whether this technology does the job for which it was intended at a competitive price. Just make sure that anything painted in AI colors is going to make a decent ROI for your business.