Give Me Another Dollop of That AI
5 February 2018 | 10:31 am
You can be forgiven if the way we talk about Artificial Intelligence makes you think you can order it up like a scoop of ice cream. It seems that way because we constantly read that:
- AI solves all our problems
- AI experts cost an arm and a leg
- AI analyzes data than any person can
- AI will make us all unemployed
While each of these statements might turn out to be true (well, we hope the last one is wrong), they all suffer from the same problem. They act as though all AI is the same. That all AI is one monolithic thing that can be added to any system if you just have enough money.
It’s not true.
First off, there are many different kinds of AI applications and they require different techniques. Voice recognition is not the same as text analytics is not the same as optimizing search results. These applications are different from each other and they use different techniques to perform their “magic.” Most of them use multiple AI techniques. And they usually depend on the existence of data.
I have been phoned up by more than one expectant client who wants to solve this problem or that problem with AI. Often, that is perfectly reasonable, but just as often I have to tell them they need to take several steps first. Often, they need to set up a standard process that collects data in a standard way so that the AI techniques have something to work with. Luckily, even taking these initial steps has business value, if you do it right, so the clients are usually easily persuaded to move forward.
Wanting to use AI is not a problem. Forward-looking organizations are always pushing the envelope and AI is just the latest way to do it. But let’s make sure that we are getting the business value we expect and that we are ready to take the preliminary steps to get there. We shouldn’t make AI a problem looking for a solution.