Why We Should Change the Conversation Around “AI”

 In Data Intelligence

The confusion around artificial intelligence (AI)

AI is not a technology.

But that hasn’t stopped people from talking about it like a technology. It seems every software vendor claims their tools are powered by AI, and businesses report they’re investing in AI. McKinsey published The State of AI in 2020, based on a survey they conducted on trends in how businesses were using AI.

How is all this possible if AI isn’t a technology?

“AI” in fact refers to a multitude of different solutions, rather than to a specific technology.

“AI” just means “(traditionally) hard to do with computers”

As a term, “AI” has been around since the 1950s, when it was used primarily in the context of machine translation. Since then, “AI” has been applied to problems as diverse as image recognition, optical character recognition, speech recognition, computer vision, chatbots, threat detection, insurance underwriting, loan underwriting, self-driving cars, and robotics.

What do all these diverse use cases have in common? They’re all about automating tasks that have traditionally required human intelligence to perform. Human intelligence encompasses, of course, a host of capabilities—some easy for machines, such as adding numbers, and some very difficult, such as folding laundry.

The fact that a task traditionally required human intelligence to perform has less to do with the nature of human intelligence than it does with the traditional capabilities of our machines. Something is hard for machines to do because we haven’t found good ways to make machines do that—not because there is necessarily anything uniquely “intelligent” about the task. We would describe the ability to add numbers, for example, as a manifestation of human intelligence, but wouldn’t call it “AI.”

Because AI always refers to “the hard stuff for computers to do,” it’s cutting-edge by definition. Tasks that used to be referred to as AI, such as optical character recognition and speech-to-text transcription, start to lose the AI moniker after they’ve been around long enough to be seen as commonly performed by machines. Researchers refer to this phenomenon as the “AI effect”.

Focus on use cases, not on “AI”

The next time someone approaches you about “AI,” ask the following questions:

  • What exactly is meant by “AI” in this context?
  • What is the particular use case?
  • What is the business value of the use case?

Because there are so many different use cases, saying that you’re “investing in AI” is about as helpful as saying you’re “investing in software.” The operative question is AI for what? Software for what? What does it do? AI, like software, encompasses such a variety of use cases that the term lacks explanatory power for differentiating between solutions.

There are a lot of great technologies and solutions out there using “AI,” but it’s misleading to suggest that “AI” gives us a useful way to talk about them.

We need to change the conversation. Whether we call these solutions AI or not doesn’t matter. Business leaders should keep in mind that AI isn’t one specific thing, it’s really hundreds or thousands of individual solutions.

The confusion around “AI” brings to the forefront the need for business leaders to plug their organizations into an innovation ecosystem that links new and emerging technologies to the specific use cases that will provide business value.

That’s what we do at Innovation Labs.

Contact us for more information about our Innovation Labs capabilities.