The AI Effect: How We Keep Moving the Goalposts on Artificial Intelligence

We have all been there. You type a prompt into an advanced language model, and seconds later, it generates a human-quality poem, diagnoses a complex medical symptom, or writes fully functional code. For a fleeting second, you are stunned. Is this machine actually conscious?

Then, the feeling fades. You brush it off with a casual, “Well, it’s just statistics and pattern matching; it doesn’t really know what it’s doing.”

Congratulations. You have just experienced the AI Effect.

What Is the AI Effect?

Coined by computer science researcher John McCarthy, the AI Effect is a psychological and linguistic phenomenon defined by a simple, stubborn rule: Once artificial intelligence successfully solves a problem, we stop calling it “real AI.”

Instead, we demote it. We recategorize it as mere “automation,” “advanced statistics,” or “sophisticated programming.”

Think about it historically:

• Optical Character Recognition (OCR): In the 1960s, teaching a computer to read handwritten postal codes was considered cutting-edge artificial intelligence. Today, it is built into the camera app of your smartphone, and nobody considers their photo scanner a sentient being.
• Chess Engines: When IBM’s Deep Blue defeated world chess champion Garry Kasparov in 1997, it was hailed as a monumental milestone for machine intelligence. Today, you can download a chess app on your phone that would easily crush Deep Blue, and you view it as a basic utility, not a marvel of synthetic mind.
• GPS Navigation and Recommendation Algorithms: Decades ago, pathfinding and predictive filtering were frontiers of AI research. Now, they are just boring background infrastructure.

The Moving Goalposts of “Intelligence”

The core driver of the AI Effect is our tendency to define intelligence by what we cannot yet explain or replicate.

As long as a task remains exclusively in the domain of human biology—such as creative writing, strategic reasoning, or artistic composition—we view it as a hallmark of true intelligence. But the moment a silicon chip learns to execute that task efficiently, our definition shifts. We redefine “real intelligence” to exclude whatever the machine just mastered, moving the goalposts further down the field.