Moving Goalposts: Understanding the “AI Effect”

Have you ever wondered why breakthroughs that once felt like science fiction—such as recognizing human faces, translating languages on the fly, or generating photorealistic images from simple text prompts—suddenly feel utterly ordinary?

As soon as a machine successfully performs a task that previously required human intelligence, society doesn’t celebrate it as “true AI” for long. Instead, we shrug and say, “That’s just standard algorithms; that’s not real thinking.”

This psychological phenomenon is known as the AI Effect: the tendency for people to discount the intelligence of artificial intelligence as soon as it works. Coined by computing pioneer John McCarthy, the concept highlights a profound paradox: AI is defined not by what it can do, but by what it hasn’t yet mastered.

The Shrinking Definition of “Intelligence”

The history of computer science is a long chronicle of shifting definitions. In the mid-20th century, computing pioneers believed that if a machine could play chess, solve complex mathematical proofs, or process structured data, it would undeniably possess machine intelligence.

Yet, history proved them wrong time and again:

• Chess and Strategy: When IBM’s Deep Blue defeated world chess champion Garry Kasparov in 1997, critics argued it was merely brute-force calculation rather than genuine thought. Decades later, when AlphaGo mastered the profoundly intuitive game of Go, skepticism persisted.
• Optical Character Recognition (OCR): Reading handwritten postal codes or scanning printed text was once considered a frontier of cognitive automation. Today, OCR is embedded invisibly into mobile apps to scan receipts, stripping away its status as “AI.”
• Speech and Translation: Real-time speech-to-text transcription and instant multilingual translation once required human interpreters. Now, they are taken for granted as basic utility features on smartphones.

Once a technology becomes reliable, affordable, and woven into the fabric of daily life, we strip it of the “AI” label and reclassify it as routine software.

Why Do We Move the Goalposts?

The AI Effect is driven by several psychological and linguistic factors:

“True intelligence is always whatever machines cannot do yet.” — Larry Tesler’s Theorem (adapted for AI)

1. The Familiarity Bias: When a technology is novel, magical, and mysterious, we attribute high cognitive agency to it. Once we understand its mechanics or use it daily, the mystery vanishes, and with it, our perception of its intelligence.
2. The Demystification of Labor: Tasks that humans find intellectually demanding—such as calculating huge sums or sorting through millions of data points—are things computers do naturally well. Conversely, tasks that feel effortless to humans—like walking, recognizing a familiar face, or understanding subtle sarcasm—require immense computational complexity. When machines master human intuition, we devalue the complexity because humans do it effortlessly.
3. Marketing and Hype Cycles: The tech industry itself contributes to the effect. By overusing terms like “next-gen intelligence” for minor software updates, companies desensitize the public, forcing the definition of true AI to continually escalate.

The Modern Paradox

In today’s landscape of generative models and foundation systems, the AI Effect is more visible than ever. We interact with systems that can write code, draft legal briefs, and compose symphonies. Yet, the moment a model hallucinates a fact or stumbles on basic common sense, critics rush to declare that “it doesn’t actually understand anything.”

This constant reclassification creates a moving target for researchers and ethicists alike. It forces us to ask a deeper question: Are we chasing an illusion of human uniqueness, or are we simply unwilling to acknowledge machine cognition on its own terms?

Ultimately, the AI Effect teaches us as much about human psychology as it does about computer science. We protect our definition of intelligence as a uniquely human sanctuary, updating the rules of admission every time technology comes knocking at the door.