The AI Effect: Why Artificial Intelligence Never Quite Gets to Keep Its Mind

Imagine standing at the edge of a horizon, chasing a brilliant sunset. No matter how fast you run, the horizon retreats at the exact same pace, maintaining its distance.

In the world of technology, human ambition has its own version of this optical illusion. It is known as the AI effect—a peculiar psychological and cultural phenomenon where, the moment artificial intelligence successfully cracks a capability long thought to require human thought, society stops viewing it as intelligence at all. Instead, it is swiftly reclassified as routine calculation, mundane automation, or “just math.”

As computer scientist Larry Tesler famously quipped, capturing the essence of the paradox: “AI is whatever hasn’t been done yet.”

Moving the Goalposts of Mind

The AI effect acts as a perpetual shifting of the goalposts. Historically, almost every major milestone in computer science followed a predictable script:

1. The Aspirational Stage: Experts and science fiction writers declare that a certain complex task—such as playing grandmaster-level chess, translating spoken languages in real time, or recognizing human faces—is the ultimate benchmark of true cognition. It is widely agreed that once a machine can accomplish this, it will possess genuine intelligence.
2. The Breakthrough: Engineers and researchers spend decades building neural networks, training models, and refining algorithms until the machine finally achieves or surpasses human capability.
3. The Dismissal: The system succeeds. Yet, rather than marveling at machine intellect, humanity shrugs. Critics point out that the program is merely executing billions of matrix multiplications, using brute-force search trees, or relying on statistical probabilities. “That’s not real thinking,” we conclude. “It’s just an algorithm.”

Consider the milestones of yesterday. In 1997, when IBM’s Deep Blue defeated world chess champion Garry Kasparov, it made global headlines as a watershed moment for machine cognition. Yet, almost immediately, commentators downplayed the victory, arguing that Deep Blue wasn’t actually “smart”—it was just calculating millions of positions per second.

Decades later, optical character recognition (OCR), speech-to-text transcription, and GPS route optimization are deeply woven into our everyday lives. We use them effortlessly to scan receipts, dictate messages, or navigate unfamiliar city streets. Because they work reliably and sit quietly in our pockets, we no longer label them as AI. They have been demoted to mere software.

Why Do We Keep Moving the Target?

This shifting baseline is not just a quirk of human stubbornness; it is rooted in how we perceive our own minds.

• The Anthropocentric Bias: Humans tend to equate “intelligence” with subjective experience, consciousness, and biological emotion. Because a computer does not feel tired, experience joy, or possess an autobiographical soul while calculating, we struggle to attribute mindfulness to its outputs. We define true intelligence by how we think (messily, intuitively, and consciously), rather than what can be achieved.
• The Domesticating of Magic: Novelty breeds awe, but familiarity breeds contempt. Once a technology becomes commercialized, transparent, and ubiquitous, its mystical veil drops. A technology that requires a user manual feels advanced; a technology embedded seamlessly into a smartphone app feels invisible.

This creates a strange catch-22 for researchers: an AI milestone is either unproven hype or, the moment it actually works, stripped of its “AI” status and rebranded as standard engineering.

The AI Effect in the Age of Generative Models

Today, the AI effect plays out on an unprecedented scale with modern generative artificial intelligence. Tasks that once required years of specialized artistic training, professional writing, or software coding—such as generating photorealistic images, writing functional code snippets, or drafting coherent prose—can now be executed in seconds by large language and diffusion models.

Yet, true to form, the goalposts have shifted upward once again. Critics and users alike point to hallucinations, lack of genuine agency, or absence of true emotional understanding as proof that contemporary systems are “just stochastic parrots” spitting out predictable word patterns.

While these technical critiques carry weight regarding how these models function under the hood, they also reflect the classic symptoms of the AI effect: minimizing monumental functional capabilities because they do not map onto human-like sentience.

Embracing the Illusion

The AI effect reveals less about the limitations of machines and much more about the ego of humanity. We have a historic habit of defining our uniqueness by whatever skill we have most recently taught computers to copy.

Perhaps the solution is to change how we measure progress. Instead of asking whether a machine possesses a soul or mirrors human self-awareness, we might look at intelligence through a functional lens: What problems can be solved, and how does this technology expand human potential?

Until then, the horizon will keep moving—and whatever incredible breakthrough arrives tomorrow will, by next week, simply be called “normal software.”