Reality Check: The Surprising Relationship Between AI and Human Perception

TL;DR – What You Need to Know in 30 Seconds

• Our brains don’t passively receive reality; they actively predict what’s coming next and then perceive that prediction as “reality.”
• This predictive ability is a survival shortcut—faster decisions in uncertain environments mean less time fretting and fewer calories burnt.
• AI works in a similar way—ChatGPT and other generative models also “guess the next word,” converging on a form of predictive processing.
• Expectations shape perception—we can literally “see” different worlds depending on context, prior experiences, and biases.
• Misalignment of predictions explains why others with opposing viewpoints might seem to inhabit an entirely different reality.

Predictive Processing for Both AI and Your Brain

Today, I’m going to take you on a mind-bending little journey that might just change how you look at… well, everything. If you’ve ever felt that your eyes play tricks on you—or that your political sparring partner simply can’t see the “real facts”—prepare for a major twist. As it turns out, the world you think you’re observing might simply be the one your brain is predicting. And it’s not just us humans at it—AI is doing something remarkably similar. Intrigued? Let’s dive straight in.

The Shocking Realisation: You See What You Predict

Most of us assume our senses faithfully relay what’s happening in the outside world. But mounting evidence from cognitive neuroscience says otherwise: we actually see (and hear, smell, taste, etc.) what our brains think we should see.

Case in point: a quirky optical illusion where the same character can look like a “3” in one context and a “B” in another. Our high-level brain centres (like the visual cortex) inform our lower-level sensory processes what to expect and effectively shape how we perceive the very same lines and curves, depending on the overall context. This fascinating concept is linked to neuroscientist Karl Friston’s predictive coding model. Essentially, your top-down predictions fill in the details of what you “see,” often before the data from your eyes arrives.

A World Jumping (or Not)

Try this quick experiment:

1. Close your left eye.
2. Gently push your right eyeball leftwards with your fingertip.
3. Watch as the entire world seems to slide sideways!

Why? Because your brain didn’t command your eye muscles to move, so it didn’t “cancel out” that movement of the visual field. The upshot? If the brain isn’t expecting your retina to shift, it perceives the world as shifting instead—even though the only thing that moved was your eyeball.

The Milk-Orange Juice Shock

Even if illusions and eyeball nudges aren’t your thing, you may have had this experience: picking up a drink, expecting milk, and instead sipping orange juice. The result is so jarring that it can taste almost foul—much weirder than it would have if you’d expected OJ in the first place. Once again, mismatched prediction plus reality leads to confusion.

Why We’re Wired for Prediction

Predictive processing isn’t just a random quirk of the brain; it’s a strategy for survival. Our ancestors needed to make rapid decisions in uncertain conditions—like whether that rustling bush was the wind or a lurking predator—and do so using limited mental and physical energy.

• Speed: By predicting what’s likely to happen next, the brain can respond almost instantly.
• Efficiency: The brain is an energy hog (using around 20% of the body’s energy). Predictive shortcuts help reduce this cost.
• Adaptability: When the brain notices it’s made an error (e.g., that “leaf” is actually a camouflaged butterfly), it updates its predictions for next time.

AI as a Prediction Engine

Surprisingly (or perhaps not, given the same constraints of speed and efficiency), generative AI such as ChatGPT uses a very similar approach. It doesn’t “understand” the way humans do—it’s simply predicting the next word based on the context of all the words that came before it.

• Pretrained Transformer (PT): The “PT” in ChatGPT refers to the method it uses to learn from vast volumes of text, forging associations rather like our brains form connections between experiences.
• Convergent Evolution: Think of the marsupial wolf versus the canid wolf—two unrelated lineages that evolved to fill similar niches. Biological and silicon-based “brains” also converge on similar problem-solving strategies due to common pressures like speed, efficiency, and limited resources.

The Pitfalls of Predictive Perception

While our (and AI’s) predictive brains are brilliant for speed and survival, they can also lead us astray. We all know about cognitive biases, but these illusions can be even more potent when dealing with people who see the world through a different lens. If someone’s internal model of the world is shaped by divergent experiences, it’s almost as though they’re seeing a completely different reality.

• Political Polarisation: “They must be ignoring the facts!” Actually, their predictive model might be so different that they’re literally perceiving different facts.
• AI Hallucinations: Generative AI sometimes produces wildly incorrect statements—so-called “hallucinations”—because it’s only as accurate as its ability to predict and synthesise words. When it “misfires,” it can sound confident despite being wrong.

Final Thoughts

We might prefer to believe our eyes, but the evidence is overwhelming: both humans and AIs perceive a version of reality shaped by predictions. This realisation can be unsettling, but it can also be liberating, especially when it comes to understanding why others sometimes arrive at very different conclusions.

The next time you find yourself shaking your head in disbelief at an AI-generated error or a friend’s seemingly bonkers view on world affairs, remember: we’re all just brains (silicon or organic) navigating the murky waters of an unpredictable world with the best guesses we’ve got. The future belongs to those who can update their predictions when the evidence calls for it—and hopefully, that includes both you and me.

What Do YOU Think?

If we don’t see reality for what it truly is, does that mean we can never really trust our own eyes? Or is our brain’s tendency to shortcut reality the very reason we survive—and thrive—in a chaotic world? The debate continues!

Let’s Talk AI!

How are you preparing for the AI-driven future? What questions are you training yourself to ask? Drop your thoughts in the comments, share this with your network, and subscribe for more deep dives into AI’s impact on work, life, and everything in between.

You may also like:

Time For A Reality Check On Our AI Expectations

Bot Bans? India’s Bold Move Against ChatGPT and DeepSeek

The Mind-Blowing Future of AI: Ray Kurzweil’s Predictions for 2050

• For a deeper look at how our brains can misinterpret the world around us, check out the Illusions Index; it’s packed with fascinating examples of how easily our perception can be fooled.