Does AI Know the Similarity Between Cars and Planes? The Birth of AI That Sees the World Like Humans

Imagine for a moment. You have a state-of-the-art Artificial Intelligence (AI) in front of you. This AI is a true “car expert” that can identify the brand and model of hundreds of cars in existence in just one second. However, when you ask this smart AI, “What is the similarity between a car and a plane?”, it either fails to answer or says something completely nonsensical. To us humans, the common sense that “both are large means of transportation made of metal” is a given, but for this AI, it is the most difficult problem in the world. Teaching AI to see the world more like we do - deepmind.google

This is the massive wall that today’s AI faces, the so-called ‘Perception Gap.’ Teaching AI to See the World Through Human Eyes: Bridging the … Simply put, AI is like a memory genius that has memorized tens of thousands of books cover to cover, but has no idea how the content in those books relates to our lives. On the surface, it seems much smarter than humans, but its way of looking at the world is so different from ours that it occasionally makes absurd mistakes. Recently, however, scientists around the world have begun teaching AI “human eyes” and “human common sense” to bridge this gap.

Why It Matters

You might ask, “If AI can identify car models well, why is it so important for it to know they are similar to planes?” However, this issue goes beyond simply answering a quiz; it is directly linked to the safety of the AI we use every day.

Let’s use an analogy. Suppose a self-driving car encounters an ‘empty cardboard box’ on the road. A human driver makes a common-sense judgment: “That’s light paper, so it’s safe to just drive over it,” or “I don’t know what’s inside, so let’s avoid it.” But if AI only recognizes this as a ‘rectangular data pattern,’ it might fail to distinguish whether a box of a form it hasn’t seen before is a rock or paper, risking an accident.

Therefore, the work of ‘Aligning’ AI with human knowledge systems is the key to making AI Robust—unshakable in any situation—and allowing it to Generalize effectively to new, untaught situations. Teaching AI to See the World More Like Humans Do — Google DeepMind

Understanding Easily: Three Ways to Teach AI ‘Common Sense’

Scientists are using three innovative strategies to transplant human visual systems into AI.

1. Running ‘Simulations’ in the Mind: World Models

When you wake up in the morning, you can vividly imagine where the bathroom is or what the hallway will look like when you open the front door, even with your eyes closed. This is because there is a ‘map’ or ‘operating principle’ of how the world works inside our heads.

Giving AI this kind of imagination is exactly what ‘World Models’ do. World Models: Teaching AI to Think Like Humans - LinkedIn This involves building an internal system where the AI doesn’t just store its surroundings like taking a photo, but predicts how the environment will change. World Models: Teaching AI to Think Like Humans - LinkedIn It essentially gains the ability to run a simulation in its head, thinking, “If I push this cup, it will fall to the floor and break, right?”

2. Replicating the Brain’s Filter: Lp-Convolution

Our brains have very efficient filters that pick out only the important parts from a vast amount of visual information. Recently, a joint research team from Yonsei University, the Institute for Basic Science (IBS), and the Max Planck Institute in Germany introduced a technology called ‘Lp-Convolution’ that helps computers process images more similarly to the human brain. AI Horizons: Teaching computers to view the world like humans do

To use an analogy, it’s like putting ‘special glasses’ on the AI that the human eye and brain use when looking at the world. With these glasses, the AI can process information by prioritizing the outlines or three-dimensionality of objects that humans consider important, enabling much more natural recognition.

3. Perception Learned Through Games: Brown University Research

Researchers at Brown University in the US are educating AI in a very interesting way: teaching it to perceive like a human through ‘games.’ Researchers are teaching AI to see more like humans - MSN Just as a young child learns the laws of physics by playing with blocks, AI builds visual logic similar to humans by touching and moving various objects within a virtual game world. Training AI to see more like humans - National Science Foundation

Where We Stand

At this very moment, Google DeepMind is accelerating its research, publishing in-depth study results in the international journal ‘Nature’ that analyze the differences in how AI and humans organize visual information. Teaching AI to See the World More Like Humans Do — Google DeepMind

However, to be honest, there is still a long way to go. Current AI is brilliant at recognizing objects individually, but it often misses the ‘invisible relationships between objects’ that humans grasp very naturally. Teaching AI to See the World Through Human Eyes: Bridging the … The reason we occasionally feel “something is off” when reading text written by AI is that the patterns AI creates are still far from the natural common-sense systems of humans. AIDetector - AdvancedAIChecker for ChatGPT, GPT-5 & Gemini

What’s Next

What kind of future will unfold if AI truly sees the world like a human?

Experts predict that by around 2050, ‘AI teachers’ or robots could emerge with the ability to manipulate matter at the atomic level and see objects perfectly even in the dark. Technology in 2050 - experts give their predictions Beyond being machines that simply pour out knowledge, it might become possible for AI to act as true ‘mentors’ that understand, empathize with, and teach students from their level.

MindTickleBytes’ AI Reporter’s Perspective

Until now, we have been obsessed only with ‘quantity’—how much data AI processes. However, these studies remind us that ‘how one sees’ is more important than ‘how much one knows.’ AI that learns human visual patterns is evolving beyond mere performance improvements into a ‘safe companion’ that shares human values and common sense. That small ability to know the commonality between a car and a plane might just lead us to a safer and warmer technological future.

References

1. Teaching AI to see the world more like we do - deepmind.google
2. Training AI to see more like humans - National Science Foundation
3. Teaching AI to See the World More Like Humans Do — Google DeepMind
4. Researchers are teaching AI to see more like humans - MSN
5. AI Horizons: Teaching computers to view the world like humans do
6. Teaching AI to See the World Through Human Eyes: Bridging the …
7. World Models: Teaching AI to Think Like Humans - LinkedIn
8. Technology in 2050 - experts give their predictions

9. [DataAnnotation

   Future-Proof Your Career WithAITraining Work](https://www.dataannotation.tech/)

10. AIDetector - AdvancedAIChecker for ChatGPT, GPT-5 & Gemini

11. [World models: 10 Things That Matter in AI Right Now

    MIT Technology Review](https://www.technologyreview.com/2026/04/21/1135650/world-models-ai-artificial-intelligence/)

FACT-CHECK SUMMARY

• Claims checked: 14
• Claims verified: 13
• Verdict: PASS