Understanding Intelligence Through Abstraction: Human and Artificial

Tiwariprakhar
2 min readJun 11, 2024

Introduction

Abstraction is all around us. When we see a cat, we don’t focus on every detail like the shape of its face or the length of its claws. We simply recognize it as a cat. Yet, if asked why it’s a cat and not a dog, we might struggle to explain. Despite this, a child can distinguish between cats and dogs after seeing only a few examples, abstracting the essential features. This ability to simplify complex realities is fundamental to intelligence.

Human Intelligence and Abstraction

In human intelligence, abstraction helps us simplify complexities and focus on broader concepts. For instance, novels are abstract representations of ideas, while mathematics and language are powerful tools for understanding and communicating. Some argue that mathematics is more real than the physical world, allowing us to theorize about unseen concepts. In art, Picasso revolutionized perception with cubism, an abstract style that changed our understanding of reality.

Artificial Intelligence and Abstraction

AI deepens our understanding of abstraction. Since AlexNet’s success in the 2012 ImageNet competition, neural networks have excelled at image classification by mimicking human visual processing. Initially, these networks focus on low-level features like edges. As the layers deepen, they abstract this information into high-level features, such as shapes and objects, much like the human visual system. This emergent property of abstraction is not explicitly programmed but results from training.

Large Language Models (LLMs) like GPT-4 work with text, an abstract representation of the world. Using self-attention mechanisms, they abstract and make sense of vast amounts of data, reasoning about the world even without direct sensory experience. This suggests that AI can understand our world through abstraction, similar to humans.

Future Directions

The next step in AI development might be familiarizing models with the actual world through extensive video data and self-supervised learning, similar to how LLMs are trained on text. This could enable AI to understand our world more deeply. Current efforts, like diffusion models that associate language with objects, show promise.

Conclusion

Abstraction is a critical component of both human and artificial intelligence. While this blog focuses on abstraction, there are many more facets of intelligence to explore. I look forward to delving deeper into these aspects in future posts. Please share your thoughts and critiques; I would highly appreciate your feedback.

Sign up to discover human stories that deepen your understanding of the world.

Free

Distraction-free reading. No ads.

Organize your knowledge with lists and highlights.

Tell your story. Find your audience.

Membership

Read member-only stories

Support writers you read most

Earn money for your writing

Listen to audio narrations

Read offline with the Medium app

No responses yet

What are your thoughts?