1 of 6

Deep Dive into AI Perception

​

​

Presented by: Keria Varro�Fundaments Of Computing�19th November 2025

2 of 6

How AI Sees & Hears Us👁️ �Seeing the World (Computer Vision)�

Imagine teaching a machine to see numbers on a puzzle:

Old School: The machine uses template matching, basically comparing the picture to exact, pre-saved copies.

Problem: Fails if the number is bigger, tilted, or handwritten. It's too picky!

3 of 6

�����������New School: The machine uses feature extraction and image processing (like edge enhancement and smoothing) to identify shapes (a vertical line is a '1’).���Problem: Still struggles with messy real-world photos where things overlap. Humans still win at recognizing complex scenes!����

4 of 6

💬 Talking the Talk (Natural Language Processing)�

  • Turns out, human language is super messy compared to simple computer code. We rely on context and hidden meanings.

​

  • To understand us, AIs need three levels of analysis:

​

  • Syntactic: What's the grammar? (Who is the subject?)

​

  • Semantic: What do the words mean conceptually? (Who did the action?)

This Photo by Unknown Author is licensed under CC BY-SA-NC

5 of 6

����Contextual: What's the vibe? (Does "bat" mean baseball gear or a flying animal?)��The goal is to get AIs to understand us as easily as we understand each other�—even when we're sarcastic or use slang!���🤔The Big Question: ��Is an AI just acting smart (Weak AI) or is it smart and conscious (Strong AI)? That debate is still ongoing!���

This Photo by Unknown Author is licensed under CC BY-SA

6 of 6

THANK YOU FOR LISTENING �GOODBYE