Introduction
You never truly understand something until you teach it. That’s why I’m writing this – after taking a DataCamp course on ChatGPT taught by James Chapman, I realized that while we all use AI daily, few of us actually understand how it works. Let me break it down in simple terms.
What Is ChatGPT?
ChatGPT is an AI chatbot developed by OpenAI. Unlike traditional chatbots (which follow rigid scripts), ChatGPT is general-purpose—it can handle almost any topic, understand complex language, and integrate with various apps.
How Does It Work?
Input (Prompt): You type, speak, or upload an image.
Processing (LLM): A Large Language Model (LLM) analyzes your prompt.
Output (Response): The LLM generates an answer based on its training.
Fun fact: There are also Small Language Models (SLMs)—like those by EqualyzAI—designed for specialized tasks (more on this later).
Where Does ChatGPT Get Its Answers?
The LLM is trained on massive datasets (books, articles, code, etc.). An algorithm fine-tunes this data to help it “understand” and generate responses.
Key point: ChatGPT doesn’t “know” anything—it predicts answers based on patterns in its training data.
5 Big Limitations of ChatGPT
Knowledge Cutoff: It can’t access real-time data (hence constant updates like GPT-4o).
Bias: Trained on human-created content, so it inherits cultural, political, and social biases.
Hallucinations: It confidently invents false answers when unsure (like a “drunkard making up stories”).
Context Loss: Struggles with long, multi-topic conversations.
Legal Risks: Who owns the output? You, OpenAI, or the original data sources? Always check usage policies.
The Future: Why We Need Small Language Models (SLMs)
While ChatGPT is a jack-of-all-trades, specialized fields (coding, marketing, African languages like Yoruba and Hausa) need tailored AI. That’s where SLMs shine—efficient, focused, and ideal for low-resource settings.
Want to go deeper? Read EqualyzAI’s whitepaper: “Democratizing Generative AI for Social Good in the Global South.”
Final Thought
AI is powerful, but it’s not magic. Understanding its limits helps us use it smarter—and pushes us toward better solutions, like SLMs.
