When people talk about AI today, ChatGPT is usually the first thing that comes to mind.
But the more I explore this field, the more I realise that there's a lot happening behind these tools that we rarely talk about.
I've been particularly interested in understanding how modern AI applications actually work, beyond simply writing prompts and getting responses.
A few concepts have really caught my attention.
LLMs: What's happening behind the responses?
Language models do much more than simply repeat information. Understanding tokens, context and next-token prediction gives us a better idea of how these systems generate responses.
Embeddings: Finding meaning, not just matching words
Imagine searching for "learning programming" and getting an article about "becoming a better developer."
The words aren't identical, but the meaning is related. Embeddings help AI systems capture these relationships.
RAG: Making AI work with external information
One thing I find interesting about RAG is how it allows language models to work with information from external documents and knowledge bases.
It's particularly useful when building AI applications around company-specific or frequently updated information.
AI Agents: Beyond answering questions
This is another area I'm keen to explore.
Instead of just responding to a prompt, AI agents can use tools, work through multiple steps and perform tasks with some level of autonomy.
There's still plenty to learn about how reliable these systems can become.
Why I'm exploring all this
For me, learning AI isn't just about keeping up with new tools or trends. I want to understand the technology behind them and how different pieces fit together.
I've also been writing about these topics and sharing my learnings in my LinkedIn newsletter, The Algorithm Behind AI https://www.linkedin.com/build-relation/newsletter-follow?entityUrn=7447966269116522496.
It's been a great way to explore these ideas more deeply and exchange perspectives with others interested in the field.
I'm curious: Which area of AI are you spending the most time understanding these days?
LLMs, RAG, AI Agents, or something completely different?
