AI, Explained Simply
Everyone's used a chatbot by now, so let's skip the "what even is AI" bit. What actually gets skipped over is what's running underneath these things, and why the big three all feel a bit different to use. Here's the quick version.
Wait, what's an LLM?
Every chatbot you've used - ChatGPT, Gemini, Claude - is powered by something called an LLM: a Large Language Model. Worth knowing what that actually means, because it explains both what these things are brilliant at, and where they still trip up.
An LLM has been trained on a genuinely massive amount of text - books, websites, code, conversations - and it's learned the patterns in how words and ideas fit together. When you ask it something, it's not pulling a stored answer out of a database. It's predicting, piece by piece, what should come next, based on everything it's seen. Do that well enough and you get answers that read as properly thought-out, even though nothing in there is "looking anything up."
That's also why they can sound completely confident while being flat-out wrong sometimes - they're not fact-checking, they're pattern-matching. Worth keeping in mind before you trust anything important it tells you at face value.
You can run your own LLM at home too
Here's the bit most people don't realise: you're not locked into using a company's chatbot in the cloud. You can download and run an LLM directly on your own PC - free, and completely offline once it's downloaded. Tools like Ollama make it genuinely a one-command install: grab it, pick a model, and you've got your own private chatbot running locally, nothing leaving your machine.
The catch is hardware. A small model can squeeze into as little as 4GB of RAM - fine for basic stuff on almost any PC. The rule of thumb from there is simple: more RAM (and ideally a decent graphics card) means a bigger, smarter model you can run. Roughly:
- 4-8GB RAM - small, basic models. Enough to have a play and see how it works
- 16GB RAM - solid mid-size models, genuinely useful quality for everyday questions
- 32GB+ RAM, or a decent GPU - bigger, sharper models start becoming realistic to run
Doesn't need to be perfect on day one - even a small model on a spare laptop is a fun way to see how this stuff works under the hood, and it slots in nicely if you've already got a home lab going.
My simple 1, 2, 3 of the big names
- 1. ChatGPT - the one most people have heard of. General-purpose chat, good all-rounder for everyday questions
- 2. Gemini (Google) - built into Google's own stuff, and my pick if you want images generated
- 3. Claude - my go-to whenever actual code needs writing. Coding king, in my book - it's what builds this blog
The takeaway
Same tech under the hood, different strengths on top. Pick based on what you're actually trying to do - chat, images, or code - rather than assuming one does everything best.
Using any of the big three already? Which one's your go-to - drop it in the comments.