Large Language Models
Large Language Models are the culmination of all that came before them. All the capabilities enhanced with the crazy amount of computing power we have today.
These are models trained on unthinkable amounts of text. They’re trained on the internet. And, the more we use them, the more they learn.
This amount of training makes them really good at guessing what words should come next in a sentence. That’s how they can generate new text that sounds natural.
That’s why they can sometimes feel like you’re talking to a human.
But, as you know, you’re not.
You’re feeding input to a machine that analyzes your words. It pays attention to the important parts of them. It does this to guess what the most probable best response should be.
You’re seeing it predict one word at a time, really fast. Based on what you say, it tries to respond in a way that makes the most sense.
It’s not thinking like a human. It’s using patterns from human language to sound like one.
Even though it’s not human, it’s important to note - it’s trained to respond like one based on human content.
Tons of human generated content. Content from people like you and me.
This is why tools like ChatGPT are so popular. It finally feels like we’re really talking to computers that get us.
And, in a very real way, they do get us. They learn from us. They grow with us.
The more we train AI and improve how it works, the better it’ll get.
If you want a little more technical insight into how they work, check out the LLMs: Behind the Scenes course.
If you want to learn more about how to prompt LLMs, check out these links: