If you grew up in Germany you know immediately who to look for when you need to easily understand how something works: “The Show with the Mouse”
The term “LLM” is often used when talking about AI. But what does it mean?
LLM stands for “Large Language Model”. You can imagine it as being the “brain” of an AI application: it processes information, recognizes patterns, and generates responses based on this.
However, an LLM can neither think nor understand. It can only perform calculations. But it does this very well and with staggering speed. And it is very good at recognizing patterns and then building something itself that uses the appropriate patterns.
To us humans, the result can seem like communication which is meaningful, and even empathetic. In reality, behind it all is a gigantic probability calculation that figures out which words fit best and then strings these together. Because this can all seem a little abstract, the German children’s TV show “Sendung mit der Maus” explained how it all works in a way that is easy to understand, using the example of storytelling (Link to Youtube, video in German language):
You can turn on English subtitles in YouTube (as well as other languages).
Of course, there is still so much about the inner workings of an LLM that we do not understand. But even if it might sometimes seem so, an LLM isn’t sentient – it has no awareness, feelings, or goals, just a powerful ability to predict and generate text.
In the end, the big question remains: Where will this technology lead us? Hopefully, the mouse will cover this topic soon!
Summary:
What could easily be considered “intelligent” or even “empathetic” in AI is in reality extremely quick calculations by a huge language model. The mouse explains how this works.