How to Think Outside the Box

There is a beautiful phrase by Ludwig Wittgenstein: "the limits of my language mean the limits of my world". In other words, we always think inside the box. And if we want to think outside the box, we need to go beyond the limits of our language. 

Interestingly, it's not only humans who have this problems, LLM are also limited by their context. The bigger the context becomes, the harder it is for them to come up with novel ideas. 

However, there is a way to fix that. The approach that I'm going to share with you today uses a knowledge graph as a map to represent your thinking. This map shows you where you are, what ideas you're thinking about, how they relate to one another, what clusters they form. It gives you a really good visual representation of your thinking, the territory. 

In fact, this is a visualization of everything that was written so far:

We can see the main ideas, the topics, relations between them. In fact, this knowledge graph representation approximates how we think and also how LLMs represent information. 

However, the most interesting part is what is missing from this map and what exists beyond the periphery of our thinking. And this is exactly what the graph can be very useful for, because shows us the periphery of our knowledge and this periphery can be used to transcend beyond our bubble and think outside the box. 

Let me show you how this works. 

For instance, the "context limits" cluster above is at the periphery of the graph, let's focus on it for a moment:

It's at the very edge of this text and it's linking this whole topic to LLMs: how AI can be limited by the context.

Just by looking at this topical cluster, we can already start developing this whole discourse in a new direction: linking it to context engineering, LLMs, and exploring the different ways to make AI think in ways that are more creative. 

This is the point where we can leave our thinking bubble and explore some news ideas that may be completely unrelated at first but that can produce some interesting and meaningful connections to whatever it is we are thinking about now. For instance, maybe there's an interesting insight on creativity in AI science? Maybe we can apply some methods already used in LLMs (temperature adjustment) to think more creatively as humans? 

In fact, we can use the built-in AI insight module to ask the machine themselves: 

• select the topical cluster at the periphery of the graph
• click AI: Transcend Selected Topic to have AI generate an idea that originates in this cluster but takes us beyond the limit of the knowledge graph

This is a super interesting idea, actually. The proposition is not to try to think outside of the box (into the emptiness) but to, rather, think of some other, completely different boxes, which may seem unrelated, and to then think of connections between them. 

As Alfred Korzybski said, "a map is not a territory". In fact, a map will always have a boundary and one of the most interesting ways to use it is to explore what's lying beyond the boundary. If the map represents language, we can start with words. But as soon as we start using new words (or maybe stop using the words completely) we enter a completely different realm: a part of the territory we haven't explored yet, rich in novel ideas and insights.