The Half-Full Glass Principle for Using AI

We’ve already reached a point where using AI can no longer be avoided, especially in the creative industry and anything related to digital.

The use of AI is getting more and more widespread, from multinational companies all the way down to small businesses and government.

Of course, the impact felt so far is convenience and a lot of work being made easier with AI. Faster processes, more ideas. Effective and efficient.

But too many people are dazzled by AI’s capabilities to the point of forgetting how it works and how we should actually be using it to the fullest.

The Half-Full Glass Principle

I’m reminded of when I first got into the world of media and journalism, where I was introduced to the Half-Full Glass Principle.

This principle essentially teaches that as a “journalist” you can’t be in an empty state, and you can’t be in a full state either. In this context, it’s about the information you dig for, or dealing with a source.

It means you already have some initial information about what you want to dig into, or about the source you want to interview. But it isn’t full — there’s an empty space that you still need in order to make it whole.

The empty space left in the glass is the information you’re looking for, to complete what you already have.

Meanwhile, what’s already in the glass is the foundation that keeps you from being easily fooled, because you’ve come prepared.

The same goes for using AI: ideally, when you ask AI for something, you already have a foundation or basic knowledge — whether from books, podcasts, journals, or whatever else.

That way you won’t be easily steered around and you won’t immediately believe whatever AI serves up.

The Importance of Understanding How AI Works

The discussion above has an important connection to understanding how AI itself works, instead of being dazzled by results that look wow.

AI today produces data from training data. That means AI (an LLM, in this context) is trained on an enormous amount of data so that it can generate answers as if it’s talking to us.

In short, AI produces answers not from a thinking process, but by predicting words.

That’s why we still have to be skeptical of the data or text that AI produces. Because there’s a chance that its training data is wrong, or even that its word predictions when generating an answer are less than accurate.

Use AI to Learn

The knowledge that has been fed into AI as training material is incredibly vast, and it may well be the largest amount of knowledge that exists in AI so far.

I agree with a statement Ferry Irwandi made in one of his Instagram stories back then, that AI is most powerful when used for learning.

Because we can learn about any knowledge in the world from AI, since it has all of it.

Conclusion

Because AI isn’t 100% correct, in the end it still comes down to us. We’re the ones who can filter whether to use AI’s output, believe it completely, or stay skeptical — with the final result still coming from our own minds.