Making AI Sound More Like Us

Simplifying Output Through the Flesch Reading Score

Ever felt like chatbots or AI programs try too hard?

They often use big words and long sentences, adverbs, passive voice and hyperbolic language that might sound fancy but feel unfriendly.

Some of the cringe-inducing regular words I see appearing frequently include: "Pivotal" "Transformative" "Paramount" "Navigating" "Fostering" and "Crucial".

I bet you noticed some regular AI words like these too, let me know in the comments what the dead giveaway words you spotted that immediately give you a clue that you're reading something LLM-generated.

It's like having a conversation with a textbook. The truth is, when we chat, whether with friends or colleagues, we keep it simple. That's what makes our conversations feel natural and easy to follow.

And simple language matters. In the USA, 53% of adults have literacy levels below that of a 6th grader. But with the default "voice" of LLMs you can come off sounding, well, robotic and tryhard, even unreadable. Worse still, the wider international audience of non-native English speakers may not be able to catch your drift.

Now, there's a way to make AI sound more like us: By using the Flesch Reading Score. This score tells us how easy it is to understand something written. The higher the score, the simpler the text. By aiming for a score of 80, 90 or above, we can make AI write in a way that's easy and pleasant to read - almost like it's coming from a friend.

Why does this matter? The main point of using AI, whether for work, learning, or just fun, is to make our lives easier, not harder. When AI talks like a human, using straightforward language, we can understand it quickly and use it more effectively.

It's about making technology work for us, not the other way around.

In practice, this means programming AI (via Custom Instructions or the Prompt itself) to choose simpler words and shorter sentences. It's like teaching it to communicate in a way that's natural for us. This doesn't just make AI more helpful; it also makes our interactions with it more enjoyable.

By focusing on making AI's language clear and straightforward, we're taking a big step toward technology that truly fits into our lives without being so annoying. It's not just about making machines smarter; it's about making them more relatable and easier to use.

So, the next time you come across a piece of writing from AI, imagine how it would sound if it scored 80, 90 or above on the Flesch Reading Score. Give it a try.

Wouldn't it be nice if all technology communicated that clearly? It's a goal worth aiming for as we blend technology into our daily lives.

To execute this hack, GPT 4 pro users can add the following sentence to custom instructions, and for the rest of you using the free version of GPT, you can simply add this to any prompt you want to sound more.... human!

Let me know how you get on!

"Write at a Flesch Reading Score of 80 or higher"