Think about this: writing is really cool. We take our thoughts and ideas and put them into words. This way, other people can get what we mean. But even in those words, things aren’t always straight. Some parts are tricky and surprising.
If you think about it, writing is really cool. It’s our way of sharing thoughts and ideas by shaping them into words so others can get what we’re saying. Yet, the way we use language has its own set of quirks and surprises. So, you might wonder, can machines learn to deal with these quirks in language?
Well, yes, they can. Usually, when people talk about the tricky parts of writing, like how unexpected or complex it can be, they dive deep into math and hard-to-understand theories. But I’ll keep things simple here. This way, you can improve your writing skills without getting bogged down in numbers or overly detailed analysis.
Talking About Words and How Much We Use Them
Basically, it’s all about how we use words and how much. For those who write or make content, our writing can be easy to guess. Plus, sometimes, the words we choose come out in clusters or “bursts.” Let’s take a closer look at both parts.
What is Perplexity?
It’s a math formula used to figure out perplexity.
This formula helps figure out how easy it is to guess the next word you’re going to read. Think about reading a story and trying to guess the next word. If guessing is easy, the formula’s number is low, which means it’s not too hard. If guessing is hard, the number is high, meaning it’s tricky.
What does this mean if you’re writing something? Well, if your writing is too easy to guess what comes next, it might be dull. But, if it’s too hard to guess, it could be confusing and tough to follow. The key is finding a middle ground to keep your readers interested and make sure they understand your message.
What does Burstiness mean?
Sometimes, us writers stick too much to certain words, using them over and over. Think about writing a story at a birthday party. You might use “cake” a lot at first when setting the scene. But once you’ve described it, you likely won’t mention “cake” much again. That’s what we call burstiness.
When you read a story and meet new characters, their names pop up a lot at first. But by the end, not so much. Why? Because you know them well by then. That’s called burstiness.
Why should you care, especially if you write stuff? Well, using words more in some parts of your story can shift what people pay attention to. Some writers do this a lot, like it’s their signature move. And, repeating words can really hammer a point home. Just like with puzzles, getting the balance right is super important for making your writing good.
How Do Confusion and Sudden Changes Work Together?
Confusion and sudden changes might look totally different, but they actually overlap. If there are lots of sudden changes in a piece of writing, predicting the next word becomes harder because the writing isn’t as uniform. Some writings or articles might have more sudden changes than others. A lot of this depends on the writer’s own style.
As AI gets better and tools start to understand the small details of how people use language (including confusion and sudden changes), you’ll see these tools do more than check grammar and spelling. Soon, they might help writers know if their writing is too easy to predict or tell them to add more surprises in their words. This way, writers can make their work more interesting, change up their style, and have fun with how predictable their writing is.
How Does This Affect Your Writing?
In almost all writing, adding bursts of unique words is important. Picture adding war words into a love story book. This tells the reader that a fight or big problem is coming up. Different kinds of writing might use lots of unique words more than others. For example, science and study books often have special words here and there that make patterns of these unique word bursts. Sometimes, writers choose to use lots of unique words on purpose.
Martin Luther King Jr.’s “I Have a Dream” speech is a good example. He repeats some words over and over to make his ideas stronger and easier to remember. When you write, changing how much you use these special word bursts can make what you write more interesting. This can help keep your readers interested for a longer time.
Why Talk About Math and Guessing Words When Teaching AI?
Why bother with math stuff and guessing or finding groups of words? Why is it important? This stuff helps teach AI and tools like Originality.ai that spot when something is AI-made. When training AI to get and make language, it learns from lots and lots of text. It tries to guess the next word from the words it already knows. At its heart, it’s really about spotting patterns.
Perplexity helps figure out how good an AI is at guessing the next word. When it guesses right a lot, its perplexity is low. But when it gets it wrong often, its perplexity is high. So, when people make AI, they try different ways to make one that guesses the best. They look for options with low perplexity because that means it’s a better model.
Now, let’s talk about something called burstiness. It’s a common thing in how we talk and write, like when a new person shows up in a story, and their name pops up a lot quickly. AI needs to understand and copy this way of talking to work well, especially when dealing with stories or novels.
We need to make sure the AI doesn’t use the same words too much or keep saying things the same way over and over. It’s important for the writing to feel real and easy, as much as it can. This means the people making the AI must mix things up, just like writers who aren’t robots do. They also need to teach the AI about lots of different kinds of writing, not just one kind or one sort of story. This helps the AI learn different ways to write.
Why is This Important?
There are two big reasons why we teach and train AI (and also why we work on AI detection systems):
First, the people who make AI want it to talk well with humans. To make this happen, it needs to sound very real. To sound real, the makers need to manage how confusing or surprising the AI’s speech is. This makes the AI’s text sound more like what a person would say.
The more we learn about how good the AI is doing, the more the people who build it can make it better for the next time. It’s like they don’t have just one way to make it perfect for everyone. They keep making changes to make AI write and talk more like us.
In short, knowing about how confusing or unpredictable writing can be is really important. Not just for writing but also for making AI better at finding and making writing that sounds real. By adjusting how AI and machines learn, they get better at mimicking the way we talk. This makes what AI writes sound more real and helps the tools that check for AI writing get better too.