Paul_Chance
The Watcher
- Joined
- Dec 23, 2011
- Posts
- 21,436
It's important to understand how these tools work, so you can decide whether or not you want to rely on them for assistance. For those who are unfamiliar with what they're actually doing, and why sites (like Lit) are discouraging and opposing their use.
All large language models work on the same principle.
When you write a sentence and pass it through the tool, it goes like this.
1. The tool loads your sentence into it's language model. (So, you are essentially helping train the LLM, for free.)
2. The tool analyzes your sentence and breaks it down into component parts.
3. The tool then compares your sentence with its library of similar sentences and phrases.
4. Then, it provides you with the most common expression of similar sentences as it's suggestion/recommendation.
5. So, when you select that suggestion to "clean up" your writing, you're electing to write your sentence the exact same way hundreds of thousands of other people have written it.
6. Depending on how many times you follow the suggestion of the AI tool it turns your original, creative work into a bland repetition of the same sentences and phrases. It's a ruthless run to average, unimaginative, bland writing.
7. You might have a great idea for a creative and original story. But the AI tools that are "helping you" are making sure, subtly and powerfully, that you're expressing it in the most common way. Literally by comparing it to millions of other examples and suggesting you change it to the way it's been written millions of other times.
The AI detectors are simply LLM's that are doing the same thing and simply flagging it when it's done. They take your sentence, compare it the model, return the same result (the most common way it's been expressed in the past), and simply say "Hey, this is what an AI writing tool provides."
All large language models work on the same principle.
When you write a sentence and pass it through the tool, it goes like this.
1. The tool loads your sentence into it's language model. (So, you are essentially helping train the LLM, for free.)
2. The tool analyzes your sentence and breaks it down into component parts.
3. The tool then compares your sentence with its library of similar sentences and phrases.
4. Then, it provides you with the most common expression of similar sentences as it's suggestion/recommendation.
5. So, when you select that suggestion to "clean up" your writing, you're electing to write your sentence the exact same way hundreds of thousands of other people have written it.
6. Depending on how many times you follow the suggestion of the AI tool it turns your original, creative work into a bland repetition of the same sentences and phrases. It's a ruthless run to average, unimaginative, bland writing.
7. You might have a great idea for a creative and original story. But the AI tools that are "helping you" are making sure, subtly and powerfully, that you're expressing it in the most common way. Literally by comparing it to millions of other examples and suggesting you change it to the way it's been written millions of other times.
The AI detectors are simply LLM's that are doing the same thing and simply flagging it when it's done. They take your sentence, compare it the model, return the same result (the most common way it's been expressed in the past), and simply say "Hey, this is what an AI writing tool provides."
