AI text vs Human text

[JuanMena]

Okay, first:

Now... based on what, are we able to tell "hey, 101% of this text is AI generated" ???

Based in the amount of typos?
Structure of the code?
Amount of fancy words utilised in the main text?

For fucks sake, the last time I asked ChatGPT to generate code, it didn't fucking worked (about 8 months ago) and spent 2 days trying to debug it but realised it was non-sense mostly! But that might be on me since I requested C#, so not sure if it can "handle better" other languages.
The AI can't even handle simple .cfg files instructions even and when it kinda does, it doesn't work because there's no logic nor structure in them.

So, whoever used AI to write code at a 101% reading efficiency, pleade let me know what fricking blessed and logical AI you used!

What's your thoughts?

 

[AngelaBThomas]

Okay, first:

View attachment 504447

Now... based on what, are we able to tell "hey, 101% of this text is AI generated" ???

Based in the amount of typos?
Structure of the code?
Amount of fancy words utilised in the main text?

For fucks sake, the last time I asked ChatGPT to generate code, it didn't fucking worked (about 8 months ago) and spent 2 days trying to debug it but realised it was non-sense mostly! But that might be on me since I requested C#, so not sure if it can "handle better" other languages.
The AI can't even handle simple .cfg files instructions even and when it kinda does, it doesn't work because there's no logic nor structure in them.

So, whoever used AI to write code at a 101% reading efficiency, pleade let me know what fricking blessed and logical AI you used!

What's your thoughts? geometry dash lite

If you're unsure, try looking for subtle repetition or unnatural phrasing.

 

[JuanMena]

If you're unsure, try looking for subtle repetition or unnatural phrasing.

Unnatural as ???
Or Unnatural as a foreign language I haven't bothered to learn?
Or Unnatural as: "I'm sure this person is an idiot, and can't follow grammar rules... so this sudden improvement must mean one thing only."
Or Unnatural as: "I'm sure my knowledge in foreign languages doesn't utilise this verb/pronoun this way because it's wrong, so it must mean one thing only."
Or Unnatural as: "I'm not understanding what I'm reading, so that must mean one thing only."

Also, repetitions (grammatical redundancies) are common in any language.

Post automatically merged: Jun 27, 2025

"Buffalo buffalo Buffalo buffalo buffalo buffalo Buffalo buffalo."

English is flawed.

 

[Reiten]

Okay, first:

View attachment 504447

Now... based on what, are we able to tell "hey, 101% of this text is AI generated" ???

Based in the amount of typos?
Structure of the code?
Amount of fancy words utilised in the main text?

For fucks sake, the last time I asked ChatGPT to generate code, it didn't fucking worked (about 8 months ago) and spent 2 days trying to debug it but realised it was non-sense mostly! But that might be on me since I requested C#, so not sure if it can "handle better" other languages.
The AI can't even handle simple .cfg files instructions even and when it kinda does, it doesn't work because there's no logic nor structure in them.

So, whoever used AI to write code at a 101% reading efficiency, pleade let me know what fricking blessed and logical AI you used!

What's your thoughts?

I've been asking ChatGPT for java, typescript and html for quite some time and it does ok, at least on a function and class level. Though the more code that you want it to generate, the more precise your prompt has to be, as it easily gets details wrong. Though you still have to check and make sure the code works as intended as it sometimes doesn't get what you want, like wanting a comperator that will sort in a descending order, but it gives you one that does it in ascending order.

Also, how the heck can 101% of something be AI generated? Is that checker ok?

 

[JuanMena]

I've been asking ChatGPT for java, typescript and html for quite some time and it does ok, at least on a function and class level. Though the more code that you want it to generate, the more precise your prompt has to be, as it easily gets details wrong. Though you still have to check and make sure the code works as intended as it sometimes doesn't get what you want, like wanting a comperator that will sort in a descending order, but it gives you one that does it in ascending order.

Also, how the heck can 101% of something be AI generated? Is that checker ok?

Regarding code I'm not using ChatGPT for that, right now I'm studying (by myself, with documentation) Python and occasionally ask ChatGPT to explain/give documentation on whatever I might bot understand.

The only times I asked ChatGPT to actually generate code, was with GDScript for Godot and C# for Godot, which both I'm familiar with and more experienced thatn Python.

ChatGPT tends to generate somewhat working code, but as soon as you try to ask the bot to correct a mistake, it regenerates everything, which just complicates things because there's not a clear use in syntax.
Debugging code generated by ChatGPT is a nightmare... or a good exercise, depends on how you look at it.

About the later, probably the checker was AI generated... which is why:

So, whoever used AI to write code at a 101% reading efficiency, pleade let me know what fricking blessed and logical AI you used!

 

[Dark_Ansem]

All AI detectors are mostly bullshit, the original research paper about detection said so very clearly. Even more so in English, which is a language that normally doesn't allow for much creativy in sentence construction. Not like, say, Latin.

 

[JuanMena]

All AI detectors are mostly bullshit, the original research paper about detection said so very clearly. Even more so in English, which is a language that normally doesn't allow for much creativy in sentence construction. Not like, say, Latin.

Humans are bullshit: https://www.reddit.com/r/needadvice...eacher_is_going_to_fail_me_after_accusing_my/

 

[Dark_Ansem]

I mean, that too, but do you have a point overall?