Auto tag suspected AI generated posts with a label

[Unicorn Slayer]

Many sites are already doing this so if users post AI generated content steam should flag it and mark it with a suspected AI generated tag.

[Truth]

They should go further and automatically hide the posts if they detect it was done by AI, the AI spam is getting ridiculous.

[mldb88]

Originally posted by Truth:

They should go further and automatically hide the posts if they detect it was done by AI, the AI spam is getting ridiculous.

It’s mostly one user having a hissy fit over a rightful ban, one jumping on the bandwagon, and one intentionally doing it to show that AI sometimes just makes ♥♥♥♥ up if you give the right prompts as basically proof that the other two claiming AI is basically objectively correct have no basis in reality

[KalGimpa]

they are?

i have never seen this, myself.

google tags with ai and the like, but not forum posts

do not use facebook or twitter or anything, so maybe on those?

wouldn't they have to us an ai program to look through so many posts to see if it is ai?

[Haruspex]

How do we know if a post was generated by AI? I know I've been accused of being AI before. To some people, using punctuation and forming complete sentences is somehow "evidence" for a post being written by AI.

I think the best evidence is when you take the context of a user's post history into account. For example, if a user typically posts nothing but Gen Alpha brain rot all the time, and all of a sudden they spit out an eloquently written and long post that makes all relevant points to the topic at hand, there's a good chance (but not a guarantee), they asked ChatGPT for a response and copied it over.

[76561199793358131]

Originally posted by Haruspex:

How do we know if a post was generated by AI?

There is online analysis tools to detect AI usage.

[Truth]

Originally posted by Haruspex:

How do we know if a post was generated by AI? I know I've been accused of being AI before. To some people, using punctuation and forming complete sentences is somehow "evidence" for a post being written by AI.

I think the best evidence is when you take the context of a user's post history into account. For example, if a user typically posts nothing but Gen Alpha brain rot all the time, and all of a sudden they spit out an eloquently written and long post that makes all relevant points to the topic at hand, there's a good chance (but not a guarantee), they asked ChatGPT for a response and copied it over.

There are online tools like this that will detect it - https://www.zerogpt.com/#google_vignette

[Haruspex]

Originally posted by Darth Yshrak:

There is online analysis tools to detect AI usage.

Originally posted by Truth:

There are online tools like this that will detect it - https://www.zerogpt.com/#google_vignette

These are all utter garbage. Teachers have been using them in schools to "prove" a student wrote the paper with AI, and those students then turn it around by using the same tool to "prove" the teacher's own writing, or passages from the Bible were written using AI. One instance I've seen shows how you can fool these AI detectors by simply telling the AI to write in such a way that is unlikely to be triggered by these AI detectors.

Fact is, good clean writing is indistinguishable from AI writing.

[Truth]

Originally posted by Haruspex:

Originally posted by Darth Yshrak:

There is online analysis tools to detect AI usage.

Originally posted by Truth:

There are online tools like this that will detect it - https://www.zerogpt.com/#google_vignette

These are all utter garbage. Teachers have been using them in schools to "prove" a student wrote the paper with AI, and those students then turn it around by using the same tool to "prove" the teacher's own writing, or passages from the Bible were written using AI. One instance I've seen shows how you can fool these AI detectors by simply telling the AI to write in such a way that is unlikely to be triggered by these AI detectors.

Fact is, good clean writing is indistinguishable from AI writing.

They've improved quite a bit since those days, they are about 60-85% accurate now, and there are a lot of tells with formatting that helps now.

Like for longer articles most AI format it like this

One sentence summary

Then a paragraph about the topic

1. xxxx

2. xxxxx

3. xxxx

Conclusion (sometimes omitted the word conclusion)

Then often writes redundantly In conclusion, after the conclusion part

[KalGimpa]

Originally posted by Truth:

Originally posted by Haruspex:

These are all utter garbage. Teachers have been using them in schools to "prove" a student wrote the paper with AI, and those students then turn it around by using the same tool to "prove" the teacher's own writing, or passages from the Bible were written using AI. One instance I've seen shows how you can fool these AI detectors by simply telling the AI to write in such a way that is unlikely to be triggered by these AI detectors.

Fact is, good clean writing is indistinguishable from AI writing.

They've improved quite a bit since those days, they are about 60-85% accurate now, and there are a lot of tells with formatting that helps now.

Like for longer articles most AI format it like this

One sentence summary

Then a paragraph about the topic

1. xxxx

2. xxxxx

3. xxxx

Conclusion (sometimes omitted the word conclusion)

Then often writes redundantly In conclusion, after the conclusion part

with as many people posting on the boards, that is a huge number of mistakes

and a lot pissed off people telling support to remove those marks

[Truth]

Originally posted by KalCuey:

Originally posted by Truth:

They've improved quite a bit since those days, they are about 60-85% accurate now, and there are a lot of tells with formatting that helps now.

Like for longer articles most AI format it like this

One sentence summary

Then a paragraph about the topic

1. xxxx

2. xxxxx

3. xxxx

Conclusion (sometimes omitted the word conclusion)

Then often writes redundantly In conclusion, after the conclusion part

with as many people posting on the boards, that is a huge number of mistakes

and a lot pissed off people telling support to remove those marks

You'd be surprised, it takes a considerable amount of text. Feel free to use it and test it yourself.

[KalGimpa]

Originally posted by Truth:

Originally posted by KalCuey:

with as many people posting on the boards, that is a huge number of mistakes

and a lot pissed off people telling support to remove those marks

You'd be surprised, it takes a considerable amount of text. Feel free to use it and test it yourself.

i am just going off of your post's numbers

that is 15-40 wrong marks out of every 100

multiply that by the tens of thousands of posts (whatever the huge numbers are)

and you are getting way too many people pissed that they are getting called a bot

[Truth]

Originally posted by KalCuey:

Originally posted by Truth:

You'd be surprised, it takes a considerable amount of text. Feel free to use it and test it yourself.

i am just going off of your post's numbers

that is 15-40 wrong marks out of every 100

multiply that by the tens of thousands of posts (whatever the huge numbers are)

and you are getting way too many people pissed that they are getting called a bot

You should really try some of them, if you don't know how they work that explains alot. They have tools that can tell you anywhere from 50-100% accuracy if they feel its AI generated. For instance i've seen posts here that come back 100% AI generated, those are the ones i'd say to go after.

[Sir Seanicus, Esq.]

Every moderator post would be labelled AI.

[Truth]

Originally posted by Sir Seanicus, Esq.:

Every moderator post would be labelled AI.

Nope

Originally posted by Truth:

it takes a considerable amount of text.

Its not a tool used for a sentence or even a paragraph. Again if you want go to steam discussions on the first page, look at a few lengthy threads and plug the text in and see for yourself.

[Sir Seanicus, Esq.]

Originally posted by Truth:

Originally posted by Sir Seanicus, Esq.:

Every moderator post would be labelled AI.

Nope

Ah. You're right. OP said "suspected AI"...not actual AI, which the moderators are.