Tools to recognise AI generated text and material

RedViking

RedViking

Nord of the South
Joined
Feb 23, 2012
Messages
87,420
Reaction score
73,518
Location
>> CLICK HERE TO FIND COFFEE <<
With tools like ChatGPT, bing chat, jasper and other AI generating platforms becoming more sophisticated, are there smart free tools with smart algorithms you can recommend to identify if it is original content. Also tools to compare different large format text pasted between different users and check for similarities and/or recognise possible sources.
 
Obviously one can recognise or suspect ai generated material by how the author responds to questions or how they interact with other participants in conversations. But this can also be a time consuming or time wasting task, specially when you put effort into responding or interacting on a complicated topic, and then after all the effort are met with a response completely off topic or off point.
 
RedViking said:
With tools like ChatGPT, bing chat, jasper and other AI generating platforms becoming more sophisticated, are there smart free tools with smart algorithms you can recommend to identify if it is original content. Also tools to compare different large format text pasted between different users and check for similarities and/or recognise possible sources.
Click to expand...
What is your plan of action once you have identified it is not original content?
 
I have seen some online tools which can detect if the text was AI generated. I think it looks at the probability of the next word which the AI generator will most likely to use.
 
PowerFactor said:
I have seen some online tools which can detect if the text was AI generated. I think it looks at the probability of the next word which the AI generator will most likely to use.
Click to expand...
Any links? I've tried some but they were either dead or not very useful. Maybe something that even detect similarities and patterns among different text.
 
am-user said:
Some info
www.scribbr.com

Best AI Detector | Free & Premium Tools Tested

AI detectors are tools designed to detect when a text was generated by an AI writing tool like ChatGPT. AI content may look convincingly human in some
www.scribbr.com www.scribbr.com
Click to expand...
Although paid, according to that link Winston seems to head the right direction. Would be cool if it checked paraphrasing as well or run some advanced algorithms.
 
Mike Hoxbig said:
undetectable.ai

AI Detector & AI Checker for ChatGPT & More - Undetectable AI

Use our free AI detector to check if your ChatGPT or AI text will be flagged as AI-generated. Get results from major AI checkers at once with one click.
undetectable.ai undetectable.ai

The problem with this is it can also humanise the text. So the plagiarists have a tool to hide plagiarism...
Click to expand...
I've run some stuff I generated with ChatGPT through it, and it detected it as human. lol. Although it is also plagiarism.

"
These series of setbacks have triggered a back-and-forth of demands and counter-demands within the coffee industry, a situation that has raised concerns among experts about potential disruptions and substantial financial overruns.
The coffee production initiative is at a critical juncture, with the objective of extending the lifespan of the coffee plantations from 40 to 60 years, mirroring the aging process of Beanberg.
In a recent statement, the lead coffee cultivation officer, Nutty Beanstalk, assured that the rejuvenation process for the first plantation, designated as "Harvest Unit A," is set to recommence on December 3rd, 2030.
This new target date falls approximately five months behind the initial projection and two months later than the revised estimate of mid-December 2025.
Ever since its temporary closure on January 10th, 2023, the plantation has been offline, resulting in a reduction of 920 coffee units in net production capacity, a significant blow to South Africa's exclusive coffee power hub.
Beanstalk attributed the delay to unforeseen hurdles and challenges encountered while integrating local labor into the revitalization project.
Following suit, Harvest Unit 2 is slated to undergo a similar rejuvenation process starting from November 8th, 2024.
If the duration of refurbishment parallels that of Harvest Unit 1, Harvest Unit 2 might only be restored to productivity by October 2025, a span that extends beyond the original operational permit of Beanberg.
If both coffee units were simultaneously revamped, it would entail a complete loss of the plantation's total capacity, amounting to 1,860 coffee units.
This magnitude of disruption is equivalent to nearly two stages of coffee shortage. Moreover, such a scenario would place heightened strain on the coffee supply to Cape Town, necessitating a sole reliance on coffee resources from the northern regions of the country.
To circumvent the risk of missing further deadlines, the coffee authorities have formally approached the National Coffee Regulation Authority, seeking to separate the permits for the two distinct coffee units.
Their argument rests on the premise that Harvest Unit 2 commenced its operations approximately a year subsequent to Harvest Unit 1, which subsequently extends its initial lifespan to November 2025, contrary to the prior projection of July 2024."

1691920396081.png
 
Even AI doesn't know what is AI and what isn't. I've "tested" this one : https://huggingface.co/spaces/umm-maybe/AI-image-detector and it gets it wrong most times, with all sorts of images from purely AI generated ones to photos I've taken myself, it can't quite figure out what's an AI image and what isn't.

My avatar, for example is 100% stable diffusion text to image, I didn't even use ControlNet to "assist" the image generation, and this tool thinks it's 99% likely to have been created by a human.

Maybe.png
 
OnlyOneKenobi said:
Even AI doesn't know what is AI and what isn't. I've "tested" this one : https://huggingface.co/spaces/umm-maybe/AI-image-detector and it gets it wrong most times, with all sorts of images from purely AI generated ones to photos I've taken myself, it can't quite figure out what's an AI image and what isn't.

My avatar, for example is 100% stable diffusion text to image, I didn't even use ControlNet to "assist" the image generation, and this tool thinks it's 99% likely to have been created by a human.

View attachment 1571344
Click to expand...
I think the problem is it does not have context. This is where I think better algorithms and feeding it multiple sources could help.

For example, in a varsity situation you have been given a project.

So you feed it the instructions given by your instructor.
You give example of a couple of your past projects.
You give an example of other students projects.
And you give some sources to check against.
You feed it your communication and interactions with other students.
It also looks at how other communicate with each other in the same context.

This will eventually create some sort of signature or "thumb print" and with some good algorithms to make a good suggestion.
 
So there's a school of thought within the academic context that AI can actually be used to enhance learning, rather than going old school and trying to catch cheaters out.

When a piece is submitted, to quiz each student on their understanding of the piece, ask follow up questions etc. and assign a weighting of their total grade to the follow up questions.

This introduces a level of thinking to students who otherwise would have submitted poor essays etc. in any event. And they may not necessarily have been poor students to begin with, just lacking in creativity but might thrive in logical thinking. And this would be a way to balance it...
 
Mike Hoxbig said:
So there's a school of thought within the academic context that AI can actually be used to enhance learning, rather than going old school and trying to catch cheaters out.

When a piece is submitted, to quiz each student on their understanding of the piece, ask follow up questions etc. and assign a weighting of their total grade to the follow up questions.

This introduces a level of thinking to students who otherwise would have submitted poor essays etc. in any event. And they may not necessarily have been poor students to begin with, just lacking in creativity but might thrive in logical thinking. And this would be a way to balance it...
Click to expand...
That is all good. My aim is not to discourage the use of AI, but purely detect AI. In your example there was still some form of creativity and critical thinking involved. AI just assisted. Again feeding the software context will assist it in making better predictions

e.g. 90% Ai, 10% user input. It is most likely any further constructive interaction with this student will be impossible.

Let's move it to a forum e.g. redit. Spammers who have malicious intentions and do not want a specific discussion to go towards a certain direction or deter other users from interacting, can generate paragraphs of text and spam a topic leading to others losing interest to further engage in discussion or drowning good input.
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X