How To Get Around ChatGPT Detectors
VloĆŸit
- Äas pĆidĂĄn 15. 06. 2024
- Looking to get more out of ChatGPT? Get my free E-Book: myaiadvantage.com/newsletter
In this video, I'll be running AI-generated text through various plagiarism detection tools, to make sure that no one gets caught up in any plagiarism-related issues. We'll be using a traditional plagiarism detector, and then an AI-specific one. We'll also take a look at how to get around the plagiarism detectors, and how to use the rephrasing technology to make sure the text passes the AI detector. In the end, I'll be showing you a bunch of creative ways to use ChatGPT, so stay tuned!
---------------------------------------------------------------------------------
Links:
ChatGPT: chat.openai.com/
Duplichecker: www.duplichecker.com
Small set tools: smallseotools.com/plagiarism-...
GPT-2 Output Detector Demo: openai-openai-detector.hf.space
Quillbot: quillbot.com
Prompts:
đ Write me an essay about the importance of avoiding plagiarism in universities.
đŹ Write it in a tone that is not typical to AI.
đ€ Analyze the following text for style and tone of voice. Apply that exact style and tone of voice to all your future responses: [insert your writing].
Chapters:
0:00 Use responsibly
0:57 First check
2:00 AI maneuver
3:14 Paraphrase in AI
4:44 Be careful
#chatgpt #chatgptdetectors #chatgptplagiarism
---------------------------------------------------------------------------------
đ My Free ChatGPT Templates: myaiadvantage.com/newsletter
đ€Ż 1000+ AI Assistants and GPTs: shop.myaiadvantage.com/produc...
đ§ AI Advantage Shop: shop.myaiadvantage.com/
đŠ Twitter: / theaiadvantage
đž Instagram: / ai.advantage - VÄda a technologie
Will be doing a follow-up exploring this detection software in-depth just note that this video took half a day to shoot because the detector was always down.
Whats your opinion of other ai's like chatsonic?
Is chatgpt is best?
1. When I ask it to rewrite it, I always ask ChatGPT to keep the same layout because otherwise it will remove layouts like bulleted lists, etc and write coninuous texts.
2. Sometimes, adding just once a second space between two words lowers the fake rating of about 50%. I do this when I get something like a 11% fake.
3. Translating to another language (with google translate or deepl) and then back to English is yet another solution to reduce the fake rating. But not always.
4. I tried "Rewite it in a tone that is not typical to AI, keeping the same layout." and it didn't work while (as I usually do) trying a more direct way works better for me: "Rewrite your answer to pass through AI ouput detectors. Keep the same layout." works much better for me. Anyway, the important point here, is that you may try rephrasing your directive if it fails lowring the rating enough.
And now the very bad news: even as if English is not my mother tongue, it happens that those detectors are sometimes rating my very own writing as 90 or 95% fake, especially when I pay attention to write a good text. đ€Źđ€Źđ€Ź So, when it is important, even if you're honest, you'd better check your own work.đ€Źđ€Źđ€Źđ€Źđ€Źđ€Źđ€Ź
yay
Hell sir i am from Indian .I am studing inter 1st and 2nd year concepts with the help of chat gtp
.Please tell is it okay to study concelts from it and do i get always correct information from it?
What I do, In general when Iâm having a hard time brainstorming, I ask GPT to write me an outline, or even an essay. But I never neglect the research, and I never copy/paste for obvious reasons. I use it more as influence rather than a quick check in the box for my writing. Gotta remember to learn something at the very least.
Biggest problem is that chat gpt doesn't give you valid sources
@@irgendeinname9256 I was ask GPT to cite its sources and give the appropriate link in order verify. Sometimes though itâll give you a dead-end 404 page.
One obvious problem with AI detection is however that apart from the score you can't really prove that an AI generated it. Traditional plagiarize detectors link to the source and you can easily determine if the score is reliable or not.
Thay's not true, AI detectors can detect AI generated content based on pseudo-random textual watermarks, patterns of word distribution, and a mathematically calculated distribution of punctuation uncharacteristic of a human being. The patterns are based on mathematical probabilities which differentiate text produced by a machine in comparison to a human. Human generated text is much more random, whereas AI text has verifiable mathematical linguistic patterns built into the text, and that's the key difference.
@@subtronik9311 Yes. If you take text generated from an AI tool and you just put some extra spaces in it (as in change "my dog, Spot" to "my dog , Spot") by putting a space before the comma, and then do that a handful of times throughout the article it will be detected as close to 100% human written. That's how fragile the AI detection tools are.
I was running my own work through the detector stuff from years ago some of that was getting a 40% human rating...
@@smartduck904 Yeah it gave my own writing 50% AI generated, which it did for a paraphrased chat gpt content.
@@subtronik9311 i used texts long and short that i wrote completely by myself and it said it was mostly ai. It sucks ass and teachers shouldnt be allowed to use them to check hw or assignments because they are super unreliable
I am your recent follower, and you are doing a great job in demystifying this technology! :D
I am glad that what I have been doing is exactly what you are presenting in this video, to benefit from AI is truly a reflection of the individual using it, if you are creative you can benefit so much
Spot on. The best get even better
what if it's put in turn it in? does it show differently??
Love your content! While I never plan on using chatGPT to plagiarize and only ever use it as a starting point I love how informative your videos are
Wow you are a wise human
Anything chatGPT or other AI tools puts out is going to be close to 100% plagiarism free. they don't scrape the web and steal content.
Cap
Iâve been struggling with this for weeks, and this video finally cleared things up. Thank you!
Using a tool to check plagiarism and a tool to check whether the output is AI are two COMPLETELY different things. most, if not all, AI output will be close to 100% plagiarism free because it's not just copied from somewhere. For some reason, many people think AI tools scrape the web and copy text already out there, but they do not. (probably because many older "AI tools" would do just that, scrape the web and copy text and possibly spin it. they weren't AI tools like we think of with ChatGPT now.) In the simplest terms, checking to see if the text is from an AI generator is just a tool to check patterns to see if it's possibly AI-generated. If you take text generated from an AI tool and you just put some extra spaces in it (as in change "my dog, Spot" to "my dog , Spot") by putting a space before the comma, and then do that a handful of times throughout the article it will be detected as close to 100% human written. That's how fragile the AI detection tools are.
So basically I can use a response from ai and change some words around and it should show in turnitin as 100% authentic?
@@imohkryptic seems so but i wouldn't recommend it cuz u know if u only copy and past ur kinda screwt in exams
@@imohkryptic Change some of it and detectors will think it's more human. But it's quite unreliable, they need to hide some hidden hash or invisible characters when you copy paste. Just take some text you wrote and check it in any AI detector. These detectors often think original human texts are likely generated from AI. Maybe if you are bad at grammar
it will think it's 100% human XD
@@MickDarkstar1 yea lately Iâve been using chatgpt for some papers but not copying everything completely more like if I get stuck or need help starting a paragraph and I take from the response then plug it into two ai detectors or also run it through a paraphrasing site and add some of my words and so far itâs been great lmao
its great. but how to deal with the fake citation and references generated by chatgpt ?
GPT-2 output detector demo is broken. Just write anything with correct grammar and it will say, it is fake.
yea
Lol true
It works for me. I have tried many examples. Here is one of them.
"Zig is designed to be easy to use and provide a high-level interface for common tasks, while also providing low-level control when needed. It has a simple, expressive syntax and a standard library that provides a wide range of functionality. Some of the key features of Zig include static typing, automatic memory management, and support for multiple programming paradigms. " - written by ChatGPT which gets detected as fake. But by telling ChatGPT to rephrase it in a way that will not be detected by an AI detector, it does that too.
"Zig is an imperative, general-purpose, statically typed, compiled system programming language designed by Andrew Kelley. The language is designed for "robustness, optimality and maintainability", supporting compile-time generics, reflection and evaluation, cross-compilation and manual memory management." - Wikipedia (detects as real)
@@shazam314 Depends how you ask ChatGPT to rephrase. My own text is sometimes detected up to 99% fake !!
@@yoops66 Maybe your text is too short?
the biggest issue is with bad gpt detectectors, I know several people who are receiving 0% on their assignments and getting in trouble for plaigerism even though they didn't have a clue chat gpt existed when they wrote the essay. I heard about this and checked my engineering essay(which I wrote entirely by myself) and spent the next 4 hours changing shit until the gpt detectors were convinced I'm human...
hi. could you please share what would be the best platform to detect Ai writing?
@@bollywoodjam5304 Open AI has their âai classifierâ which is based off the same language model as chat gpt and admits when it canât tell. Iâm not sure if itâs the best though, better platforms may have been released since I looked
It is not the best, it says unclear, it is wrird
the same thing happened to my sister. she got accused of using chat gpt. she had rewrite the holw essay even if she didnt
this just happend to me, i used zero gpt to check my essay and it tells me that my text are 58% written by AI, i've never used ai to write my text because its still a new thing in my country
OpenAI is working on introducing watermarked essays that basically use vocabulary probability to watermark their outputs. For example, if the essay output has a lot of nouns and verbs that are used around 1-5% of the time when compared to their synonym counterparts, then this consistency could be used as a watermark. Forcing the student to at least re-word their essay to avoid GPTzero detection. Doing this for articles, conjunctions, and other parts of speech could make the watermark so powerful that the output becomes unusable without a complete re-write; only using the AI output essay as inspiration.
as inspiration is actually pretty fair
@@jochemvangalen9464 sadly there are free tools out there to paraphrase essays or they could just use grammarly or something
So it would just talk really formal and in common simple non complex vocabulary?
Okay this actually makes me hate OpenAI a little less now. They're doing something to stop people from being complete zombies with the usage of AI.
@@prabean69 lol
If you can make 2 aiâs, one for detecting ai text and one for writing ai text you can make the detecting one try to detect the making ones text and have them both learning of each other. Then you sell older versions of them so you are always ahead of everyone else.
2000 IQ
Great tutorial! This is exactly what I needed to get started.
The first time I used chat GBT was a a skelaton were i build up my reaserch paper from. put it through these detectors. For the AI i got a 0.03% AI and for plagerism i got an 11%, that 11 were the quotes only. Thank gos I have these tools today they save so much time.
Amazing content as always. You never disappoint!
This was super helpful, thanks! Subscribed for more.
I used chatgpt to do a paragraph for my essay because I couldn't word it and it was 2 am, I put it through a chatgpt detection website (it was the first one you went to in the video) and it didn't detect the paragraph after putting it through quillbot however it suspected some of the text that I had actually wrote to be ai, I usually write pretty formally and have decent grammar and punctuation but it was a surprise to me
I actually like to use ZeroGPT to get the determinable % of AI likeliness to around 20% or less then take that and throw it into a plagiarism checker. but i never thought to ask it to write in a way not typical to ai, i have been saying "Make it more unique", "I need it to be more engaging and human like."(not very affective). think this has saved me a lot of time in future rewrites by minimalism the number of times i ask it to rewrite. thanks for the great tips!
This Video is very Excellent. Thanks for Video.
I also was experimenting with that and I found that if you tell the ai to make grammar and punctuation mistake will make it almost undetectable
This video is a game-changer! Thanks for the insightful content.
In the still of night, when all is calm and bright
The moon shines down with silver light
The stars twinkle like diamonds in the sky
And the world is at peace, as the hours go by
The night whispers secrets, unknown till now
And the world seems to slow, as if taking a bow
The worries of day, slowly fade away
Leaving us with a sense of peace, to stay
So let us cherish this moment, fleeting and rare
And be grateful for the beauty that's everywhere
For tomorrow is a new day, with possibilities untold
And the night is a reminder, to cherish what we hold.
This poem is purely generated by chat gpt but no websites are able to detect that.
One flaw with using chat GPT for papers is, you'll end up typing more words trying to give it a good prompt, than you will if you just wrote the damn thing yourself
Not to mention the fact that OpenAI is creating a "watermark" for it's outputs. Even if you ask it to write something in a way atypical to AI, the fact it came from an AI will be detectable.
Nope. Didn't experience that. What kind of "watermark" do you think they may put in text? Sounds like an urban legend to me.
@@yoops66 They're currently developing it. Not employed yet.
@@yoops66 cryptography
@@alaminsagor7618 Looks like you don't know what you're talking about. How to put cryptography in clear text? And what for?
It could be done in pictures, because there are so many nuances of colors and the, for example, one out of two nuances could be used for the picture, and the second for the picture and hidden information.
For example, if the picture was only in grays, only even grayscales could be used for the pictures, and odd for hidden informations. And then it would be possible to encode data in binary base (0 would be even, and 1 would be odd). The human eye would barely see any difference with 256 levels of gray.
But with pure text, there is no such opportunity.
@@etterath7049 Ha, you mean you know secret things ? Which are your sources and references ? (URLs please)
Over time, people will become skilled at writing like AI, making these AI detectors obsolete as people's writing styles converge with that of AI.
Good point right there.
Oh why did i have to graduate 5 years ago this would have helped me soo much. cant tell you how much stress i had working on 5 diffrent assignments at the same time all wanting me to reserch compleatly diffrent things and somehow having to talk about it as if i knew what i was talking about. its not too bad when its a topic you know but when it ssomething you know nothing about and worse is a boring topic you eather use wikapida to write it for you or just copy from a fraind. having a ai do it for you means theres next to 0 chance of you getting caught and even better theres almost no limit to what they can write about so i think they could compleat an entire uni corse for you
Where could I find a link to your ebook?
The secret is in the complexity of action. The problem with putting thoughts into written material is that, in order to be original, we need to address a very specific problem in a unique way and mix it with a new idea with a specific tone. Then, we need to know our own word usage because of common sense (everyone has an idea of how we speak and choose words). Formula: Write what you want, then rewrite it with your intention, specify the level of language. Then do it again and differently, then ask the AI to compile the text based on both ideas. Finally, change some words and it will be perfect.
please elobarate on the formula its kinda confusing
@@preeflea Sure
1- First, write down what you want to say. Your first draft might be: "Exercise is good for your health and can help you feel better."
2- Next, revise it with your intended message and language level. Your revised draft might be: "Regular exercise has numerous benefits for your physical and mental health. Studies show that exercise can help reduce the risk of chronic diseases like heart disease and diabetes, while also improving mood, reducing stress, and boosting overall well-being."
3- Repeat the process a few times, using different language and organization to refine your ideas. Your third draft might be: "Exercise is a powerful tool for promoting long-term health and well-being. From reducing the risk of chronic disease to improving mental health and cognitive function, there are countless reasons to make exercise a regular part of your routine."
4- Use AI to help you combine your different ideas. You might use a tool like GPT-3 to generate new ways of phrasing your ideas or suggest new points to include.
5- Finally, make some small changes to your wording to create a final version that is polished and effective. Your final draft might be: "Whether you're looking to improve your physical health or boost your mood, regular exercise is one of the most powerful tools at your disposal. From reducing the risk of chronic diseases to enhancing cognitive function and improving overall well-being, exercise offers a wealth of benefits that are hard to ignore."
God damn. That is an amazing content mate!
Also code written by AI is hard to tell apart from a senior developers. If you tidy your code, know good coding practices, abstraction and be quite generic of naming conventions for some piece of code, it will look the same as something an AI wrote (cause it also has access to all that knowledge you have), but it just writes it faster. It's a great tool, as when intellisense was new. Just speed things up
great video!!
Danke fĂŒr dein super Content :) kennst du gute KI Detector Tools fĂŒr den deutschsprachigen Raum? die meisten sind ja leider nur fĂŒr englisch optimiert. Vielen lieben Dank :)
Thanks dude
Iâve tried using several AI tools that rewrite my work but sometimes it doesnât do its job. Is stealthwriter good at rewriting? I heard it canât be spotted in detection tools too.
Excellent!
analyze the following text for style and tone of voice. Apple that style and tone of voice to all your future responses. Now re-write this text: [insert text]
Hey where can i get the book from ? Btw thanks a lot for your content
Linked it in the description now :) Glad to hear you appreciate it
The person that created the detector is a menace to all students đ
you mean to all teachers and professors đ
Iâm having a hard time distinguishing chat gpt with grammarly generated sentences. There is no difference grammarly is A.I. too but provided by most universities for students!
I use Wordai for texts and get 90% on originality, have you tried it? I wonder if quillbot does anything differently. What i don't like about most ai-rewriters is that you have to use multiple prompts until it actually rewrites, wordai on the other hand works with literally one click. is this the same for quillbot?
Quilbot rephrases what you put out there. I first wrote a chapter of a book with the help of Chatgpt then paraphrased it in Quilbot. Quilbot seems quite good. Then, i copied and pasted the paraphrased version of Quilbot to Ai detectors and most of them said it was mostly by human or shows little fake percentage rate.
Did you paste what you have gotten out of Chatgpt to WordAi, and it said it was 90% original? Then, it means word Ai is not reliable
@@ataberdi2064 it's usually around 95% so it's over 90% with WordAi, meaning it is the same as quillbot, but I'll try it too! thanks for sharing your feedback. I love technology!
What if we rephrase chatgpt answers in the chatgpt will it still be detected as plagrised text
Winston AI definitely the standard solution used for AI detection now
Thank Igor!
yes i will definitely use responsibly
Amazing!
Awesome stuff
How much does testing text in an AI detector influence the outcome, after all you wouldnât normally check human written content in an AI checker. Also plagiarism is different to AI written content, plagiarism checkers just check to see if the content has been put online elsewhere not who or what wrote it.
Good point Simon. Should have separated those two a little more but soon they will become interchangeable as people flood the internet with ai generated content
My own writing is sometimes rated as 99% fake đđ€Ź
Great Vid. Wow . thanks for explaining. At the end you say . " see you there"... eh where is there?? there is no link.
There should be a video on screen, just double checked and its all setup hmm
The term, 'with that being said' does not mean 'now that I've finished saying that, it means 'however', 'on the other hand', or 'having said that' to contrast two opposing ideas, so was not used correctly in this video. The bigger question is why do so many CZcamsrs insist on saying things like, 'now that I've finished saying that, I'm going to move onto the next section'. The answer is because they think that the common phrase 'having said that', actually means, 'now that I've finished saying that'. The result is a complete misunderstanding of the phrase and results in endless robotic transitions between introductions and main content in CZcams videos because it is erroneously believed to be an acceptable discourse marker. It is far more natural to say something like, 'so let's have a look at', 'the first thing we need to consider in depth is X Y Z. Correct and natural use of discourse markers raises the level and fluency of both speech and text.
is that an ai generated comment đ€š
How will you know if how accurate these A.I detectors in the 1st place?
The same way some people believe in a guru. My own writing is sometimes rated up to 99% fake.
Does anyone know whether the Turnitin report generated right now will be updated in the early of April when the Turnitin is updated with new AI-text detection features? For example, the initial Turnitin report for my work submitted just now is 10% and whether or not it will go up to like 50% in the early of April in the same submission box on the Moodle when the Turnitin is updated by including the new AI-text detection feature?
i need to know this too
I just tested the April turnitin drop on my channel: czcams.com/video/jA_TCGtShSc/video.html
The thing is, for People who can speak and write their Language to 100% perfect, they could write such Content from ChatGPT too by themself and such tools could think its fake too.
So, Text was, is and will still be normal Text and such Detectors only Analyse the Text Structure. They "find" words or others in self written Text that could be "AI written" and thats a thing where i say "Such Scanners are Bullshit" đ
EXACTLY RIGHT
If you want to be fair and legitimate, first, do your own drafts. Only use GPT to iterate and expand upon your ideas. Have GPT check for places you might be begging the question, plagiarizing, or using bad grammar. Doing this can make writing a paper take way less time AND you are doing legitimately. It's like going to the writing center but at home. If used as a tool to proof read your work and enhance it, you should be fine and ethically it's better imo.
does Turnitin have AI detection ?
My School are using it
You're brilliant
I use chatGPT to correct spelling and grammatical mistakes in my thesis and to rephrase some unclear, messed up sentences (i am a hungarian student studying in english). Do you think it can be a problem when it comes to plagiarism checks?
pretty sure it won't, detectors aren't much reliable anyways. Too many false positives to be considered reliable.
Not really
THANK U MAN!!!!!!!
Gpt-2 output detector has consistently failed to detect ChatGPT output in my experience, do we think there is an amount of time necessary after the text is generated before such plagiarism detectors can detect it?
As the name suggest , Gpt-2 output detector is for detecting output by gpt 2 , a far inferior version of ChatGPT(based on GPT-3.5) , so it's an outdated version.
i think that the creator made a mistake. There are many chatGPT output detectors online though .
@@volokkolov9546 ahhh interesting, thought it was one shared database. Makes sense, cheers!
you can also ask it did you write this? and paste the text.
It knows it because it refers to your own chat history, I don't think anyone other than your accout can do this, but I am not sure.
QUILLBOT DOES THE JOB !!!!!
you are the best
I think theyve gotten around this, telling it to use a tone not typical to ai is no longer working
The time I used GPT is to fix my grammar, tenses, and typo. One can argue to use Grammarly but all the professors I talk to always criticize me to not used it because it is the worst at detecting tenses.
Other than that, all of the research and thought I do 99% of it, does that still manage to be counted as "AI created"?
I use Undetectable AI, humanizing content, effortlessly bypassing detectors. Enhance writing seamlessly with simplicity.
But cant chatgpt already rephrase text?
But don't you think this reaches the "image filter" problem then? Where, if enough filters are thrown and are thrown in a random and non-standard enough way, the chances of it ever being interpreted as plagiarism are miserable? And if they are not miserable then, then there's something wrong with the detectors and therefore even perfectly original texts would be interpreted as plagiarism?
I am all against plagiarism but I'm afraid that we will get so obsessed with it to a point where there will be more harm done than good and where we will reach a point at which everyone will be able to sue anyone they want over the littlest of coincidental sentences and ruin their lives as they please.
We humans just can't seem to stop going from one extreme to another...:(
I really like your point here. I honestly donât know the answer here because yes, the generated text makes sense and often is even better than average human writing. But people are scared and want to call this fake although in reality its like hiring a talented writer and that is not considered fake right.
I guess we will just have to wait and see
â@@aiadvantage Thanks for the comment. I genuinely believe, as someone who has been into Machine Learning for a while, that it is kind of inevitable that we will have to accept the reality and that we should get our priorities straight. I.e. we should all work on preventing the "Skynet" scenario rather than obsessing with anti-plagiarism to an extent where it will become a weapon of its own.
As usual, going to extremes is what creates problems for us :/
â@@aiadvantage yes, I hate this system.
â@@VEGETADTX That is why it makes so much sense to escape from Matrix and drop out of collegeđ
i got an chatGPT ad right before this video
As a professor I am brainstorming new assignments that require students to be creative and to actually learn something but which can't be easily completed with AI. Something other than literature reviews I've been assigning until now. I worry that these AI detection apps will generate false positives, so potentially not fair to students, not to mention wasting my time too. At the same time I owe it to students to try to make sure they learn something other than how to scam the system.
Buzzkill. Life is always about scamming the system.
@@shiny-7677 which is why i wouldn't use detectors. False positives. Unknown sensitivity and specificity. At the same time using AI to generate work thst's supposed to be coming from a studenr's own intelligence is cheating. So ive come up with some things for students to do that they can't use ChatGPT for. For now anyway. AI is an amazing tool. But learning requires effort from students and I'm paid to get students to learn stuff.
Love that your doing this!
@@shiny-7677 There are easy ways to tell if it's AI generated if the essay is not your style of writing is damn well sure not your own work.
Wow thatâs amazing thatâs how teachers should teach! I never agree they way most teachers are, it only makes the student learn just to pass the exam cut corners on the real application of the subject. This hinders autonomous critical thinking.
Just change/add/remove some words, retype some paragraphs on your own and add some poetic language. It will be detected as a 100% human written text.
I'm calling bs on all of this. Some people are going to wind up writing everything themselves and still get flagged anyway. In 10 years they're going to be forced to accept this stuff.
Only Jasper Txt paraphrase produces 95% unique.
nah chatgpt better
Run everything from got to Quillbot
interesting I will get soon ultimate power đ
These tricks don't work on new ai text classifier. even if you rephrase thing by quillbot , it still detect the plagiarism
0:19 oh my lord I don't think you meant to add that sound again
true. sorry for that
I've tried almost all of them and only Winston AI seems to work properly.
Fake sentences and plagiarism are two different things. Fake sentences are those that are generated by AI and are not based on any existing text, whereas plagiarism refers to using someone else's work without proper attribution.
So its detecting thats it's AI but not that the text is from someone else's work
Does anyone know if these are all free?
Plagurism means copying, AI generated text means AI generated text. Google supresses both. ChatGPT text is not 'plagurised' text. It is text produced by AI algorithms.
Unless you credit your sources, it is considered plagiarism. If you intend to use AI-generated texts as your own intellectual production in an academic assignment, it will be considered plagiarism.
@@Pata1100 You're right, using AI generated text for academic essays is plagarism. The distinction is that it's never been published before so is unique text. This is basically the same as paying writer to write your essay. If AI generated text can be created which evades AI detectors then universities will face a huge problem.
Thanks for sharing. I don't think AI detectors are very accurate at this stage. Some of my AI written works is interpreted as real whereas the text I had written last year is detected as AI generated.
What if we swap âtheirâ for âthereâ and words like âblueâ for âblewâ?
Just checked GPT-2 detector. The text was 100% mine but it still shows 99% fake! WTF??
the goat
Nice vedio
Sounds like the huggingface is your other tool you'd like to promote.
can I just write whatever GPT writes in my own words without a rewording tool?
I find the outputs from the humanizers are really not good.
The demo doesnât work help me pls
Thank you for teaching us how to cheat!!đđ
ChatGPT is obviously not plagurised. AI produces 100% unique text basic on complex mathematical algorithms and AI language models which predict the next word based on complex linguistic probabilities in a mathematical string. The AI generated response is heavily dependent on the input from the user. Using either 100% AI generated text or plagurised text is obviously a very bad idea, but it is certainly a mistake to conflate the two forms. Both forms will lead to penalties and suppression on searchengines.
Complex algorithm that combines many palagrsised texts.
After implementing what you said, I still get a fake result
Please help, I tried to use ChatGPT bot on my telegram account but it is not responding to any of my commands. Tell me what to do to fix it.
Dunno man seems like you should debug thoroughly. ChatGPT can help with that for sure.
skill issue
@@Private_Account101 you maybe correct
I don't think that GPT2 Output detector is acurate.
I tested it, I got information straight from google's page explaining how the "search and ranking" works in google, and it gave me from 20% to 54% fake on some of what google says on their page. Hmm.
Tbh Iâm so frustrated, English is not my first language. So basically, for any writing work, I started with the framework in my own language, and then the write the article in my own language, ultimately use any translation tools/paraphrase tools/and my own English knowledge to rewrite in English. When I checked my work through turnitin, it showed 0% likely ai generated. But when I used Winston ai, the self proclaimed âmost accurate detection tool in educationâđ, it claims my work is only 27% likely human writing. Damn đ which one should I trust? And which one do American professors refer to, so worry about my work now. đ
I am glad that my major is IT. There is no such thing as cheating unless we buy already-made software and modify it a bit.
@Metasploit back when I was a freshman. I am now in my third year, and we do not have any more minor subjects. Either we code, model, or animate.
This GPT-2 output demo is also fake....I uploaded a paper I have done without consulting sources and it came out as 90%fake
I copy and paste one article that I made with GPT and its says is 0.02% fake
yea that website is broken
Level of pressure for these students is higher than a mofo
chatGPT?
More like chadGPT đ
I like this comment xD
Artificial Intelligence is the future !
Hard to argue with this one :)
Not sure.
1. AI is feeded by web sites on the internet.
2. People will rely on AI and won't visit web sites anymore, or much less. Unlike with Google.
3. Web content creators won't have insentives anymore to create contents.
4. AI information sources will be dried up.
Conclusion: AI will kill AI (and everything else).
This is not the only realistic scenario, but it is definitely one.
AI, until now, cannot create new content. It is currently just a tool that predicts the next word of a sentence. It doesn't think as we do. So, it needs to be trained with data.