How to Bypass TurnItIn AI Writing Detection (ChatGPT)

Video Statistics and Information

Video
Captions Word Cloud
Reddit Comments
Captions
hey guys Justin from Gold Penguin here and I want to make a video today showing how to use chat GPT and some other AI rewriting and sentence rephrasing tools to bypass turnitin AI detection so the way this is going to work is we're going to ask chat gbt on gpt4 which is the most advanced generative software right now to make us an essay and then we're going to put it through a few tools one is undetectable AI then we're going to put it through coolbot there's one called netis neetus Ai and then we've got word AI so I'm going to show you how all four of these work and then we're going to test it on a few of these AI detectors so let's go ahead and have this essay get generated and I also have not done this before so full transparency not done it before I've tested this before but I've never done it with this exact article I just made this up a few seconds ago so I I did this for the sake of saying um showing kind of the real genuine aspect of if these tools do slash don't work because they are wonky in in some regards but generally they work and the way that you have to use them is combining AI plus your own human intuition and editing to really create something that is not quote unquote AI so the reason that I kind of disagree with it with these detectors especially like turnitin is if turnitin falsely Flags somebody of using AI That's pretty detrimental and now take that and multiply it across thousands of students that might be getting falsely flagged that's pretty messed up so I think a lot of lawsuits are going to be coming out in the coming months because of this but here we go we've got our article that's getting written right now and let's go ahead and test this out so I don't have access to turnitin because I'm not an educator but I'm going to paste this into um into our detectors so we've got a few of them here we've got content at scale originality Winston and copy copy leaks content at scale I've noticed is the most like liberal of these detectors it's if it says AI you've probably used it but it's not going to just call AI on things randomly this already says highly likely to be a human so sometimes out of the box you actually get things that do work but I will say originality is probably going to say 100 AI let's go ahead and just check for AI um and so since I can't use turnitin to detect this I think a benefit of originality is it actually over diagnosis people and this tool went from being fairly good a few months ago to a little bit less reliable now because well of course this is 100 AI but I've tested this with my own essays my own articles my own blog posts anything that I've written and I've gotten 90 to 100 AI when I have not opened chat GPT once so take that as you wish these detectors really just be are based on predicting the words to the left like if I were to blindfold you and I would say finish the sentence and I say some proponents of the Electoral College and then you have to finish that sentence what would you say would you say argue that it why I can't okay would you say argue that it protects the interest of the small states you might say some variation of that but that exact sentence uh nine out of 10 times 99 out of 100 times if you train this bot which is chat GPT based on millions and billions of data sets and of textbooks it's basically like a really smart student that you're just feeding a bunch of textbooks to and you say hey based on everything you know write me what this sentence is going to be or finish it after this line more than likely it's going to say this so if you were to ask chat jbt to generate this article 10 times you're not going to see this exact sentence but you'll see very similar variations of this when human writing doesn't really work like that where humans are very creative complex and we're not perfect I think these Bots really try to become perfect not by not on purpose it's just when you're trained on a bunch of textbooks you become perfect just by definition so I have noticed that originality over predicts but uh we're gonna use this for now so we have 100 Ai and let's try to get rid of a hundred percent we'll use Winston this is another one this is really meant for educators to test things um which I have noticed is is good sometimes but not as ridiculous as originality here we go human score 37 so I think this is a bit more fair uh not to really say something is 100 unless you're really positive uh unless originality is but again I've had those issues with it but we do see we do see this this red where it says likely AI generated okay so now let's check copy leaks which I also tend to tends to flag you as AI more than not and this says human text so some of the times you'll test this stuff out and you'll see it actually like works right off the bat that's kind of why these detectors are not the most accurate and this whole kind of logic behind it is a little skewed I get some reasons you want to check for AI but there's some times that you will write human writing and get flagged as AI when you are not using AI which is kind of one of the purposes I want to show here is can we even use AI to get around this when I write my own articles and get flags sometimes so if we could turn AI into not AI that's pretty impressive so the first tool we're going to use is undetectable AI I love this tool I've used it quite a bit and they give you a few options to choose from so you could change the readability level to like kind of what your vocab is going to look like if you want a high school level of vocabulary or if we want to use a doctorate level of vocab so I'll go ahead and keep it at University level to keep it kind of general and if you want the purpose you could change this uh to whichever kind of tone you want we'll keep it on General writing for now but you have a ton of different options to choose from based on what you're doing you also could choose a more readable balanced and more human which is really more readable is going to be better sentences but highly likely High uh not highly I'm sorry more likely to get detected as AI while more human is going to add more variability and Randomness which might skew the sentences a little bit make them kind of wonky and funky which might not be the best so I'm going to keep it at balanced and we'll do a run from here let's check it for AI and it says it was written by Ai and the way that it does this is it it kind of Compares what your writing is to about a thousand or so samples on each one of these different writing tools so they've tested content at Scale based on like a bunch of different writing pieces and said okay more likely than not would this get detected with consonant scale yes and so they kind of apply that logic for all of these but yeah so let's go ahead and humanize this we've got 550 words in this article depending on how long uh your your piece of text is will depend on how long it will take to give you a new result this says about 40 seconds uh but we'll go ahead and wait it'll probably be faster what undetectable does is it rephrases your sentences and kind of rearranges how everything looks and will give you a brand new article sometimes even being a bigger word limit you'll actually get like 700 words you know in a response when you put in 550. after we use this we'll test it um we'll test it on on originality just to see what it gets out of the box but really the two methods that I want to use today are combining undetectable and quill bot together so let's go ahead and copy this and let's go into originality and let's start a new scan and let's see how it does but right off the bat we can see that we are at a 55 original which is a lot down from a hundred percent uh the Electoral College which was established in 19 in 1787 is an outdated system currently used to determine the winner of the U.S presidential elections despite its initial purpose as a compromise between Congress and the popular vote it no longer aligns with Democratic principles perfect um okay I do notice there should be a comment here so this tool is not perfect but look at that we already just reduced this by over half by not even really changing much besides putting it into undetectable which takes 10 seconds now let's show you how coolbot works so let's paste this uh oh I actually can't um do too much because I do not have a premium account uh cobot is free for not too many words and then undetectable costs I think it's also free for up to like 250 words but let's go ahead and paraphrase this and you will see that coolbot will re redo these sentences like right in front of you undetectable doesn't really give you that option they kind of just say blah like here's your new article but now let's take undetectable and let's combine that with quill bot you can go ahead and change the synonym level if we want fewer changes like let's do is there like a medium level ah okay let's do this I guess um the absolute electoral college system which was created uh you could see obsolete if it's here this means it was just added so we could see that this was new which was created it changes established to created is currently used to choose the victor of the presidential election see I don't like this uh let's say the winner there's some things here that like are just they don't make sense so you really have to like read this and see how it goes regardless of the fact that it was intended as a compromise between Congress and the people vote it no longer adheres to the values of democracy in order to guarantee this article makes its case for doing away with it uh ensuring voting rights are equal okay so let's go ahead and fix this grammar um voting rights I'm not sure if that's supposed to be Capital but for us we could see that there's some issues with commas here the way to votes cast in various States is inherently unequal because of the Electoral College so let's go ahead and paste this back into originality this could be worse than before this could be better again I have not done this specific example before but a lot of this is kind of just tweaking and seeing how things work perfect this actually is worse so straight off the bat undetectable gave you a better result than originality did I mean I'm sorry then this plus cobot did but it really depends on what you're trying to do what you're trying to bypass but you really just have to play around with it so for this specific example I'm just going to keep it with undetectable but you could use cobot if you're on longer essays and you really don't want to change as much if we change this to fewer changes and then rephrase this we could see it's going to be a bit more similar to what we had before but let's go ahead and take this undetectable text that we had originally and let's place this across these other testing tools so in consonant scale we are at 73 okay that's that's uh now let's check Winston we went from 37 human let's hope that we can be a bit higher should only take a few seconds okay we are at 79 human and we will check copy leaks which already said it was human text but we could go ahead and check it again and let's see it also says human text so you can hover over and you could see 84 likely as human um yeah isn't that weird that sometimes these things work sometimes they don't but straight off the bat undetectable did work um cobot honestly it does work for me this is one of the few times that it kind of gave me that wonky these wonky settings um really you have to mess around with it you could do simple change the synonyms but if you want to go a step further than this there are two other tools that I want to recommend today before letting you all go one of them is word AI which works the same way as undetectable we could go ahead and paste our chat GPT text into word AI I'm sorry avoid AI detection into here and you could also change that that variable if you want to change less change more change more is going to make it less likely to be detected but your sentences might be a bit weirder and not human-like so we're really hitting kind of this Paradox of these ad detection tools getting stricter on these rephrasing tools which kind of create this never-ending dog tail Chase of like really what what is AI anymore like I don't I don't even know what what's going to happen in a year from now I think a lot of problems are stemming from this but anyways back to this uh the Electoral College let's see establish 19 this is an obsolete system Paul originality originally intended to as a compromise so you could see that there are some differences between these two paragraphs and you really just have to play around with them and test them out on these detection tools if you really really want to be positively overconfident use originality to detect your AI I think it's more intense than turnitin because again I do not think turnitin is going to use or is going to falsely flag students as often as something like originality originality really loses nothing if they say something is AI they're just kind of trying to be as strict as they can but if turnitin Flags students as using add when they're not which is kind of the opposite example of this but I have seen it happen we actually have an article on turn it and flagging students at two different schools and getting in trouble that's not good um but yeah so I would say tests on originality because it's stricter I don't think turnitin will be as strict they definitely have a bit more safeguards in place I'm not saying they're not strict but I don't think they're going to be as false positive as other tools would be but go ahead and throw these into undetectable first if you're still having trouble throwing the quill bot and really only change what you need to to kind of be weird make sure the stuff is human written again this is kind of like an awkward time to be trying to write things because I don't know there's a lot of stuff going on about what is AI what is not does it even really matter but go ahead and put it through cobot through word AI after undetectable and then you could also use a Nexus AI it's kind of the same thing I'll just show you how it works same exact thing we pasted in oh okay I guess it's not working um okay well I'm not even going to use them today but yeah same kind of logic I just kind of wanted to give you as many options as possible good luck um this sucks it's also very kind of unfair I think to students and to really anyone that's writing things writers across the world that are getting flagged as AI when they're not using it and literally in in the last few minutes I showed how you can use Ai and bypass it so it's kind of a whole cluster of problems that are coming out of this but best of luck and the links to all these tools are in the description if you enjoyed please leave a like or leave a comment on your thoughts on this and really let me know how you feel because I'm very interested to hear how people across the world feel about all this stuff going on thanks for watching and have a great day
Info
Channel: Gold Penguin TV
Views: 32,234
Rating: undefined out of 5
Keywords: chatgpt, openai, bypass ai detection, turnitin, ai detection, ai paraphrasing, rewriting
Id: WUpHGg1_mKw
Channel Id: undefined
Length: 14min 47sec (887 seconds)
Published: Wed Jul 05 2023
Related Videos
Note
Please note that this website is currently a work in progress! Lots of interesting data and statistics to come.