I think the main question here is and thats why I came here: Can a professor at Uni actually let my thesis fail because some AI says it is not written by me? I strongly doubt it, because there's no way to prove that.
I agree. Also, run it through an online rephrasing service and again, it gets past this too. I’m pretty sure pressure is being applied to Open AI (esp with Microsoft) to speed up watermarking in some way. Would you chance it @david? Be interested to know.
@@TheDigitalLearningConsultant In my original text I was not referring to copied texts, I was referring to ai written texts. Okay, I've never heared of watermarking of a text and wouldn't know how that works. Especially to prevent it from being unwatermarked by another ai haha I would take my chances with letting ai write big parts of it, yes. But only because in case I actually get failed I have the money and hope in our justice system to go to a second round haha
@David text water marking can be either done with special characters that are invisible or with text patterns. just rephrase the ai text, proof for validity, and it's not traceable. I have autism and some of my texts have such a bad flow that people would believe a computer wrote it
@@Luciezka So would AI tools like this help with your written communication / articulating your thoughts more clearly? Do you use similar tools like this already? Would love to know what the benefits could be. There's a lot of scaremongering about with this
@Digital Learning Consultant With my autism its just harder to keep a proper flow so that neurotypical people have an easier time reading since many connections need further explanation, ai does not help with that. The strength in these tools is that they can be used for finding subjects and ideas more easily, but i can imagine for people using speech systems(such as steven hawkings) that it could provide a lot of help. The main reason i use AIs like github co pilot is for creating boilerplate code since its a repeating pattern. Another great ussage that this ai can have is to shed light on different view angles, especially communicational problems
The education system has always been about grades. Maybe its time to evaluate students differently. Rather than grading their essays, Interview them about their essay, ask them questions to see how well they researched. Who cares if they used chatGPT. ChatGPT is a great assisting tool. Why limit the tools they can use? "Work smarter not harder" is not a bad thing. We should encourage them to use all available tools. What matters is whether they learned or not, and how confident are they about the subject
I think this is a really valid comment Xonbiel - not sure where you are from but the UK Further Education sector has had funding cuts over the past 10 years with a reduction in contact time with students - this seriously limits the opportunity to take time to have these conversations. I’m wondering if AI could help with this too?
The amount of people that are getting in trouble because of how incredibly unreliable these AI detection tools is insane. You really need to mention this, because most of these teachers don't know anything about artificial intelligence works. To give you an example, the US constitution written in 1787 with a feather, got 89% AI. GPT Zero is notorious for giving false positives, as well as advanced detection tools such as Originality AI. Even if you can get the best detection tools, I think this is a temporary solution to a much bigger problem. Once AI becomes much more advanced, it will be near impossible to get caught.
I agree @mementomori9176, AI technology will adapt and that will for detectors to adapt and the cycle will continue. It’s already easy to navigate detectors now with some rephrase tools. Education institutions can try and shut these tools down (I believe ChatGPT is banned in Italy?) but the cat is already out of the bag. I see these generative AI tools as that, tools that we use as part of our armoury to communicate, learn and support human creativity, not replace it.
Yeah, the problem is GPT-Zero doesn't work, it's a marketing gimmick written by a guy who has tons of media connections. However there are other ways of detecting ChatGPT / A.I. generated content which are also free and exited long before GPT-Zero.
Just watched your vid Patrick - interesting and in depth. I don’t know about GPT Zero being a marketing gimmick - although I heard about it via the BBC - the test I did worked…until I ran it through an online rewriter and then didn’t pick it up as AI at all. This was my first response to seeing what we do about Chat GPT - especially when it’s not in-depth essays students are submitting. I’m using GPT as a prompt at the moment mostly - almost an idea generating tool. It was described as ‘A drunk person fumbling through a forest picking average fruit’. But it’s early days - where do you see things progressing/regressing?
@@TheDigitalLearningConsultant Also to further answer your comment - I am working on another more in depth video about the future of Large Language Models (LLM's) ... I wouldn't describe them as drunk, I would describe them as A.D.H.D. because they use a mathematical construct known as an attention mask, which...if you're not much of a math geek, you can rest fairly comfortable in knowing that, "attention mask," really is just like it sounds, it's the model's focus and attention on various key words in a text. So if the particular query you gave it happens to have attention focused on the words, "love ice cream," in the phrase, "I love icecream," it may ignore the fact that the subject was, "I" and ramble on about loving ice cream in general. Conversely if it focuses on "I / icecream," it may ramble on about your interests and sweets, less in depth about ice cream. Does that make sense? So basically being that LLM's are probabilistic in nature, they can only get so good, however research teams are constantly setting up benchmarks and trying to improve upon those benchmarks. Not even the best machine learning researchers can predict with much certainty how much better a particular benchmark will get over a year or so, but it's pretty fairly certain that there is not going to be any major explosion in intelligence, it's more of an iterative, slowly building thing. It's not the age of the singularity, as a lot of people like to think and talk about, it's the age of benchmarks. This is what I will go into in my soon to be published video when I get time to finish editing it.
@@patdel I’m subscribing so I can watch when it’s ready. I’m about to undertake and action research project looking at the potential impact of AI image generators like MidJourney will have on arts education and authenticity/authorship. You had any dealings with that? Great to connect!
@@TheDigitalLearningConsultant Yes. I would be happy to make yet another video specifically on image generation after I'm done with the one I'm doing. In short, it's done with the same underlying math, using the attention mask, but rather than paying attention to words in strings, it gets trained on word-picture couples and so when you do a query, it kind of smears out those words and pictures into a new picture. That's very high detail and it requires a bit more clarification to really get it, but that's the basis of it.
This is not accuarte at all, please don't fail your studnets because of this site. I literally use AI to generate all my content, and the site doesn't even recongize most of it. Its pretty inaccurate.
I wouldn’t fail them solely on this Tejas, but it might be another tool to flag some work. Knowing my students (as is the answer to most educational issues) is key. What other ai checking tools are you using? What would be your recommendation?
@@TheDigitalLearningConsultant this is my livelihood. I work as a content writer, and I've been using chat gpt. And I've tried every ai plagarism checker out there, it doesn't catch it. It's possible when a student structures the blog himself, and takes sections from the AI. In his own style of writing
@@TheDigitalLearningConsultant Teachers will probably fight it with everything they have, because we know they're usually a lets say...conservative group of people. I thik a right way to handle this is to use. Cause why fight it if it's very likely well have that tool in our pocket, available at any time in our future live. If there will be no need to gather information into a new text in the future, then why should we teach it?
@@TheDigitalLearningConsultant I don't it will be possible to check whether AI wrote a piece of text (especially when the AI uses NLP). Sure, you can say it's "most likely" written by AI because of certain paragraph structures, but there is absolutely no way to prove this. The best thing that teachers can do is promote the fact that people do homework and write essays for their own gain. They don't do it for the teachers, but for themselves, to improve their own skills and abilities. If a student is too lazy and doesn't want to learn and practice writing essays, it's their choice.
@@rRhyan haters gonna hate 😂😂 there’s some easy work arounds. I’m using it myself to map out a coaching assignment I’m writing! It takes out some of the prep work so I can get stuck in with the details!
Do you really @qal4real372? Which jobs do you think will be affected the most? I’m not so sure myself. At the moment, these tools still require human input, especially with deciding the quality of the responses. Without knowledge of what good/ purposeful writing actually is, we wouldn’t know if ChatGPT’s responses are any good. Also, right now, ChatGPT chooses the average of all of its sources, so throws back the average response. But this is just the start - be interesting to see where we go next!!
@@TheDigitalLearningConsultant You won't need a lot of coding skills to let a bot go through the Internet, detect what the latest trends are, then write an article on that all completely by itself. This way it won't even require human input and it would ruin at least quite a few careers when it comes to authoring articles. I think it would actually be pretty cool if there would be a website that's completely ran by AI (of course, some human filtering would be required to get rid of profanity in some cases). On the other hand, information that comes onto the internet still needs to come from somewhere, which is usually from other articles written by actual people who report on a certain thing that's actually happening.
You think so @LifeOfLevi? I think it will definitely be one part of our toolkit but I’ve heard that technology will revolutionise education so many times… “Books will soon become obsolete in public schools. Scholars will be instructed through the eye... Our school system will be completely changed inside of ten years” Thomas Edfison, 1913. What has really changed? We have calculators to help us with math problems, spellcheckers to help with English, how do you think AI will help in future?
@@TheDigitalLearningConsultant I know I’m answering your question with another question but how can you stop it? Books are obsolete and only becoming more so, the education system as a whole up until tertiary education has become more and more disconnected from todays day and age and young people are learning more from youtube videos then what a teacher can write on a chalkboard! Traditional schooling is dead and needs an overhaul… AI can help with that.
@@LifeOfLevi I don’t disagree! I think we can’t shy away from this stuff, it’s here anyway and it’s just the beginning! I just don’t see it replacing us. But maybe it will? I’ll be dead by then though. Or retired.
I think the main question here is and thats why I came here: Can a professor at Uni actually let my thesis fail because some AI says it is not written by me? I strongly doubt it, because there's no way to prove that.
I agree. Also, run it through an online rephrasing service and again, it gets past this too. I’m pretty sure pressure is being applied to Open AI (esp with Microsoft) to speed up watermarking in some way.
Would you chance it @david? Be interested to know.
@@TheDigitalLearningConsultant In my original text I was not referring to copied texts, I was referring to ai written texts.
Okay, I've never heared of watermarking of a text and wouldn't know how that works. Especially to prevent it from being unwatermarked by another ai haha
I would take my chances with letting ai write big parts of it, yes. But only because in case I actually get failed I have the money and hope in our justice system to go to a second round haha
@David text water marking can be either done with special characters that are invisible or with text patterns.
just rephrase the ai text, proof for validity, and it's not traceable. I have autism and some of my texts have such a bad flow that people would believe a computer wrote it
@@Luciezka So would AI tools like this help with your written communication / articulating your thoughts more clearly? Do you use similar tools like this already? Would love to know what the benefits could be. There's a lot of scaremongering about with this
@Digital Learning Consultant With my autism its just harder to keep a proper flow so that neurotypical people have an easier time reading since many connections need further explanation, ai does not help with that.
The strength in these tools is that they can be used for finding subjects and ideas more easily, but i can imagine for people using speech systems(such as steven hawkings) that it could provide a lot of help. The main reason i use AIs like github co pilot is for creating boilerplate code since its a repeating pattern.
Another great ussage that this ai can have is to shed light on different view angles, especially communicational problems
The education system has always been about grades. Maybe its time to evaluate students differently.
Rather than grading their essays, Interview them about their essay, ask them questions to see how well they researched. Who cares if they used chatGPT. ChatGPT is a great assisting tool. Why limit the tools they can use? "Work smarter not harder" is not a bad thing. We should encourage them to use all available tools. What matters is whether they learned or not, and how confident are they about the subject
I think this is a really valid comment Xonbiel - not sure where you are from but the UK Further Education sector has had funding cuts over the past 10 years with a reduction in contact time with students - this seriously limits the opportunity to take time to have these conversations. I’m wondering if AI could help with this too?
i don't like submitting a essay without checking for plagiarism thx for this
The amount of people that are getting in trouble because of how incredibly unreliable these AI detection tools is insane. You really need to mention this, because most of these teachers don't know anything about artificial intelligence works. To give you an example, the US constitution written in 1787 with a feather, got 89% AI. GPT Zero is notorious for giving false positives, as well as advanced detection tools such as Originality AI. Even if you can get the best detection tools, I think this is a temporary solution to a much bigger problem. Once AI becomes much more advanced, it will be near impossible to get caught.
I agree @mementomori9176, AI technology will adapt and that will for detectors to adapt and the cycle will continue. It’s already easy to navigate detectors now with some rephrase tools. Education institutions can try and shut these tools down (I believe ChatGPT is banned in Italy?) but the cat is already out of the bag. I see these generative AI tools as that, tools that we use as part of our armoury to communicate, learn and support human creativity, not replace it.
This doesn't work. I pasted a test paper completely written by GPT 3.5 and it came back as "completely human". Time to re-write the code or update.
What is another alternative for the Turnitin plagiraism checker? Like a free and reliable one?
Yeah, the problem is GPT-Zero doesn't work, it's a marketing gimmick written by a guy who has tons of media connections. However there are other ways of detecting ChatGPT / A.I. generated content which are also free and exited long before GPT-Zero.
Just watched your vid Patrick - interesting and in depth. I don’t know about GPT Zero being a marketing gimmick - although I heard about it via the BBC - the test I did worked…until I ran it through an online rewriter and then didn’t pick it up as AI at all. This was my first response to seeing what we do about Chat GPT - especially when it’s not in-depth essays students are submitting. I’m using GPT as a prompt at the moment mostly - almost an idea generating tool. It was described as ‘A drunk person fumbling through a forest picking average fruit’. But it’s early days - where do you see things progressing/regressing?
@@TheDigitalLearningConsultant Also to further answer your comment - I am working on another more in depth video about the future of Large Language Models (LLM's) ... I wouldn't describe them as drunk, I would describe them as A.D.H.D. because they use a mathematical construct known as an attention mask, which...if you're not much of a math geek, you can rest fairly comfortable in knowing that, "attention mask," really is just like it sounds, it's the model's focus and attention on various key words in a text. So if the particular query you gave it happens to have attention focused on the words, "love ice cream," in the phrase, "I love icecream," it may ignore the fact that the subject was, "I" and ramble on about loving ice cream in general. Conversely if it focuses on "I / icecream," it may ramble on about your interests and sweets, less in depth about ice cream. Does that make sense? So basically being that LLM's are probabilistic in nature, they can only get so good, however research teams are constantly setting up benchmarks and trying to improve upon those benchmarks. Not even the best machine learning researchers can predict with much certainty how much better a particular benchmark will get over a year or so, but it's pretty fairly certain that there is not going to be any major explosion in intelligence, it's more of an iterative, slowly building thing. It's not the age of the singularity, as a lot of people like to think and talk about, it's the age of benchmarks. This is what I will go into in my soon to be published video when I get time to finish editing it.
@@patdel I’m subscribing so I can watch when it’s ready. I’m about to undertake and action research project looking at the potential impact of AI image generators like MidJourney will have on arts education and authenticity/authorship. You had any dealings with that?
Great to connect!
@@TheDigitalLearningConsultant Yes. I would be happy to make yet another video specifically on image generation after I'm done with the one I'm doing. In short, it's done with the same underlying math, using the attention mask, but rather than paying attention to words in strings, it gets trained on word-picture couples and so when you do a query, it kind of smears out those words and pictures into a new picture. That's very high detail and it requires a bit more clarification to really get it, but that's the basis of it.
This is not accuarte at all, please don't fail your studnets because of this site. I literally use AI to generate all my content, and the site doesn't even recongize most of it. Its pretty inaccurate.
I wouldn’t fail them solely on this Tejas, but it might be another tool to flag some work. Knowing my students (as is the answer to most educational issues) is key. What other ai checking tools are you using? What would be your recommendation?
@@TheDigitalLearningConsultant this is my livelihood. I work as a content writer, and I've been using chat gpt. And I've tried every ai plagarism checker out there, it doesn't catch it. It's possible when a student structures the blog himself, and takes sections from the AI. In his own style of writing
@@nullprophet10 great job!!!! I’m finding Chat GPT a great jumping off point!
The education system as we know it is totally screwed.
How do you think we'll adapt/respond? Do you reckon we'll incorporate it into education as a tool, or try and fight it with AI tools?
@@TheDigitalLearningConsultant Teachers will probably fight it with everything they have, because we know they're usually a lets say...conservative group of people. I thik a right way to handle this is to use. Cause why fight it if it's very likely well have that tool in our pocket, available at any time in our future live. If there will be no need to gather information into a new text in the future, then why should we teach it?
@@TheDigitalLearningConsultant I don't it will be possible to check whether AI wrote a piece of text (especially when the AI uses NLP). Sure, you can say it's "most likely" written by AI because of certain paragraph structures, but there is absolutely no way to prove this. The best thing that teachers can do is promote the fact that people do homework and write essays for their own gain. They don't do it for the teachers, but for themselves, to improve their own skills and abilities. If a student is too lazy and doesn't want to learn and practice writing essays, it's their choice.
im a student lmaoo
L video
Not sure what that means PAKS? I’m old! Even googled it…
@@TheDigitalLearningConsultant Probably a student using AI to write their thesis haha. They basically thumbing down your video
@@rRhyan haters gonna hate 😂😂 there’s some easy work arounds. I’m using it myself to map out a coaching assignment I’m writing! It takes out some of the prep work so I can get stuck in with the details!
i already see soo many jobs being lost to this ai in the near future 😂
Do you really @qal4real372? Which jobs do you think will be affected the most? I’m not so sure myself. At the moment, these tools still require human input, especially with deciding the quality of the responses. Without knowledge of what good/ purposeful writing actually is, we wouldn’t know if ChatGPT’s responses are any good. Also, right now, ChatGPT chooses the average of all of its sources, so throws back the average response. But this is just the start - be interesting to see where we go next!!
@@TheDigitalLearningConsultant You won't need a lot of coding skills to let a bot go through the Internet, detect what the latest trends are, then write an article on that all completely by itself. This way it won't even require human input and it would ruin at least quite a few careers when it comes to authoring articles. I think it would actually be pretty cool if there would be a website that's completely ran by AI (of course, some human filtering would be required to get rid of profanity in some cases). On the other hand, information that comes onto the internet still needs to come from somewhere, which is usually from other articles written by actual people who report on a certain thing that's actually happening.
You are fighting a losing battle AI is the future of education.
You think so @LifeOfLevi? I think it will definitely be one part of our toolkit but I’ve heard that technology will revolutionise education so many times… “Books will soon become obsolete in public schools. Scholars will be instructed through the eye... Our school system will be completely changed inside of ten years” Thomas Edfison, 1913. What has really changed?
We have calculators to help us with math problems, spellcheckers to help with English, how do you think AI will help in future?
@@TheDigitalLearningConsultant I know I’m answering your question with another question but how can you stop it? Books are obsolete and only becoming more so, the education system as a whole up until tertiary education has become more and more disconnected from todays day and age and young people are learning more from youtube videos then what a teacher can write on a chalkboard! Traditional schooling is dead and needs an overhaul… AI can help with that.
@@LifeOfLevi I don’t disagree! I think we can’t shy away from this stuff, it’s here anyway and it’s just the beginning! I just don’t see it replacing us.
But maybe it will?
I’ll be dead by then though. Or retired.
@@TheDigitalLearningConsultant jst make that retirement thats all that matters