Plagiarism

Speaker A: 00:00:00

Hello and welcome to the Cognitive Engineering Podcast produced by me, Fraser McGruer, for Aleph. In this series of podcasts, we take a look at interesting topics and discuss what we think they tell us about analysis and decision making. I'm here with Jordan Fermanis and Nick Hare of Aleph, and this week we're discussing plagiarism. Nick, go for it.

Speaker B: 00:00:25

Yeah, well, the reason we're discussing this actually is because it's topical. So you may have heard tell of a lady called Claudine Gay, who was president of one of these new American universities called Harvard between July last year and January this year. Okay. Wow, that was brief. Yeah, well, so I don't know how often you're meant to be, how long you're meant to stay as president of Harvard. But anyway, she was criticised for not having sort of adequately condemned the October the 7th attacks in Israel. And then I think for, she also gave evidence to Congress, there were criticisms about this and the way that they'd handled sort of anti-Israel protests. And people, I think, perceived that there was this sort of hypocrisy about the way that, you know, they were extremely intolerant of criticism of all, anything except, you know, when it comes to Israel, when apparently that's, it's fine, that's freedom of expression, but other kinds of freedom of expression, they're very keen on policing. I think that was the perception. Anyway, it led to people sort of starting to scrutinise her academic record, and it didn't seem to be hugely impressive. I mean, in terms of volume, anyway. And then a sort of right wing activist, well, conservative activist called Christopher Ruffo and a journalist called Alan Siberian, went through her work and found what they claim is about 50 examples of plagiarism. Right, so some of these are like, well, I've looked at some of the examples, to be honest, I wouldn't personally get too worked up about them. If something someone wrote had these in them. We've got to remember, she's the president of Harvard, and they have very strict standards. These are standards that they would apply. These would fail the standards that they apply to their students. But I mean, none of them, it is sort of rewordings. It's, you know, so I mean, this is, say, for example, there was a paper by Palmquist and Voss, which she plagiarised from, and it contains, this is one sign that the data contain little aggregation bias if racial turnout rates change depending upon a precinct's racial mix, blah, blah, blah. And she has written, this is one sign that the data contain little aggregation bias if racial turnout rates change depending on a precinct's racial mix, dot, dot, dot. So it's a direct quote. And there's that kind of thing. There's also sort of discussion of case studies and things. And the big issue is none of these are attributed, right? Obviously, she's not quoting them, but she's directly used their words. So I think the thing is that, you know, not by themselves, I think, I would sort of, I would think, well, this wouldn't be enough to, you know, to ruin someone's career. But in her case, I think she, you know, there is an argument that she should be upholding, should be held to a much higher standard. Anyway, before this could all be resolved, was it plagiarism? Was she going to get fired? She resigned, thus depriving us all of an exciting, dramatic, and controversial case. As you know, we like to delicately sidestep cultural war issues. So we're going to focus on what is probably the most incidental aspect of this, which is the plagiarism, which is the plagiarism of a question. So because I think, you know, really, the plagiarism is a stalking horse, I think, for trying to get rid of someone that they see as plagiaristic. Yeah, there's other stuff going on there, right? Yeah. So what is it? Plagiarism? Is it a binary thing like yes or no? Or is it continuum? Is it can something be more or less plagiarised? You know, does it have to be intentional to be plagiarism? Is it, you know, is it necessarily bad? How do you spot it? Those kinds of things. And then I think we probably want to touch on this very pressing question of whether generative AI is plagiarism. So

Speaker A: 00:04:15

yeah, lots of questions there. Yeah, I think this is fascinating. I think we can have a good go at this. Because it immediately sparks lots of questions in my mind and certain fields. It's, you know, how does one not plagiarise? Well, you know, standing on the shoulders of giants is one thing and, you know, flat out plagiarism is another. Yeah, you know. So how are we going to go about this? I mean, it feels like to me that we should be doing some definitions, defining some of this stuff, some of the parameters and so on. And then maybe later on, we'll look into the AI stuff that you've talked about, Nick. Jordan, lead us in a bit. Well, yeah, the immediate thing

Speaker C: 00:04:54

that jumped out at me was this quote that I remembered from somewhere. It's not attributable. Talent borrows, genius steals. And I think that it was connected to the creative arts in some way. And I think it reminded me that like, in music or in art, people are often borrowing techniques, styles from each other. And it's not really considered plagiarism. It's almost like a homage to the originator in some cases. And certainly not grounds to be extradited from the in group or something like that. But in academia, like in this case that Nick raised, it's a crime, almost punishable by sacking someone from a prestigious job. So I think the immediate thing that jumped out at me was why we have this discrepancy between plagiarism being seen very seriously in some cases, and then in other contexts, not so much.

Speaker A: 00:05:55

Yeah. So I guess we need to talk about context. And I guess in just briefly, just thinking about, let's say in the arts, it's difficult to fire someone because often you're not doing a job, as it where you're not employed. However, you might get sued. And that definitely happens, right? So one, you might get fired and or sued, and the other you get sued. But anyway, sorry, keep going. Okay, really? Okay. All right. So therefore, okay, Nick takes down a bit further.

Speaker B: 00:06:20

Yeah, no, but I mean, I think the first thing is, I think that you absolutely can have plagiarism in the arts. And that's a thing, you know, visual arts, as well as, you know, literary arts. If you look at the things that people, that tend to appear in definitions of plagiarism, and bear in mind, obviously, there's no, well, the law, there isn't really a law about plagiarism. So it's institutional, it's like different institutions have different definitions. And so I had a look at Wikipedia's got a list of like various academic institutions from the US and what their definitions are. Princeton has the deliberate use of someone else's language ideas or other original not common knowledge material without acknowledging its source. So they've used they've said deliberate, and they've said ideas, then Stanford have the use without giving reasonable appropriate credit to another person's original work, whether such work is made up of code formulas, ideas, language. So they're talking about work, not ideas, which is interesting. And so looking at it all, and trying to squinting a bit, I think the key elements are use of something that someone else has created or invented, right? So whether that's work or words or ideas. But it can't be it can't be something that's, you know, if you like, well, I guess you'd call it prior art in different contexts, but something that is known, right? Does this come under intellectual property? Well, I think it's related to that, but it's not I mean, so we'll get on to the legal side of it in a bit. But the key, I think the key things are like without attribution, and an implication that it's deliberate. And I think that's really important here. Because I think it's the fact that you are doing it, knowing that you're nicking someone else's ideas and trying to pass them off as your own. So it which it makes it more of an intent thing, I think, an intent and personal knowledge thing. And I think that's the essence of it. I think where Jordan, I mean, Jordan was talking about this problem of, you know, if I'm inspired by something, and I, and I think that, you know, that impressionism is cool, I'll become an impressionist. That, that, you know, I'm being upfront about that. I saw Monet, I liked it, I copied that style. And here's some paintings that I've done. This is more along the lines of I see, you know, I go to Monet's house, and he's not famous. And he paints this lovely picture. And I go, oh, that's really nice. Can I borrow it? And then I put it on an easel and claim I painted it. You know, that's, that's the that's what I think we're talking about here is a situation where I am deliberately, I know in full knowledge that this isn't my own work, and I'm pretending it is. So that's that I think that's really important. Because it means it isn't so much the act, it's the intent. And I think that can condition a lot of other stuff that we look at.

Speaker C: 00:09:09

But then what about things like paraphrasing? And, you know, like, because that example used with Claudine Gay, it's like she didn't, it wasn't so clear cut, right? She, as you said, she reworded things that appeared in other people's work. So it's not like, it's not like

Speaker B: 00:09:24

d, Got Wild and Got a Life in: 2006

Speaker A: 00:11:39

And I think also in her case, it sounds like, from what you've said, that it's not like it was whole chunks. It sounds like a sentence or two here or there.

Speaker B: 00:11:47

Yeah, let's say as part of your argument, you want to talk about a study that someone did or a kind of theory that someone has. And so, you know, one way of doing that is put a block quote in from them and say, as they say in this paper, here it is. But she went down the approach of like taking paragraphs, slightly rewording them and just fitting them into her text as

Speaker A: 00:12:08

though she'd written them. So I've got a few questions, but I don't know if it's too early to ask these. But I mean, one is I think context and, you know, when we're talking about academia versus, let's say, just for a moment, music, let's say, right. I think that's interesting exploring the differences. But also, how do you avoid it sometimes? And what's the line between an intention and, you know, doing it by accident? And I think it's a bit hazier than we might expect. And then just finally, does it matter? Does it really matter? And why does it matter

Speaker B: 00:12:44

if it does? Yeah. Well, I've got a few thoughts about the entropy of different sort of English, which is one way to look at this. What's the probability of accidentally copying someone else's thought, right? So the average entropy of an English word is... What do you mean by entropy in this context? Sorry, I'm assuming everyone out there knows about information theory. Well, I know about entropy a bit, but I don't really know about entropy as it pertains to this. It's basically how surprising something is. So if you think about, like, if you and I are both choosing a word, it's sort of how many options we have to choose from and how probable it is. So, like, what's the entropy of the first word of a sentence? It's actually quite low, because it's very likely to be the or a, for example. So it's much lower. But if I say, what's the entropy of the third word? Well, it could be contrapuntalism, you know? But rather excitingly, I can see how this ties in with AI later, but keep going. Right. So if you have 10 words in theory, absolutely, the lowest probability you could reasonably claim that this was would be about 1 in 10 to the 36 that two people would... Essentially, it's like you and me picking the same grain of sand twice each, you know, out of the entire world. It's absolutely minuscule. So, like, I think it isn't really possible, certainly with language, with a sentence like the one that Christine Gay put together. You're going to be banked to rights. Even if you had the thought, even if the thought was functionally identical, the probability of you expressing it the same way in 10 consecutive words is very low. But then there are cases where it really does look like accidental plagiarism. So, you know, the very famous case of George Harrison, My Sweet Lord, great song, was judged to have subconsciously plagiarised He's So Fine by the Chiffons, which also a great song. There's no doubt putting them side by side that they're really exactly the same tune. You could call My Sweet Lord a cover, I think. But George Harrison, very nice guy. I don't believe he realised. But, you know, that really is straying into Jordan's zone here. Like, you know, it's not that difficult to come up with a chord structure and a rhythm and stuff. And if you've heard it before, you know the story about Paul McCartney when he wrote Yesterday. It felt so complete and perfect to him as a song that he couldn't believe he'd written it and apparently spent weeks asking people, what's this song? I think I might have copied it from somewhere. And of course, well, George Harrison's slightly less thorough with his due diligence. And then there's a case of Doris Kearns Goodwin, who wrote a book called The Fitzgeralds and the Kennedys, who said she unintentionally plagiarised from various sources. But she said in that case it was, and actually Jane Goodall did something similar in a book called Seeds of Hope. But apparently that's, and I kind of believe this kind of dodgy note-taking. Like, you know, if you're taking notes, you're reading through a book, you write down, you might write down a direct quote and then later think that that was a summary that you've done or something like that. So I think it's possible, it's possible not necessarily because, you know, you are picking a chunk and put it, but it's just through various processes or forgetting where you've got something. You know, like, in order to know you haven't plagiarised something, you'd have to remember everything you'd ever read and you can't. So there we are. I think this thing about deliberateness is interesting, but essentially there is a kind of, there is a grey area, even though it feels to me like, well, plagiarism has to be deliberate to be bad, to be naughty. Like, I don't have a downer on George

Speaker A: 00:16:38

Harrison. Quite. I want to bring in Jordan, but just before I do, this does tie into, I think, why it matters or not, because, and it is somehow tied in with intention, which is that, I think it can be a laziness. And I remember when I've either plagiarised probably in an academic sense or got close to it, it's a kind of laziness where you're doing your note taking and it's to do with cognition. And that's why it matters, because sometimes it's just easy, you can drift into taking notes to go, oh, it's just easy to write down what they've got here. And then at the next stage, if you're producing an essay or even with an exam, you go, and to what degree do you know, actually, this is actually exactly what they wrote and I'm writing it. And you either know that or maybe you don't. But that's why it's important maybe because of the cognition and this, because it's about sort of new, genuine research or thought, let's say, which if you're president of Harvard, setting an example is definitely important. But also it's what the audience or the, they get from it is somehow, is there a relation between originality and value to audience as well? Anyway, Jordan?

Speaker C: 00:17:45

I think the other thing that's interesting about it is it's often difficult to separate the person from the plagiarism. So like, because it's a slippery concept, it can be used as a bit of a tool. So like in the Guardian article about Claudine Gay, they say that like, um, had she been an undergrad and committed this kind of plagiarism, she probably would have got quite a light punishment. But because she's the president of Harvard, her job is now at stake for this kind of this incident. And so there's an interesting question about whether like, is the kind of, is the level of, is the severity connected to the position of the person that's, that's doing the plagiarism? You know, like, do the stakes kind of go up, depending on your status within the academic circles that you're in? Or, you know, like that

Speaker B: 00:18:41

kind of? Well, I don't, yeah, I mean, but I feel like that's true of anything. It's not a plagiarism. I mean, it's like, you know, if a man on the street, you know, fails to, doesn't do anything when an old lady falls over. That's one thing. But if the Archbishop of Canterbury doesn't help out, then you're likely to think, hang on. So I think it's fair, it's fair enough to hold somebody who's president of Harvard to higher standards. I mean, it is. Unless those standards are actually unattainable. I mean, obviously, that's, that's the question here is like, are we demanding an unreasonable level of robustness about plagiarism? Yeah. It's a funny one, because

Speaker A: 00:19:20

she looks bang to rights. She's meant to be an exemplar. And so therefore, on that level, it does matter. However, it sounds like the way the motivations and the person who unearthed this, I doubt they give a toss about that stuff. We know it's really about something else. And had it been not discovered, what effect would have it had on Harvard itself and the work that's produced there at undergraduate, graduate level, and elsewhere in the world? Soddle probably. And it's

Speaker C: 00:19:47

a historic case. It's not, it's plagiarism she committed during her PhD, not while she was president of Harvard. To be fair, it's because she hasn't produced any research, maybe, you know, but you know, like, she didn't she didn't commit this act as president. She committed it as like

Speaker B: 00:20:02

just an ordinary PhD student. Yeah, I mean, she hasn't, she's not getting her PhD stripped off, I don't think or anything. I don't I, after she resigned, I haven't really, I haven't really followed up on what she's up to. But, but I mean, there was certainly no appetite to strip Martin Luther King's PhD off him, for example. And I think, you know, the agreement would was, well, let's just all pretend this didn't happen. Yeah, you know, because the negative consequences of doing that, especially posthumously, were sort of, yeah, so yeah, who you are definitely matters whether you're going to get but I but I suspect that is a judgment that again, I would think comes down to some sort of judgment about intent. So, you know, I, as I said, I believe George Harrison, because I happen to know he's a really nice guy and has high integrity. With Christine Gay, I'm not so sure. I'm not sure I can. I mean, for all I know, she could be one of these extremely ambitious, you know, conniving, self, you know, self promoting academics who, you know, are quite willing to nick other people's stuff and pass them off as I don't know, I haven't, I haven't seen enough evidence either way. Yeah. Um, where do we go on our discussion? AI? Well, I've got

Speaker A: 00:21:13

to talk about the really interesting questions. Let's talk about the law. And then let's get onto

Speaker B: 00:21:18

this IA thing. Yeah. Yeah. So if we look at the law, it's interesting, because the law actually, you can see it's trying to skirt around some of the same issues that we've just been discussing. So just right from the off by itself, plagiarism is not a criminal offence or anything like that in the UK. You know, it's governed by, as I said, by institutions. So you know, if you're an institution who can give out PhDs, you can perfectly, you're perfectly entitled to take them off people. If, you know, if you've got a policy against plagiarism. Obviously, we do have copyright law. And that is, that's to protect your original works of authorship, right, which can be dramatic or literary or musical or whatever. But it protects the expression of the ideas rather than the ideas. Okay, so you can express it, you can express this thought in some words, and I'm allowed to then express that thought in my words, but I'm not allowed to use your words. Right? I mean, other than there are some exceptions. So for example, I'm allowed to use copyright, that's copyright, copyright, fair use, right, which means I'm allowed to talk, I'm allowed to print a paragraph of your work to go on about how rubbish it is. In the US, they also have a thing called statutory damages, which is like you get you're going to get some damages anyway, regardless of whether you've caused loss. So that's that's copyright is the specific expression. And that would be true of, say, a painting, like if I took a photograph of your painting, and then exhibited it and made money from exhibiting it, obviously, you'd be entitled to sue me for copyright breach. So patents and trademarks, which is more on the ideas front. This, which is, you know, obviously, that's the it's kind of copying the ideas, I think, which is a tricky bit here. What do you need for it to get a patent? It's got to be novel, which means it can't be something someone's doing already, even if it's fairly cutting edge. There has to be an inventive step involved, apparently. So in other words, it can't be blindly obvious. It can't be you can't sort of say, well, I've invented a new way of screwing a screw into a table or something. But you know, because anyone could have spotted that. It has to be used first of an industrial applicability. So artworks are not entitled to get patents. And it doesn't apply to anything like scientific theories or mathematical methods or anything like that, games, for example. And you're apparently not allowed to patent things that are immoral. So you can't have apparently you can't create like a patent was rejected, I think on about creating human animal hybrids, for example. Yeah. So So anyway, I mean, I mean, I think that's that's interesting because they've tried to wrestle with this idea of what is the idea that you own? And what is it that I would be stealing if I stole an idea? And I think I so I mean, you know, that's that's that's a way of a way of trying to do. Of course, all that does is punt all of those difficult questions down the road a bit, because, you know, what is what is an idea? What is an idea? Ultimately, you know, how do you prove novelty? How do you prove invention? And, you know, how closely is an idea and its and its expression related? You know, how much of Shakespeare's value relies in the expression or the idea? And it feels like it makes sense to draw a distinction between the two. But where is that distinction? What's the difference between inspiration and copying something we, you know, we discussed? And then there's this whole question of, you know, is it what's in between? If we can't ever prove it, because it involves in some sense, intent, you know, then it becomes difficult to sort of is the act and the intent the same or, you know, so there we are. Anyway, lots of interesting, I don't think we haven't got time to tackle probably most of those. But yeah, it's interesting that, you know, there's a lot of philosophy embedded in our legal concepts, which relate to, you know, stealing ideas, and so on, which I think are really difficult, interesting, and left very much to the judgment of individuals. You know, the law simply doesn't try to define these things, specifically, because it knows it can't.

Speaker A: 00:25:27

I mean, this feels we're, we're close to that podcast, the episode we did on the Romans, thinking about the Romans, and we talked about what is an idea? You wouldn't think these two things are close, but I think they are. And also, I mean, as you know, I'm no lawyer. But I think this is one of the things that I think that's the basic principle of law, isn't it? It's trying to, it's the world is messy, right? And vague, and it's just trying for our own sakes, to apply some kind of rigor or rules to that, which is not always, it's not, well, it's not possible, necessarily. Okay, there's some stuff I want to go on to, but Jordan?

Speaker C: 00:26:04

Yeah, well, I think there's an interesting question about how all this will be affected by generative AI. Yeah, recently, at one of our company meetings, we had a discussion about the adoption of large language models into workflows and stuff, like, should we signpost the fact that we're using large language models to our customers, because that might indicate some kind of expertise or, yeah, level of experience that people we work with, or the people that we're offering services to might not have? Or if we did that, would that be some kind of admission of plagiarism, I suppose? Are we taking outputs from the model and repackaging

Speaker B: 00:26:47

them and passing them off as our own work, for example? Yeah, is it like, you know, should we have chatGPT as a co-author on the things we write? Or is that as ludicrous as making Microsoft Word a co-author, because you've used them to write a document?

Speaker A: 00:27:05

Yeah, which also immediately makes me think, so I want to come back to you, Jordan, that, you know, I've started exploring using chatGPT myself for some of my professional stuff. And I don't think I'm very good at using that tool. And it's talking about, so we start getting to levels of expertise with actually using these tools, which then suggests to me that you've got a sort of a right, you're entitled to use it, perhaps. Because you know how to do something that someone else doesn't with this tool. I don't think I've explained that very well. But anyway, Jordan, keep going.

Speaker C: 00:27:38

Well, yeah, I just think it's a kind of interesting question about whether people, like, is it more ethical to plagiarise a human or an AI? Is using chatGPT in itself? I think Noam Chomsky, I came across, he called it, he called just using chatGPT high-tech plagiarism. Because obviously, like, the words are coming from somewhere, they're coming from the training data, I suppose, was his point. So even though they might be arranged in a different way.

Speaker B: 00:28:08

Well, I think that is the key question here. You know, what it would, I don't think it's correct to describe what generative AI is doing is, as a kind of, you know, it's not really producing a collage of existing stuff, people might, that's naively how you what you might expect it to be doing. So you know, when you ask it to produce a painting of a dog in a, you know, with a starry sky in the background, you might think it is taking a picture of a dog from some database, and a picture of a starry sky from a database and sort of melding them together. In which case, you could say, well, that's clearly, that is very much plagiarism. It's got all of, you know, you might think, well, it's got all of this stuff. And it just retrieves the relevant bits and strings them together in such a way that, you know, so that would be like a plagiarism machine, a thing that just creates new things from pieces that exist. But it doesn't, it's much more interesting than that. It is trained to identify patterns and connections in existing works of art, and text. So that, you know, for example, you know, if a piece of text starts sounding a bit like Shakespeare, you know, it knows that that sentence is probably going to carry on like Shakespeare, for instance, even if it's talking about, you know, bananas or some other thing that Shakespeare didn't know about, you know, he knows the thing will know that there are features in Shakespeare to make it sound like Shakespeare. And it understands how those features interact with, you know, whatever else has been written. Likewise, with bits of art, you know, what is, what makes a van Gogh look like a van Gogh is going to be features of the relationships between the pixels in that image, you know, which are really complex. And we perceive intuitively, but we don't, we don't, you know, it has managed to identify those, you know, using essentially mathematical techniques. So what you have when you've trained it, and it's trained on everything ever written, stroke everything ever, every image out there and every text description of an image, right, but what you get is not a database of images, there is no image inside, you know, a generative art generator, there's no text inside, you know, there aren't blocks of text and sentences inside ChatGPT, that's just not how it works. What it does is it knows how to generate new text. But it doesn't, it's not, it's not getting that from anywhere other than a text. So plagiarism is impossible with AI? Right, I think what we're looking at is very much like the state of a, it's George Harrison, I think, in the sense that if it does plagiarize stuff, it's got no means of knowing that you can't look at all of the billions and billions of different numbers inside ChatGPT and say, that bit is coming from Shakespeare, that bit is, you know, Herbert Marcuse, you can't, you can't, that is just not what's there. What it is, is a whole set of, you know, of parameters that describe the relationship between one word and another in sentences. And so it might well, if I were to start by writing something that was a direct quote from somewhere, it might well finish that quote off, because, you know, actually that text comes after that text, but it doesn't know it's doing, it's got no means of knowing, and we've got no means of inspecting it to find out whether it's plagiarizing stuff. And it's very much, I think, like George Harrison listened to thousands and thousands of songs and came up with a new song, which turned out to be one he'd heard before, but he didn't know.

Speaker A: 00:31:45

Yeah, because pre-AI in all sorts of fields, we shouldn't necessarily mistake pattern with intention, right?

Speaker C: 00:31:51

But I think the issue is attribution, because it's, because there's, I think there's some high profile case in the US where the New York Times, I think it was, was suing OpenAI, because they used New York Times articles to train the model, but they hadn't, I think they'd asked or paid for those articles, they hadn't attributed the New York Times authors to them. So like, that kind of is stealing.

Speaker B: 00:32:18

Sorry, where were the, how were the- But you read articles in the New York Times, you think about things, and then you write things. Sometimes those things cover the same material that you read. I mean, that's, is that plagiarism? But to access them, well, no, not if you're a subscriber and you've paid for them. Well, okay, okay, forget, that's a separate issue, I think, whether they're accessing things they shouldn't have been allowed to access. But the fact that their model is partly trained on New York Times data is no different to you writing something after having read some articles in the New York Times.

Speaker A: 00:32:46

Sorry, was the issue here, I slightly missed what you were saying, that some information, some articles were put out there that had been AI generated from New York Times articles, but were not connected to New York Times, is that right? Or was it from, yeah, sorry.

Speaker C: 00:33:00

No, like, I think the issue was the New York Times were saying that OpenAI had trained their model using some of their content, but they hadn't paid for that content.

Speaker A: 00:33:10

Understood, okay.

Speaker C: 00:33:11

So they were basically just, they were making their AI aware of all the stuff that was published in the New York Times without paying for it. Right. Yeah, I see.

Speaker B: 00:33:18

Okay, so I mean, and that is, that's a, I think that's a different, I mean, I think that's interesting and relevant, but it's a different thing, that is, I mean, so the issue, let's just take a more simpler example of someone who, where, you know, something they've written and it's free and it's just openly accessible is used to train. And that's very much the case with lots of illustrators and artists, you know, who publish all this stuff online, the image generators are trained on it in much the same way that you are entitled as an artist to look at it and then paint your own painting, and then image generators are churning out artwork that in some cases is going to look very like it, you know, particularly if you use the artist name, you can say, well, I want one that looks like Van Gogh, you know, and it'll make a Van Gogh for you.

Speaker A: 00:34:06

Fascinating. Yeah. Yeah. There's lots of stuff I want to ask and talk about, but we can't, we just don't have the time. However, I've got time for a quick question. Go on. And actually, this is going to sound dumb, but it's not dumb. Good. And it's my usual questions, but I think it's very interesting. Okay, let's have it, we'll be the judge of that. Um, either personally or just out in the world, best example of plagiarism, worst example, you can choose one or the other or both. I don't mind kicking off because I've got some ideas. I've got some thoughts around this.

Speaker B: 00:34:37

But what we've got to do is predict whether he's a victim or a perpetrator. I'm going to go with perpetrator. Yeah, you're right. I think basically there's two types of people, there's victims of plagiarism and there's perpetrators, and I think we all know which one you're on.

Speaker A: 00:34:51

Yeah, the worst one. I might edit this out. I don't know. But so this is my personal worst one. I've got a good one as well. But years ago, I was freelance. You always think I used to work for the EIU, right? The Economic Intelligence Union. Yes, because you claimed. No, no, it's because I used to work for a company that did similar things. Right. And there was my downfall.

Speaker B: 00:35:09

Was it Dun & Bradstreet?

Speaker A: 00:35:10

Yeah, it was D&B, Country Risk Services, who doesn't exist anymore. But what happened, I was there for a while and then I went freelance, which involved writing really lengthy analytical reports over like dozens of pages, which as you can probably guess, like, why the hell was I doing that job? Yeah, yeah, yeah, that's not your bag at all. Yeah, especially as an economist. And so I was churning out stuff for them. And then one day I got a letter from some editor that I'd never met saying, Fraser, we've come across some stuff in your report that looks like it's come from somewhere else. And here's some examples where it's basically text from some EIU report.

Speaker B: 00:35:48

They got you bang to rights. They got me bang to rights.

Speaker A: 00:35:50

But what they really were not happy about was not only were the words the same, but my forecast was the same, like the numbers, the data.

Speaker B: 00:35:57

Yeah, well, that would. But then, you know, at the same time, two equally brilliant, the EIU and Fraser McGruer. Yeah. Happened to generate the same forecast. That just makes it more plausible.

Speaker A: 00:36:05

Well, what's interesting is I know others later or earlier freelancers got brought up on exactly the same things. And as you know, I'm an honest guy, apart from when it comes to writing economic reports. And I said, yeah, no, you got me. You're right. And they said, as long as there's nothing and actually other economists, freelancers being called up on the same thing. And they use your argument saying, no, no, absolutely not. I came up with this. Maybe they copied me. And they said, that's fine. No worries, as long as you let us know. Is there anything else in the report we should know about? I said, no, it's all fine. It's all fine. And then you can guess what happened. Actually, it was loads of rest of the report when they looked into it. So I lost my, I got fired from my job as a freelancer, lost that.

Speaker B: 00:36:42

So you're actually, you've been fired for being for plagiarism. Yeah, yeah, yeah, yeah, yeah, yeah, yeah, yeah. Aren't you currently in a very high senior academic position? It's a concern.

Speaker A: 00:36:53

Yeah, no, I'm senior lecturer in photography and AI research at Oxford. And it's a joint thing between Oxford and Princeton and Harvard and Cambridge, actually. So yeah, I should be careful. That was worst case. Best case. Now, this is, does it matter? I got really upset recently when I saw an internet argument about Led Zeppelin. Talking about Led Zeppelin being famous for plagiarizing and all this kind of stuff. And I thought it was outrageous. And I'd heard rumours of it, but I never totally believed it. But I sort of delved down. I was looking at some of the songs and recordings. And I went, oh, you know, it does look like Led Zeppelin totally stole some of this stuff and later admitted it and paid out to people. And the question is, why did you just bloody say so at the time? Why, you know? But in a way, it doesn't matter. Or does it? Because I fucking love Led Zeppelin and I love their songs. And yes, you can totally hear how they just took it. But then they did do their own Led Zeppelin thing with it. And they are phenomenal musicians, wonderful music. And I kind of don't care. Or do I? Because it's not fair on those poor people who, you know, who did some original work and didn't get the money for it, let's say. Or anyway. Yes, that's best plagiarism, let's say. Those are my two. Choose one or the other or both as you like. Let's have Jordan.

Speaker C: 00:38:13

The one that stands out to me is music as well. I think it was... Is this good or bad? Well... Or maybe I'm being too simplistic. Yes, I don't know. It was like a Marvin Gaye song that was plagiarized by Pharrell Williams, I think. It was about 10 years ago. Judged to have been plagiarized.

Speaker B: 00:38:32

But it wasn't even the song, it was just the rhythm, I think. The rhythm, yes.

Speaker C: 00:38:36

Preposterous. Like the intro and then the backbeat of the song or something that was... And then I think they actually settled in court or something.

Speaker B: 00:38:45

Oh, yes. I mean, I think it was outrageous, really. But, you know, something that really anyone could have come up with.

Speaker C: 00:38:54

Yeah, it really was. It was a very sort of simplistic beat, really. And they sort of said that it came from Marvin Gaye's track and... And they sailed out of court. Yeah, yeah. I guess the stakes are quite high when you're talking about that level of music business as well. Yeah, yeah. No, fair enough.

Speaker B: 00:39:11

Fair enough. Okay, that's quite depressing. Yeah, yeah. Nick? Well, I've unearthed a case of plagiarism committed by one Jason Kay of Jamiroquai. Oh, right. Yeah. Oh, what, you yourself discovered this? Yeah, I didn't do anything with this information. Okay. But I knew, I happened to know someone who... Don't get sued. Was a bit of a JK groupie when I was at school. And I actually quite liked Jamiroquai. But anyway, they're track Half the Man. You may, it has a kind of intro which goes, do, do, do, do, do, do, do. It's quite, quite a nice little riff. Anyway, I was playing it and my mum said, that's one of Chopin's ballads, I think. And it was, and lo and behold, there it is. Absolutely note for note. And I looked in the notes, the sleeve notes, which they had in those days. No mention whatsoever of Mr. Frederick Chopin. Disgraceful. So if I was Frederick Chopin, I would have taken him, his ass to court. But, you know, and I assume that Chopin is out of copyright anyway. But I just thought it was really annoying not to credit Chopin with, I don't, and I just don't get it. I don't, like, that would be such a cool thing to do, to write a song based on a bit of an excerpt from Chopin and then to have Chopin as a co-writer. I mean, what are you going to do? What are you going to do? I have to pay out for him. Anyway, so that was that was an early, an early, you know, an early plagiarism incident for me. But I was a victim of, I consider myself a victim of plagiarism in the sense that when I was working on Chopin, when I was when I was. Yeah. But this was this was this was something I had actually had a big hand in creating, which was a book called The Quick Wins for Busy Analysts, which was a kind of well, sort of comprehensive. Was it about quick wins for Busy Analysts? It was. Yeah. But it was, it had a, like, an organising system in it, which kind of divided analytical methods into different functions and had a kind of process of how you go through and ask questions and refine questions and that sort of thing. I was pretty proud of it. It's really good. It's still in use. And we put this out, my team, and there it was with our names on it and contact details and stuff. And lo and behold, about a month later, I discovered that the Cabinet Office, their equivalent of our team, had printed a whole, like direct copies, but with their branding and their contact details on it. And I was, I was, it's all the government in it. It's not like I owned it. I wasn't making money off it. I mean, you know, it was, it's owned by the government. But it really irked me. I would be furious. And it kind of underlined to me that they didn't, I think there are some people out there who don't realise you can create things, who think they have to come from somewhere else. And I assume they looked at this and probably thought, oh, Nick's team has found this thing somewhere and put it into a book and not really thought that, you know, this was genuinely new, kind of interesting, creative stuff that they had, that, you know, had not existed before. And I presume this all came out in court, which is great. Well, you know, it was certainly a factor in making me think I didn't want to work there anymore. Yeah.

Speaker A: 00:42:37

Okay. Interesting. Yeah. Lots of fascinating stuff to sort of delve further into still there. Anyway, thanks as always for listening to the Cognitive Engineering Podcast. I'm Fraser McGruer. We've been here with Nick Hare and Jordan Fermanis of Aleph. Until next time. Bye-bye.