Speaker 1: Writing in academia is inherently messy. You have to come up with references, you have to come up with prose, you have to come up with evidence to support what you're saying. So there is a lot of to and fro. You may find something that you didn't even know about that you want to include after you've written a large portion of your introduction or your literature review, for example. And here's the process that I would use in academia if I wanted to make sure that this process was as streamlined as it could possibly be. The first thing that I like to do is have a look to see what has already been published that I can use to structure my own response. There's no point reinventing the wheel and large language models are perfect at coming up with the structures that you need. So I like to use GPT-4 Advanced Data Analysis. Here you can add a file and that file, in this case, is gonna be an Excel document with different examples of what I want to write. For example, here I've got abstracts and I've got these from my own papers. Essentially, we need a couple of examples that we can put into ChatGPT, which would then give us an opportunity to structure our own response. So I've got this, I put it into ChatGPT by clicking here, and ultimately, I end up with this prompt. And I'm gonna put all of these prompts in the description below for you to use. This is a file containing multiple abstracts from scientific papers. Use these abstracts to create a scaffold from which someone with no experience could structure their own abstract for their paper. So as you can see, it goes through, it says a couple of questions, like yes, would you like to identify common structural elements? Yes, please, great, let's continue. And so it works all the way down, and you can see here, we've got abstract one, and then it's broken it down into what is actually the structure. Introduction or background, objective or aim, methods, results, and it does that for both abstracts. And then I've asked it here, hey, can you create a scaffold for a scientific abstract? And this is what it looks like. If you can put it into a table, if you prefer that, but essentially, that is what we're using to structure our own academic writing. It's so easy if you just look at what people have already done. It doesn't just work for abstracts. You can do this for introductions, for example. So I did exactly the same thing for introductions. I got introductions in an Excel document, and I just pasted them in here. You can see these are the same papers that I got my abstracts from, but essentially, we put it in using exactly the same prompts. And this is a file containing multiple introductions from scientific papers. So I can use these introductions to create a scaffold from which someone with no experience could structure their own introductions for their papers. It goes through, goes through, goes through. And at one point, it actually said, hey, look, there's a lot of input that's been truncated. However, I can still identify structural elements often found in scientific papers from my training data. No, we don't want that. So I said, no, please use the data provided, and it apologizes, and then it uses the actual data that I provided. So a little bit of a trick if you find you need to sort of go back and forth with ChatGPT. Here, we've got background information, problem statement, and it's done that for both of the different introductions. And you can see here that I said put it in a table. And so here, we've got section, background information, problem statement, significance, gap in literature, methodology preview, and outline of contributions. But ultimately, this is what I'm going to use to structure my own response. So scaffolding and coming up with examples from the literature that have been accepted in different peer-reviewed journals or have been accepted where you want your publication to go is a great way to start. And that is how I would first use any AI tool to do this. You can also do this in Julius AI. So if you don't pay for ChatGPT+, you can also do this in Julius AI. You can see here, I did exactly the same thing. I put in my abstract, and I said exactly the same prompt. This is a file containing multiple abstracts, and then it goes through and exactly the same thing it's come up with. And then here, it's got an example with each section from the abstracts that I've put in. So it does it a little bit differently, but ultimately, it's the same outcome. It gives you a structure by which you can work with. Julius AI is free at the moment, but if you can pay for Plus, it's a really great tool. Once you've got your scaffold, you've got to ask yourself, right, now I need to start writing. So you know what you need to write. Now you just need to generate the words. And there's a few tools that I would use to do this. The first one, obviously, is ChatGPT. So here, for example, I could say, I'm currently writing an abstract for a paper and need help writing the abstract. Create a referenced introduction or background statement about the organic photovoltaic devices that I could use for the first two sentences. So here, we can see that it says, in the quest for sustainable and renewable energy sources, blah, blah, blah. So it's come up with that first thing. It hasn't referenced it. It hasn't provided me with references. This is default 3.5, so it's free for most people to use. And I can actually prompt it a little bit more by giving a little bit more information about my particular study. For example, it says here, I'm currently writing an abstract for a paper and I need an introduction and I need to state the problem or challenge that my research addresses. In my case, removing the need for harmful solvents during manufacture and using water-dispersed semiconducting nanoparticles. So here, you can see it's come up with a load of information that we could just simply cut and paste into a working file and that's what I've got right here. I did this with the abstract. I did this with the introduction and I've got this working file and I'm just going backwards and forwards, backwards and forwards, asking myself what I need and then looking at ChatGPT or any other AI tool that I'm using to help work as like a research assistant with me to produce the text and then, the next step is to reference it and that goes backwards and forwards for ages. So if you get stuck in that loop, that is a good place to be. It means you're refining your ideas, you're refining the arguments you're putting forward and you're looking more at the literature that has been published before. So that is just the nature of academic writing. It's very, very messy. There are a few other places that do a lot better than ChatGPT for actually referencing what they write. So for example, if you want something free, you can look at perplexity.ai and you can see I've put in the same prompt where I'm saying, look, I need to introduce the field of study and then I need to state the problem. This is the problem that my research addresses and you can see it's come up here with an abstract that I could potentially use. Now the thing is, if we go back to my reference sheet, you can see that the citations aren't really useful. I mean, they're broad, it's Wikipedia, it's a .gov site, another Sigma Aldrich, another Wikipedia, it's got nature. So this one's actually an article that I could reference but a lot of the time, it really doesn't grab onto that scientific sort of credibility from papers that we need. So we can use things like site. So here you can see this is site.ai and I put in exactly the same prompt and it says here that it's got organic photovoltaic devices, blah, blah, blah. So it's actually gone away and it's found us some references. Now the thing about all these references, you can't just blindly cite them, you need to go and look at them, that's part of the messiness that is academia but you can just sort of like use this as a starting point. So here we've got reference one, reference two and it does sort of show you why it's mentioned these things. So it is very important to actually sort of like read deeply into these references once you've actually cited them. Another tool you can use is Mirrorthink. I actually quite like Mirrorthink. You can see here I've put in the same prompt here. I'm currently writing a paper for a peer review and need help writing the introductions that tells the AI what I want to achieve. It says the paper is about OPV devices and addresses removing the need for harmful solvents during manufacture and using water dispersed semiconducting nanoparticles. So here I've said, right, the background information, the problem statement and the significance. Where did I get these from? Well, I got it from the scaffold that I created previously. So if you look here, I've gone in, scaffold for introductions and this is exactly what I've asked it to do. I want it to give us background information, problem statements and significance. So I'm using the scaffold to create the prompt that I'm putting back into any AI that I'm using. So here, Mirrorthink, you can see that's where I got that information from. And then overall, it's done a really bloody good job at just sort of like providing background information, helping me with my problem statement and addressing the significance. I copied and pasted this into my introduction section here. You can see I've got the table here and this is what it's ended up with, background information, problem statement and significance. That's just copied and pasted in. But the ultimate thing here, once again, it's not giving me references like a tool like Sight would. So now I need to go and find those references and this is how I do that step. Finding references to support what you're saying is so very important that AI tools that I would use are, first of all, something like illicit. So here I can see that all I need is something to support what I'm saying. So ultimately, I need something to support this sentence and I just sort of like copy and paste that sentence into illicit. Organic photovoltaic devices have emerged as a promising solution, blah, blah, blah. And I got that from the first sentence here. Organic photovoltaic devices have emerged as a promised solution. So I've taken all of that sentence and I've put it into illicit. And I think it's sorting it by relevance. So here you can see this is a super new thing and it's about organic photovoltaic devices for indoor applications. That's not something I've considered yet for my peer-reviewed paper. So it may be worth reading this and putting in something about the applications and the novelty of OPV. This paper tells me something that I didn't know about the field before. And now this is a backwards and forwards, backwards and forwards. Once I find something, I read it. Is it interesting? Great, put it for later. Is it interesting for this paper? Great, put it in the paper. The one thing I love about Cite is this prompt. Find a reference for this statement. Organic photovoltaic, blah, blah, blah. I put in that same thing and you can see it's come up with a load of references that I could include in my research article. Overall, that is what I would do. I would spend a lot of time going backwards and forwards in that text generation stage and finding references to support it. And I'd go backwards and forwards, backwards and forwards until I was confident enough that what I had written is not only credible, it contains factual information and I have covered the information as outlined in the scaffold that I produced at the beginning. And that's when it's ready to go off to your supervisor for checking. Now that's all the academic writing you've got to do. What about literature reviews? Well, go check out this video where I talk about the tools that you can use to actually automate your literature review writing. Now, it's automatic in the fact it will generate words, but it's not automatic in the fact that you have to look over it to make sure that everything that it writes is factual, true, and represents your research and what you want to say. So literature reviews are a bit of a different beast. You do have to end up in that really to and fro, read the papers, put them in, rejiggle what you're saying. But ultimately, using AI tools at every single step of that can really help improve your productivity and move things forward a lot quicker. So there we are. That's how I would use AI to write anything in academia that I needed to write. Using AI tools is going to be the way of the future and I'd love to know what you think. Let me know in the comments below. Also, remember, there are more ways that you can engage with me. The first way is to sign up to my newsletter. Head over to andrewstapleton.com.au forward slash newsletter. The link is in the description. And when you sign up, you'll get five emails over about two weeks. Everything from the tools I've used, to the podcasts I've been on and how to write the perfect abstract and more. It's exclusive content available for free. So go sign up now and also head over to academiainsider.com. That's my project where I've got my eBooks, I've got my resource packs, I've got the blog, I've got a forum and it's all over there to make sure that academia works for you. All right then, I'll see you in the next video.