Speaker 1: I have been blown away by what ChatGPT 4 can do and I think it's perfect for research. Now in my last video, go check it out over here if you haven't seen it already, I used ChatGPT 3.5 and I can tell you that stuff has come on in leaps and bounds, especially for scientific research and literature reviews. It is incredible. Okay, the first thing I think we need to address is the referencing situation. So in my last video, the references that ChatGPT gave me were absolutely fake. It emulated what an actual research reference looks like. Let's see if ChatGPT 4 is able to give me real references. That will be incredible. Okay, so over here you can see that I've paid for ChatGPT Plus and it's about 20 US dollars a month, I think, but it is well worth it. Here we've got the different models. This is where I was before, ChatGPT 3.5. Now we're at ChatGPT 4. So we'll make sure we're on that and we'll just ask it. We'll just ask it for simple references. Let's say you're starting out in a research area. Let's have a look to see what it can provide and if it is real. Provide me with references and links to papers in the organic solar cell field. Okay, let's hopefully see if this one works a little bit better. Here we go. It's a science paper and it's given us a link. Let's see if that link works. I think it's a real paper. I think it's a real paper. Alright, so I completely messed up last time by not clicking on the links and now it looks like it's really trying. Now this is only for stuff until September 2021, but it's still a massive improvement over the last one. So let's try a couple more. Let's have a look at this one. Polymer solar cells, that's a real paper. Let's try this one. The emergence of perovskite solar cells, that's a good one. What else we've got? Next generation, let's try this one. Not found. Alright, so it's not foolproof but it's way better than it was. So you do have to go through and double check all of this stuff and I'm sure that you could go and find these. Let's have a look. Let's see if they're actual real papers by typing in to Google. The real, so it had additives for morphology control by Liao and it's there. It's there but not all the links work. So they're real papers, not all the links work, but that's okay. So there's a really good improvement, isn't it? Another massive improvement in ChatGPT4 is that you can give it much more text to analyze. I don't really understand the sort of like ins and outs of it but essentially you can copy and paste and tell it to read a load of stuff before you ask it questions and this is perfect for creating scaffolds. So if you need to start on something, let's say you need to start on a presentation, you need to start on an abstract, where do you start? Well, you can actually take other people's things that you like and ask it to create a scaffold for your work. Let's create a nice little abstract scaffold for ourselves. So if I go in here and say ChatGPT4, okay, please read this and say read when done and then colon new line, paste it in. Let's see what it says. Okay, read. It's read it. Let's give it one more. Let's give it this one. Please read this and say read when done. And you can do this for as many abstracts and I think the more the better. So there we are. We've got two in there but you can add up to 10 if you really want and let's just say please create a scaffold. How do you spell scaffold? Scaffold for someone wanting to create their own abstract from the stuff you have just read. Let's see what it does. Let's have a look to see introduction. So a background on the polymer photovoltaic cells and that's exactly what happens in recent progress, the development, blah, blah, blah. So it's just like a really sort of broad opening statement. Materials and techniques used, the progress in polymer organic solar cells. I don't feel like this has done it perfectly. I think that's my mistake for not giving it like the perfect command but let's say create a table with prompts to create an abstract like the ones you read. Let's see what it does with tables. So section and prompt. Section one title. Concisely describe the topic of the paper. Introduction. Briefly introduce the context and relevance of the research in polymer solar cells. Outline the materials used. Any specific techniques. Highlight the progress made in the field. Mentioning improved efficiency rate. Describe the potential application. Summarize the key achievement in your recent field and recap the main findings and emphasize the importance. So that's the sort of structure you can use. Remember this is only from two abstracts. If you were to put more in there it would allow you to find a much better and sort of succinct scaffold that you could use and you can use this scaffolding for anything and I'm talking anything. You can use it for presentations. If you find presentations that you like you can ask it to analyze those and produce a scaffold for you. If you're not sure how to write your materials section you can ask it to create a scaffold for all of that. It takes the guesswork out of writing and doing research and creating scaffolds is how I think CHAT GPT-4 can really help you personalize the outputs for your research. Okay this one is actually pretty exciting to me. When you go to conferences all of that audio, all of those presentations are normally forgotten about. Now you can use CHAT GPT and another AI tool to capture the scripts from those presentations and turn them into useful outcomes for you after a conference, after you've watched a presentation. You can even find stuff on YouTube, presentations on YouTube and you want to extract the data from that for your own understanding, for your own research. It is incredible. Let me show you how I do it. So the first thing you want to do with any audio that you've recorded at conferences, that you've recorded at presentations you've been to, even if it's like a lecture, if you're an undergraduate, getting the value from those is very difficult until now. So the first thing I do I've got a Python code that I've created from Whisper AI. In this case I can actually use it to export audio from YouTube but essentially it kicks out the other end, it kicks out the script. So let's go to YouTube and let's just have a look for some three-minute thesis winners. So let's take this one. So this can be any, oh there we are, this can be any presentation that you've personally recorded or you found online or that you've attended. So let's go here, let's grab that URL and if I run this script, down here it's going to ask me what URL I want to grab it from. So eventually that will pop up. Enter the YouTube video URL, so I'll put it there, push enter and now I just have to wait. One thing you'll notice is that it's kicked it out in a text format which is perfect for putting into chat GPT-4. So Matthew Thompson is our first finalist today from the University of Queensland, well done and here is all of his talk and it does a really good job at actually creating text from speech. So if I take all of this for example and I put it into my chat GPT, so let's start a new chat, let's choose GPT-4 and just say read this and say read when done. Okay it shouldn't take long, read. Now I can say extract the important key points and let's just see what it says. Matthew Thompson is a finalist, okay that is a very important first thing. So here it's giving us a breakdown of the actual information in the talk and you can do this for any talk you attend, you can record it on your mobile phone, you can record it afterwards if it's a record lecture or something like that and it's just about extracting the text from the talking and then getting it into chat GPT, it does an incredible job at actually synthesizing it together. You can use this to create a scaffold from a talk, you can use this in so many ways and I've been doing this with different talks and different bits of information that I wouldn't otherwise be able to access the value directly from because it was locked up in a video. Now I can get it and I can get it into a written form very easily and summarized and worked on, it's incredible. So here is the end thing, so contrary to popular belief fingerprint identification is done by humans and not computers and here all of this, this is perfect. If I wanted to sort of like put this just for my own understanding somewhere or if I was a conference and I went to like 20 talks, I was able to do this for each of the 20 talks, put it into a little document, even a PowerPoint document like I did with my literature review, so how to read like a PhD, go check out that video, that is how you extract the true value of these conferences and I think more people will be doing this in the future, you've just got now a head start on them. And the last thing I want to make you aware of is that the generative and the creative power of this new version of chat GPT i.e chat GPT 4 is incredible. You can use it to generate text. Now what I do is just like we've been putting stuff in, I've actually been putting in results and discussions and then saying write a conclusion based on these or write an abstract for me based on this stuff. It gives you at least a starting point, it isn't always perfect but generative text from the input you give it is really the way you make chat GPT 4 work for your research. Let's take one of my papers and see what it can do to generate the abstract or the conclusion after I've given it all of the data. Over here is highly conductive interwoven carbon nanotube and silver nanowire transparent electrodes beautiful title. Let's just put in the experiment, no let's put in the results and discussion. So let's go through here and see if we can just copy and paste this into something and see what it will give us. Now it won't do all of it, let's see how much it can actually do. Let's just copy that and say read this. Let's start a new chat just to make sure we're not sort of poisoning the well with other stuff. Read this and say read when done. Okay good it's now just giving me a little bit of a summary of what it's read. Now I can say please generate an abstract based on this. Now it may not be perfect but I'm sure it'll give us a really good starting point. So generate an abstract for a science paper based on this information. It's actually doing a pretty good job, I cannot believe it. All I did was copy and paste in the results section and now it's giving me an actual real abstract that sure it's maybe not perfect but it's really bloody good. There's a lot of work actually that is now no longer needed to be done by you manually. It's a great starting point and let's try generate an abstract. No generate a conclusion for this article. In conclusion our investigation blah blah see what I now choose composite films reveals promising potential as an alternative to indium tin oxide the composite film demonstrates enhanced conductivity mechanical stability and optimal balance of transparency when incorporating up to it's good it's not once again it's not perfect but I'd rather start from something like this than sort of struggle by thinking you know trying to start from zero and the last thing I think we can do which is really powerful is generate a sensational press release for these results and this is great for like a science alert article or you know getting someone excited about your research i.e the marketing team at your university. I think it's really really powerful. Also remember you could copy in some science blogs and say create a blog like this one but with my results. That is not beyond the realms of possibility with this because it can actually take a load of text and use that to generate its own sort of versions of blogs. Incredible. They've called me lead researcher Jane Smith. Our findings are open an entirely new avenue for development of so it's actually you know it's providing quotes it knows what it what a press release is actually like and it's done a bloody good job. Boom. Incredible. Use chat gpt4 like this and it will really sort of make your productivity skyrocket because it takes a lot of the grunt work out of anything you want to do. So there we have it there are the best ways I think you can use chat gpt4 at the moment to accelerate your research career make everything so much easier and all you have to do is pay 20 US dollars a month at the moment and I don't know what they're going to do in the future but I know that you can actually in the future they're planning on releasing image analysis. I wonder how well that's going to do with like scientific figures. Watch this space I'll let you know. Let me know in the comments how you would use chat gpt4 for your research because I'm sure there are a ton of uses and we've not even scratched the surface yet and there are more ways that you can engage with me. The first way is to sign up to my newsletter when you sign up to my newsletter you'll get five emails over about two weeks everything from the tools I use, the podcasts I've been in, how to write the perfect abstract and more exclusive content how to write the perfect abstract is my favorite in all of that so go sign up now and also head over to academiainsider.com that's my new project where I've got my ebooks the ultimate academic writing toolkit as well as the phd survival guide we've got the blog grown out there as well the forum and resource packs for different stages of your research are going to go on there soon so go check them out and I'll see you in the next video.