Speaker 1: The first way that you can use ChatGPT for Omni for science and research is like this. I really like using it for talking about and explaining papers. So here I've got a paper, this is one of my papers, I've uploaded it and it's as easy as dragging and dropping it into this message bar and just saying, explain this paper to me and the key results. Clearly you can do a bit better than that. But this is what I love about it. You can put in full PDFs and it gives you an introduction, experimental approach and the key results if you ask for them. This is what I'd be interested in with any peer-reviewed paper that I'm reading. And then down here it's got figures and data interpretation. This chat model gives you all the information and more. It almost anticipates what you need, which is what I absolutely love. And this is a model that combines text, audio and visual input so that you can sort of give it loads of context to get it to really hone in on what you want it to do. So I said here, can you show me the figures? This is something I've not done before with any other model and it says, here we are, no images were extracted from the papers I targeted and it sort of like self-corrects. So it finds the figures, first of all it finds kind of the understanding of the figures and then it says down here, to provide visual context I'll display these figures directly and it bloody does it. Look at this. It actually got the figures from the papers. So if you want to talk about this paper in a presentation, it's easy to extract the figures. You end up with one figure that contains all of them. I absolutely love this. You can download it and it just makes it so easy. And you can go backwards and forwards chatting to this model over and over again and it just seems to understand exactly what you like. And like I said, it gives you so much lovely context in which it sort of responds that it is really, really useful. So if I want to know what the limitations of this research are, I can just ask it what are the limitations of this research and it has awesome memory now so it can kind of really understand all of the thread and use context from stuff very, very early on. That's one thing I really like about it. And look, it gives such a well-rounded, deep understanding response to this paper that it is just getting better and better. I would almost say you don't need any other sort of like specialized tool anymore for science to do something like this. ChatGPT Omni is doing an amazing job, but it doesn't just stop there. This next one is probably one of the favorite ways I've been using it recently. One of my weaknesses when I was writing up any research was the fact that I just got lost in the words. It's so easy to explain it out loud, but sometimes you get lost in the nitty gritty of this paragraph or this word or this verb. But if you just talk about it to ChatGPT, it can give you the paragraph from your sort of like jumbled mistakes of a sentence. And this is how I love using it. So I use my phone and I go on and I just open up the chat dialog so I can use it with voice and then I just push on it. The reason I push on it is so that I can actually then just talk about things without it assuming it's all done. But this is the sort of stuff that I say, hello ChatGPT, I'm writing a peer-reviewed paper at the moment about single-walled carbon nanotubes and carbon nanotubes and silver nanowires and I would like to produce a transparent electrode that has strong mechanical strength because of the interwoven nature of the two materials. Can you give me a paragraph just to highlight the benefits of using a composite material like this? And then you release it and it does its thing. Now, it will respond to you.

Speaker 2: Here's a paragraph highlighting the benefits of using a composite material of single-walled carbon nanotubes and silver nanowires for a transparent electrode.

Speaker 1: Okay, yeah, I don't like listening to it. But one thing I like to do after that is if I go to and refresh the chats that I'm having, I can then go in and say, okay, composite electrode benefits. Then I've got this paragraph that I can use going forward which I absolutely love. This is just such an awesome way of producing a first draft that it's just so easy now to take your jumbled thoughts and turn it into precise academic language. Absolutely love it. Check this out. Combining single-walled carbon nanotubes with silver nanowires to create a composite material for transparent electrodes offers several significant advantages. And then it gives me those. And look, it was just from talking for a minute. Now I've got something I can work with. Love it. Yes. Yes. Yes. One thing that this new chat model does is it provides structures and outlines in a really succinct, really nice way. It really seems to get what you want to put out. Now, obviously, this doesn't have specific domain knowledge, but it does incredibly well across a range of different fields. And I'm absolutely so impressed by how well it's able to navigate all the different fields and give you structures for papers, for theses, for literature reviews. Check this out. One of the important things is give it context so it knows what you want it to do. So normally I say context, what I'm trying to achieve, and the output that I want. So in this case, I've got a literature review. I want it about elephants in Africa. Why not? And then also provide me with an outline. So I'm just going to send that off. And one thing you'll notice is that this is not only a very quick way of doing this. This is actually probably one of the best ways of doing this. Because something like this, creating a structure for a literature review takes hours. If you have to read the literature you've read, now you've got a starting point. So here, we've got all of the things that we could talk about, introduction. And the one thing I like about this model is that in the past, it used to give an outline. You have to say, I'll go deeper. This kind of understands what you need from it with a very simple prompt. This idea of overly complicated prompt engineering, I think, is out the window with this model. Because look, it gives me not only the sections, but also in the past, I would have to ask for this a little bit more clearly or do a follow-up prompt. But this is all in one go. And I can copy and paste this across into a document and then start fleshing out this as a literature review. It does it for a range of fields. Check it out for your field and try it for your next peer-reviewed paper or your literature review or your essay. And I think the ability to create a structure is one of the most powerful features that you can use in ChatGPT. This is one of my favorite ways of using ChatGPT for Omni, and that is giving it a load of figures. So I've uploaded one, two, three, four, five figures, and this is sort of like how I would use the visual capabilities of ChatGPT to write a paper. So I can say, I'm writing a paper and here are the figures I want to include. Create a story structure for this peer-reviewed paper. All right, I've used peer review a couple of times, but let's give it loads of context to see what it does. And so here it's looking at the figures, and the one thing it does is it can extract the text within the figures and understand what each figure is. To make this even better, I recommend that you also copy and paste in some captions because it is able to extract the text from those captions and give you even better responses. So here we've got all of the stuff here. This is one thing that amazes me about this new model is I asked it for a structure, but it's gone above and beyond. It knows about how to create an abstract. It's got bullet points to create in your discussion. It's got details and methods, described the preparation process of the P3HT blends, including the annealing and cooling, optical properties. It's got everything that I would need to start writing a peer-reviewed paper. The great thing about this is all it was was giving it some figures. It's just incredible. And in fact, these figures were from completely different papers. So these four are from the same paper. This is from one. So it managed to navigate that actually pretty well. And I think that sometimes one of the hardest things about writing a peer-reviewed paper is coming up with the structure and the story that you would tell and try to sell to a journal editor. This is how I would do it, no doubt, if I had to submit something right away. Just get the data that you've got, mush it together, ask ChatGPT to create a story structure for you, and I think you'd be surprised at what is actually submittable to journals. A lot of the time we think, oh, we just need that one more experiment. No, just put it in and see what comes up. I think you'll be happily surprised about the number of papers you can produce in this fashion. If you like this video, go check out this one where I talk about ChatGPT Vision and the insane things it can do for research. I think you'll love it.