Speaker 1: Elicit.ai, the research assistant has become so powerful and this is how you use it. Some of the really useful things they've hidden in this kind of maze of options. So in this video, I'm gonna share with you all of the secrets for using it powerfully. When you first log in, this is what you get. You get this screen here and you'll notice that everything is revolving around notebooks. And by the way, I have paid for the pro version, but you can do most of this stuff without the pro version, you just don't get as many credits. So this is a notebook view and you can see up here, you can go to new notebook if you're not in this view. And then this is the workflow that typically I go through. You've got an add notebook title. So you click here and then you can put something, whatever you're searching and your notebooks will be saved in the sidebar here. So this is a nice way to kind of like segment your thoughts and collect references or research papers around a particular theme or topic. So I'm gonna go up here and by the way, if you don't put a notebook title, it will generate one automatically for you. So you don't need to do this step if it bores you. And the first thing you'll notice is there's three ways that you can interact with the data they produce. First of all, you've got find papers, then you've got extract data from PDF, and then you've got list of concepts. We'll be going through each of these so you'll understand how to use them perfectly. But the first thing and what most people use this as is a search engine for a particular research question. So the first thing I would do is go here and ask a research question. This is perfect for anything that pops into your mind, any questions you have about the world. I use this for non-research related purposes a lot. And so the first thing you need to do is know how to generate a really great search. Now those of you that are coming over from something like Google Scholar, it is slightly different. So even though we're gonna use keywords, we have to ask a question that's clear. Oh, clear, that's weird. We need to make sure it's completely unambiguous, but don't worry, we can refine this later if you're not sure where to start. So for example, instead of just asking shiny head wax, we'd have to say what is the best way for a healthy scalp? Something like that, you know, I like that. All right, there we go. And we can see we've got this, it costs about 100 to 120 credits on average, and your credits are shown down here. Because I've paid, I've got a lot of credits. So then all we have to do once we feel like we've got a good question is click this, and it will go away and create this notebook. So you'll see that it's created a notebook title for me, Optimal Products for Scalp Health, and then it's got the question here. And then it will start going away to find research papers that answer that specific question. I love it, this would take hours when I was doing my PhD. Now it is a matter of seconds. So here we go, this is what we get. We get this summary of the top eight papers. So when you're not paying for it, it's not as many papers, I think it's limited to four, but this is essentially like a snapshot of all of the things that you need to know about, and it's referenced as well. If we click on the references, it takes you to that paper, so we can go back to that notebook. And this is how Ellicit kind of like displays its results. It's got columns, so it's got papers, it's got abstract summary, and this is how you now navigate all of this information is you can filter. First of all, you can say I want it to have a PDF, I want it to have this publication year, I want it to be this study type, I want it to have abstract keywords that contain this or does not contain this. So this is how you start sort of like narrowing your search to something that's far more specific and useful for your research field. I'm gonna go here and click cancel, and then here we've got most relevant. This is the sort by most relevant, most recent, least recent. So this is a method for sort of bubbling up the most important papers in your research field based on different sort of criteria. And then this is it, we've got all of this. Now, it is displayed in columns. Each row is a paper, but each column is a question, an AI-generated response. So here we've got the abstract summary, and this is AI-generated. Hair oil is one of the best treatments. Great, I love hair oil. So now, once we've got this abstract summary, we can add our own columns. So we can click here and we can put what kind of data we want to extract, and it will use AI to do that. We've got all of the ones that you would normally use for a research study, but now I'm gonna create my own. I want to know about that they contained human trials. Okay, and then I'm gonna create a column. Now, one thing that you'll notice is that it gives me this. Create custom column, contains human trials, and it's a description, okay? So this is optional, but it's really recommended that you spend a bit of time telling the AI what you want for that particular column. It's really important. There we are, and then create column. And so now, it will say whether or not it contains human trials, not mentioned, not mentioned, answer not found, not mentioned. So all of these are just sort of like a great way, or none of them did, answer not found, not mentioned, not mentioned. So there's not much in there, but you can do it for whatever query you've got. I should have probably chosen something a bit better. Now, with this column, we can turn on this, which is high accuracy mode. It uses more credits, but it reduces the mistakes to about 50% of what they were before. So if you've got pro, I really recommend that you turn this on, and then we can also filter this column as well. Add the keywords we want to filter by. So if we did see any human trials, we could type in, for example, and then it would filter all of the human trial ones. So that is really great. Then you'll see down here, we've got this bar. This bar is always a little bit hidden. It's always floating there, but it never really stands out to me, and this is where we can add a new step. So we can ask a new question, and this is that same notebook. So this is how we can kind of get like a ledger of all of our thoughts and processes as we're going through this. So we've got ask new questions and find paper, extract data from papers, get a list of concepts, and then if we start clicking here, you'll see that some of them start being not grayed out. We can chat with papers. We can summarize abstracts for selected papers or create a new table from the selected papers. So that is a great way of honing in your research question and starting to explore the literature much more efficiently than we did in the past. Another thing we can do if we select papers, for example, this one and this one, oh, I don't speak, is that Korean? I'm not sure what that is. There we are. We'll put that in, and then this is what I like. So sometimes AI will throw up some things that aren't relevant, and sometimes the responses aren't brilliant, but if there are two or three in there that are good, we can select them using this checkbox and then go to show more like these, and then it will say to the AI, these are the ones I like, so just sort of like use these as a basis to go in. You can see that it's found more examples of papers that I selected. Brilliant, that's what I love to see. The next thing you need to know about is going into this notebook view and this next one down, extract data from PDFs. This is really powerful if you've already got PDFs or you've found some and you want to chat a little bit more with that data. Very, very cool. Something that used to cost a lot of money just a couple months ago. So I'm going to click here this time, and I'm going to upload or select papers. So here I've uploaded papers. You can go to library and add them there as well. You can see upload papers, and you also can connect it to your Zotero. This is why I recommend Zotero. There are so many cool integrations, but we'll go back to new notebook. We'll go to extract data from PDFs. These are ones that I've already uploaded to my library, and you can see I can also upload papers. So I can drag and drop PDFs or click to browse files, and then essentially we click the ones. Let's go here and here and here, and these are the ones I want to chat with. So Elicit is going to create a new notebook from these, and you can see you've got that same table structure again. We can add columns that we want, but importantly, I want to go down here, click, click, click, and I want to say new step, chat with papers. So this is what I want. I want to chat with the papers, and here you can see it brings up this typical chat dialog that we're used to seeing with AI tools, and this is in beta, and by the way, this can be quite chatty. So you do have to steer it a bit more than other large language models at the moment, but I think that's going to change over time as they get out of beta, and so here we are. We've selected three papers, and then I can ask it anything about those papers. The one thing I'd recommend you do initially is turn off use full text because that will only use the abstracts of the papers. So if you know that the answer can be found in abstracts, turn that off because it's just going to save your money and save your credits. I'm going to turn use full text on, and then we can start chatting with the papers, and then what is a research gap in these papers? I'm going to click enter, and then this is when it starts using the full text, and we can start asking questions across a multitude of documents, which I really like. So here we go. A research gap identified across the papers involved a further need for further optimization. So a really great way of asking those very sort of detailed questions about very specific work that general text, if you just ask it to go out and find something, won't be able to do as well at. So give it the data, give it the papers, and you can really delve deep. This is something that I really like, and you can see, look, I can go in and add a new step. So these are all the next things I can do, but this is a super really cool thing, super really cool thing, that I like. So let's go back to new notebook and look at the very last thing, the list of concepts. Okay, so here we are back on the landing page once you've signed in, and I want to go down here to list of concepts. And so this is really cool, because if you're entering a research field, sometimes you're not quite sure what is out there. This is a very broad level thing that I would do at the beginning of any research project, or even if my research is taking like a new branch, a new direction, I can be like, this is what I need to do to make sure I'm covering all of the concepts, and I'm over all of the different sort of areas of a particular research field, so that I can understand that area much, much better. So here, treatment for hair loss. Elicit, come on now, don't trigger me. But I'm going to say treatment for hair loss, and list of concepts, let's go. So once I've clicked it here, you can see it's searching for papers, then it find the concept in each source, and then it will remove the duplicate concepts, and it will give you the final answer. So here we are, you can see it working. This is a concept, so it's got hormone modifiers, hormone modifiers, surgical options, minoxidil, Rogaine, surgical options, so all these are what it has found in terms of the different types of keywords that you can use and search for later on. So you can see that it's found 211 concepts in 60 papers, it found 88 unique concepts, and this is all of them, and you can download them as a CSV, and use them in Excel to make sure that you've captured that somewhere, but if you're logged in here, it'll always be in this notebook. So here we are, these are all of the different things, and this is one thing I really like is that this, in the past, to get over and around and understand a particular research field would take hours of reading. Now you've got everything you need to know in a simple table format. Really, really cool. Thank you so much, Elicit. And then once again, we can go down here, click Add New Step, and this is where we can start sort of like investigating a little bit deeper. So we can ask new questions and find papers, we can extract data from uploaded papers, we can get a list of concepts, which we've just done, but because this doesn't contain papers, we can't click or select anything, these are grayed out. So these are all of the things you can do next, and this is a really, really powerful way of ensuring that you know everything you need to know about a particular research field. Love it. If you like this video, you'll love this one where I talk about using Julius AI to unlock the data in your research. Go check it out.