Speaker 1: Julius.ai is an absolute powerhouse of an app that you should be using no matter what field of research you are in. It does writing code, statistical analysis. It is like having a data scientist in your pocket that pops up and says like, hmm, you should do all these things. It is so, so powerful. Also, it does really cool visualizations with a simple prompt. This is it. When you go to julius.ai, this is the simple interface that you're first greeted with. It's got all of the information here. But when you start using it, it's actually a little bit exciting. I'll show you in this video. You only have to do is click here, start a chat. You start a chat and you get this simple, simple sort of like chat interface. We're all familiar with this. And it's as simple as getting started by uploading data. So I have some data here. I've got some public healthcare data. And the one thing I love about this, and it's so simple, is all you do is drag it across and it like plops in like this. Blop. And it just sort of like already gets your information into an AI-friendly format for you. So as it's uploading, you can start thinking about the sorts of things that you want to do with that file. Before you do anything, you want to have a look down here. Default, here are the models. You can use GPT-4 or you could use Claude III. Go check out my other video where I talk about Claude III. But you can use both models in this chat bot. Then you've got tools. You can say yes, run Python code, search the internet, which is just awesome. And plain mode, direct chat with our model and disables, underlying prompts and tools. I don't want to do that. I want to unleash the full power. And then you've got advanced reasoning where AI creates a plan for the task and executes it. That's what I really like. I'll leave it off for the moment. But if you're ever finding that it's just struggling a little bit, click that button and it does such a fantastic job at just sort of like working its way around problems. But let's not get distracted. When you upload data, this is what it looks like. You can see it's already put it in a table format, which is just so useful. And there's a number of things you can do with it in this interface. You can start clicking on columns and say, I want to select from these things only. Or you can say, I actually want this subset of data or I want these rows. That is just so easy to use. Just click, drag, drop, whatever. But I want to know everything about this. But here's the thing. Is that when you first upload data, and by the way, this can sort of like get up to one gigabyte of data into its system, which is just so much data and it works with all of it. You can just sort of like say, what can I do with this data? So please provide me with insights about this data. I'll click go and it will start doing its thing. The first thing it does is examine resources. So having a look at everything that you've put up there. And then this, look, it's just like magic to me because it's got this, okay. It's exploring the public healthcare data. It's got Python code. And as it's running, you know, it's doing all of the different kind of observations down here. And then it's got Python code that you can actually use outside of Julius and put it in something like Jupyter or any other program you're using. And then it's got a simple language summary of the Python code. Now here's the thing. It's a little bit of a confession on my part, which is even though I was doing like chemistry and physics in my PhD, this was so confusing to me. I didn't do any statistical analysis. I used to go to like my colleagues and be like, what is going on? This is just such a great tool to have if you're like me, where this just scares you because it does it all for you. You don't need to kind of grovel to your co-supervisors to be like, oh, how do I do this? No, it does it all for you. And this simple sort of like summary here in plain language just means that I understand what's going on with that code, which I really, really like. Okay, so here it's coming up with some observations. It's got descriptive statistics. It's got observations. And now I can say, let's plot some stuff. Plot some of the best insights from this data. From this data. And this is where it gets sort of like super exciting is that it starts showing you visually what's going on. And so one thing you'll notice is it actually grabs modules for Python as it's going. So you don't need to go and install stuff. It just does it all automatic. Look at all this code. It's just spitting out. That would take me so long to produce on my own, but it is doing it completely on its own. Now this is where it gets super exciting for me because I look at this and immediately the data starts making sense to me. So here we've got distribution of age categories. We've got distribution of type of administration. And we've got distribution of severity of illness. Now all I've done with a simple prompt is produce three insights that just start me thinking about my data. So very easy. Now the great thing about all of this is you can actually download these and use them wherever you want to use them. Fantastic. Okay, cancel that one. But this is how you start interacting with your data. You can also ask further questions and the more you chat with it, the more it understands your data and what you're doing. And it gives you these awesome little sort of prompts down here that you can choose from. Can you provide a breakdown of the distribution of hospital stays by duration? Sure, I want to know that. Let's just let it do its thing. And I'm amazed at just how easy it is. It's replacing all of the tools that I was absolutely terrified by during my PhD. And this is going to be such a useful research assistant no matter what field you are in. It's doing it again. It's spitting out the Python code. Let's see what it comes up with. Now here's something that's really interesting to me is Julia's AI will self-correct. It will try something and if it runs into an error like down here, it will just try again and it will self-correct. So no more troubleshooting when there's all of these issues. It just does it for you. It makes it so simple. You just need to wait a little bit longer. But here we go. This is the corrected plot showing the distribution of hospital stays during duration. So here we are. Here's our nice histogram. Now this is something, if you ask it for a specific diagram, it will give you this option, edit graph. So you can click on this and you can start playing about with how that graph actually looks. You've got legend, you've got size, you've got title, and you've got custom options as well. And if you go here, custom, and you enter a prompt like how you want Julia to try to change the table or graph, it will actually do that for you with a simple prompt. So cool. Legend, show legend, yes. Another thing I tried was uploading some data. So I got gene expression data and you can say, I said here, can you load this data? It loaded the data, gave me a table, and then I started asking it for very simple statistical analysis. And any statistical analysis you can think of, this can do with a simple prompt, almost, I guess. Yes, almost anything. So can you perform a normalcy test on this? It gives me the gene expression, statistics, the p-value for all of this stuff. And this is the one thing I love about it, is can you create a box plot for me? It knows what a box plot is, it understands how to do it, and immediately, just with a simple prompt, it got this relatively complicated box plot for this gene expression, so very easy. All of this is so perfect. If you've got a presentation coming up, if you're writing a paper, if you want to put it in your thesis, and it just takes away all of the hours and hours of formatting in whatever you were using before, it immediately does it for you, so cool. Okay, here's the thing. If you want to use AI for writing, they say, no, you're a naughty person, stop using it. We cannot have academic credibility undermined in this way. But if you want to use it for statistical analysis, go for it. In fact, there are a load of different courses run by universities that are actually based on Julius AI at the moment, so people are very friendly and warm and welcoming to this. This is one from Rice Business School, and you can see we go all the way down here, and they say Julius is the best one we got at the moment, so all of the information, all of the course is based around what you can do with Julius, and this isn't just one university. There's plenty of them doing this, so they are embracing it, because this is a pain in the ass otherwise, so it is something that will be used with greater frequency going forward in academia. Mark my words, and you should be on the front foot. Use Julius AI. The last thing that's just so cool about this that I want you to know about is that if you are struggling, there is a very supportive community that you can go to, and here, for example, we've got a guide that people are producing as part of the community, so we've got descriptive statistics on how to prompt for and run them on Julius. We've got all of this broken down with screen grabs, and the Julius Community Forum is active. It's very, very sort of busy. You've got all of this stuff, and if you're confused, you can go on and ask questions. Absolutely love it. Here's another guide as well, running students, t-tests, and paired samples, so whatever you're thinking of using Julius AI for, there's probably someone out there that could help you if you get stuck. Absolutely love it. Great. I think that every researcher at the moment should be using Julius AI to delve deep into their data, doesn't matter what field you're in, and the prices and plans and the free options are just so generous that you can get started right now, today. It is so valuable, and like I said, it's like having a little data scientist in your pocket that's just like, don't worry, I'll sort this out for you. I really, really like it, actually, so you don't have to do it. You go get yourself a nice cup of coffee from the tea room, please, and also go get yourself a bit of that stale cake from yesterday's lunch, because you know you want it. There we are. That's what it's like. Use it. If you like this video, go check out this one, where I talk about using SciSpace and Copilot. It is also a very useful tool that every scientist should know about. Go check it out.