[00:00:00] Avery Smith: The data analysis landscape is changing quickly in today's market. More and more AI tools for data analysis are coming out like every single week, and it can honestly feel really overwhelming. But don't worry, guys. I got you. I'll be here to guide you through all the confusion and all these different releases, so make sure you hit subscribe Today.

[00:00:21] Avery Smith: I'll be comparing two popular AI tools for data analysis chat, GPT, and Julius ai, who happens to be sponsoring today's episode to see how they analyze data differently and when you'd want to use one over the other. And although Julius is sponsoring this video, I'm going to try to be really fair in my analysis between the two, in fact, to compare these two AI platforms.

[00:00:44] Avery Smith-1: I thought I would just test them both out the exact same way. So here's the test. I'm going to ask each the exact same questions in the exact same order, on the exact same dataset, and then judge their responses. That way you can see the differences and know [00:01:00] the pros and the cons of each tool. For this experiment, we're going to be analyzing 1,444 data, job listings from my own data job board.

[00:01:09] Avery Smith: Find data job.com. We'll be trying to see what trends we can learn about the 2025 data job market. So let's go ahead and get into it. The first thing I'm gonna do is go to chat GPT and upload our file. You hit the plus button right here. You add the files, you select the file that you want. I'm just gonna go ahead and upload this right here inside of chat, GPT.

[00:01:30] Avery Smith: So you can kinda see what our data set looks like. Now, I will say that chat, CPT doesn't do a great job of showing you this data set unless you open it up a little bit bigger like this. Um, and you can see that we have the IDs. You can see that we have categories, we have the job description. This is where the actual description is of what the job entails, like the requirements and all that stuff.

[00:01:50] Avery Smith: We have the link, the publication date, and then we have the actual job title, like AI business analyst, or. Ag Data Analyst, the company and the [00:02:00] location. This one's in Moose Jaw, Canada. How exciting. The process for getting your data inside of Julius is very similar. You'll just go ahead and create a free Julius account.

[00:02:09] Avery Smith: I already have one, so I'm going to log in and then you'll do the exact same thing. Hit the plus button right here, upload the file, and uh, I'll just go ahead and run that inside of Julius. Once again, you have a very similar preview right here. I like Julius's just a little bit more because out of the gate it is a little bit easier to read than chat GT's right here.

[00:02:29] Avery Smith: One of the things I really like about Julius is if you click on any of these columns, for instance, if I click on the extracted at, it'll actually generate this clarification or this note on the right hand side that kind of explains the column. This particular case, it says that this column consists of timestamps representing when job listings were extracted.

[00:02:43] Avery Smith: The timestamps appear to be mostly concentrated around June 28th, 2025, with some earlier dates also present. And I can go ahead and open this in a new image and kinda see what is going on. So it does look like we had a lot of jobs recently. So it almost like does some like. Pre-analysis for me at the [00:03:00] beginning, which I think is really fun.

[00:03:01] Avery Smith: Another thing I really like about Julius is it gives me four suggestions on things that I could do to analyze my data. For example, how are job titles distributed by category? What is the salary range for different job titles? Which locations have the highest number of job postings, and is there a trend in job title, frequency over time?

[00:03:17] Avery Smith: Now chat, GPT kind of does something similar. It says, let me know what you'd like to explore trends over time. That's kinda similar. Most common job titles are companies that was kind of similar. Skills analysis from descriptions, remote versus office roles, salary insights, but I have to actually copy and paste these or just rewrite it if I want to.

[00:03:34] Avery Smith: In Julius, it's as simple as just clicking one of these. For example. Let's go ahead and just click this one right here, which locations have the highest number of job postings and right away, Julius has already created a plot for me inside of their plots feature right here. It's showing me that most of the jobs posted are remote, but then it goes Seattle, US, a remote, United States, Bellevue, Washington, Chicago, so on and so forth.

[00:03:55] Avery Smith: That was really quick. Now I wanna go ahead and take that exact question. Which [00:04:00] locations have the highest number of job postings and write it out here in Chachi pt. And so when I did that and press entered it gave me some Python code to find the top locations, and then you'll get the top 10 spots. But here's a quick placeholder example, based on typical tech listings in the United States.

[00:04:15] Avery Smith: So chat g PT didn't actually do any data analysis here. It just gave me code to actually do it. And I think in order to actually have it do some data analysis, I can have it right or run code right here and re-ask the exact same question. So let's go ahead and do that and hit run. So it generated some Python code for me right here, and then over here on the left hand side, it gave kind of a similar response with remote Seattle, Washington, USA remote USA, Bellevue, Washington, Chicago, Illinois.

[00:04:43] Avery Smith: So kind of the same answer. A little bit easier to see it in graphical form, right? Versus just seeing this really small table that I have to expand. It was a little bit more automatic from Julius than it was inside of chat, GPT. Now, although this was interesting, the top locations we had [00:05:00] Seattle as number two and Bellevue, which is basically Seattle at number five.

[00:05:04] Avery Smith: So a more interesting question might be what state has the most data jobs listed? And I will ask that to both chat GPT. Once again, chat. GPT wrote the Python script for me, but didn't actually run it. But yes, I would like you to run it and it still just gave me the Python script. So I'll rephrase my question.

[00:05:25] Avery Smith: I'll say, what state has the most data jobs listed based on the CSVI uploaded. Okay, now it said the most common state is actually United States, followed by remote. These aren't actually US states, so the data is a little bit messy here, so we'll need to clean the location data a little bit more thoroughly.

[00:05:41] Avery Smith: Would you like me to clean this? Yes, I would. And this is one of the trends that I'm going to just tell you that I see with chat GBTA lot of the time is it really requires you to kind of hold it by the hand a lot of the time. And do things step by step, although it is pretty fast to actually do the analyzing the analysis, it's pretty slow [00:06:00] because it only does like one little part of the analysis at a time.

[00:06:03] Avery Smith: Only New Jersey appears to as a valid US state with one data job posting in your current data set. Oh, that's really interesting. Would you like me to try extracting? Yes. Try extracting. Okay. Then it looks like we have this cleaned version, which has New York at the top, California next. Followed by Virginia, Texas, Florida, and Massachusetts.

[00:06:26] Avery Smith: Rounding out the top six. Alright, so that's how Chatt PT did. Let's go ahead and go back to Julius and ask it the exact same question. Which state has the most data jobs listed? Let's hit run and let's let Julius think and see what it comes up with. Okay. Julius created a bunch of Python code. One thing I like about Julius is it actually lets me see what the code is right here.

[00:06:46] Avery Smith: Obviously. Let me edit the code, but also lets me see what the code's actually doing with this code explanation. So it's basically reading in the CSV, inspecting the location column, extracting state abbreviations from the location string, counting out the [00:07:00] occurrence of each, uh, in the top 10, so on and so forth.

[00:07:03] Avery Smith: And you can see that we actually got California, Washington. New York, Texas, Virginia, Florida. So quite interesting because we got very different numbers here. We got New York at 34 and New York at 73 here. We got California at 32 and California at 95. I will tell you that I've actually done this analysis before I was recording the video just to make sure I knew what I was doing in both of these tools.

[00:07:26] Avery Smith: I got a completely different answer from chat GBT earlier that looked a little bit more like this, where it also had California and New York, Texas and Colorado, which matches this a little bit more closely, but still not quite. If you remember earlier, we did see that Seattle and Bellevue, Washington were the top two cities, so it makes sense that Washington would be in the top five, and I'm not even seeing, I see Washington at.

[00:07:49] Avery Smith: Seven here. So something's going on and I think I trust Julius a lot more just because this number makes a lot more sense. And we also know that California is probably going to have the most data jobs because [00:08:00] San Francisco, LA, there's quite a bit of data jobs there, followed by Washington, New York and Texas are quite large as well.

[00:08:06] Avery Smith: So I think this makes a lot of sense here. The next question we might wanna ask is, well, what is the trend of job postings over time? I've really just started collecting this data, so it's not going to have a ton of insight for us, but we'll go ahead and see what trends we can find. So here's the trend based over time, I'd argue that this is not a trend.

[00:08:25] Avery Smith: Giving me just a, you know, group eye of how many jobs were posted on that day isn't exactly a hundred percent helpful. Something like a visual chart would be useful. So, yes. Please make a chart and let's see what chart it makes. I'm giving it a lot of leeway right here to create whatever it thinks is best.

[00:08:41] Avery Smith: I'm curious to see what chat GPT comes up with. It looks like we got a line chart here. Created, uh, showing the daily, the daily number of job postings, and we see we have a really big. Jump recently, and that's probably because the way I've been collecting this data. But this is what the, the chart looks like.

[00:08:57] Avery Smith: So Okay. It allows me to see certain trends like, oh yeah, [00:09:00] we're starting to find a lot more jobs more recently. Let's go over and ask Julius the exact same question and see how it does. One thing I like in Julius is they have this reasoning tab over here, which shows you the plan. So we're going to expect the data set to identify that there is a date or a timestamp column that makes sense.

[00:09:16] Avery Smith: We're gonna parse that date column and then aggregate a number of jobs and then create the plot. And it looks like I had a plot created here from Julius, but it looks like it did it by month, which isn't really useful because most of this is from one or two months. So we'll say, let's make the same chart, but.

[00:09:31] Avery Smith: A daily version of it, and this is the truth with these tools right now, is you are going to need to talk through it. It's not going to just do your analysis for you. It's there to be a copilot for you, but it still requires you to kind of think through what's going on in the analytical process. Aha. Now, Julius has done, and it's created that very similar looking chart to what we had in chat EPT, but one of the things I really like about Julius is it spit out some analysis for me.

[00:09:56] Avery Smith: So let's see. It shows that line chart has daily counts [00:10:00] across the whole spend of the dataset. Started, uh, modestly in May of 2025, grows rapidly through late May and then fluctuates day to day, even with the expected daily noise, weekends, batch scrapes, company posting habits, the overall upward momentum that we saw in the monthly views is still evident at the finer daily resolution, and spikes correlated to particularly active posting days, often midweek while dips, align with low activity periods, typically the weekends or on holidays.

[00:10:24] Avery Smith: So that's very cool that it gave us that information versus over here we didn't really get that from chat, GPT. We just kinda got the graph. Another really cool part about Julius is you can go down here to tools and change your chart theme. So right now we have no theme. We could try the Julius theme, we can try the scientific theme, we can try the dark theme, and so that way when it creates a new graph, it'll be all in that dark theme.

[00:10:45] Avery Smith: So let's actually just ask Julius to recreate that line chart with the new theme and see what it comes up with. And there we go. We have our new chart that's really just our old chart, but just dark with our new theme. And that's really nice because you can [00:11:00] actually go in here and uh, create your own theme, your own customized theme right here.

[00:11:04] Avery Smith: And that way your charts can always look the same, which is really important when you work for a company or you're doing any sort of consulting or anything like that. Alright, so another interesting thing that we can do with job listing data is take a look at the salaries. So I'm just going to ask it in a pretty general way.

[00:11:18] Avery Smith: I'm gonna say analyze job salary trends. I'm purposely being a little bit vague to see what it comes up with. I wanna see what it does. So it says, dataset you upload. It doesn't have a dedicated salary column, but it's possible salary info may be embedded inside the description field, which is the truth.

[00:11:33] Avery Smith: But it's possible salary information might be embedded inside the description or title fields, and it is in the description field. So, let's see. Here's what we can do. Extract salary mentions using RegX. Look for things like this. Okay, great. Clean and standardize those salaries to numerical values and visualize trend over time.

[00:11:48] Avery Smith: I'll start by extracting the salary figures from the description column. Here's some of the extracted salary samples from the job descriptions. The formats vary. For instance, 80,050 to 70 $16. So next I can clean and [00:12:00] standardize these into consistent numeric values like average annual salaries to analyze trends.

[00:12:05] Avery Smith: Would you like me to convert these into estimated annual annual salaries, filter out hourly in contract roles, or analyze by job? I'd say. Let's do the convert into estimated annual salaries. And I'm glad that it's doing this, uh, step, even though I wish it would just kind of do it automatically for me. In an ideal world, it would.

[00:12:23] Avery Smith: Uh, but when I did this previously, it skipped the step and thus the analysis was really off. So let's say here are the estimated annual salaries based on the values extracted from the job description. Many of those appear to be hourly rates that we scaled to using a standard full-time schedule. Would you like me to filter out unusual, high or incorrect values?

[00:12:40] Avery Smith: Show the average salary by job title or location. Create a chart of salary distribution. Let's do the create a chart. Now, ideally for this, they would show a histogram. Let's see if they create a histogram. Alright, we got our histogram and we kind of got our skew to the right. Kinda looks like our average is right around like the $90,000 [00:13:00] range it looks like, but we got a good histogram right here.

[00:13:02] Avery Smith: Good breakdown. Some of these jobs pay up to $300,000. Let's go over Julius and ask the exact same thing, analyze job salary trends and see what Julius comes up with. Here's our plan from Julius over here on the right hand side, writing some Python code. All right, Julius just finished up and it said first we extracted salary information from.

[00:13:21] Avery Smith: Titles plus descriptions with RegX to capture values like 120 to 150 and $50 an hour after cleaning. 438 postings contain recognizable pay figure, and I think that's really important that that, it told me how many, so only about a third of the jobs. It also gave me the average salary, which looks like to be 87,000.

[00:13:39] Avery Smith: Which is really cool. It also broke it down by the state it looks like, so it does look quite interesting. It looks like Louisiana had the highest at 131,000 on average, which is quite interesting. And then my Utah was number two and then Texas and New York were uh, a lot lower California as well, which is quite interesting.

[00:13:58] Avery Smith: Now it created the same histogram [00:14:00] as, uh, the previous one did in chat GPT, but it may be found, it went out all the way to 600,000. It might have found a job out there in 600,000 land. So let's go ahead and say create the histogram again, but lower the bin size. One important thing it does say is the highest state averages are often driven by a few generous listings, broader markets like California, New York, and Texas show more stable central tendencies.

[00:14:26] Avery Smith: That makes a lot of sense. Okay. And it looks like we still have some issues. So I would guess that most data jobs are not below 50,000. So I'm gonna go ahead and tell Julius, I think most data jobs pay above. 50 K. So maybe you are reading hourly data as annual data. Alright, and while this is thinking, I just remembered a really cool feature they have, which is called Data Explorer.

[00:14:51] Avery Smith: This actually lets you look at the data frames that Python screening or the tables. So for example, we know that there is about 438 different jobs, [00:15:00] right? And so I can actually just go and open this up and it will show me what the data table that it's actually looking at. I can kinda look at this and try to figure out maybe what's going on.

[00:15:10] Avery Smith: So for example, yeah, look at, there's salary min at 25, salary max 83, salary average 84. And so I don't think that's quite converting, uh, hourly jobs into salary jobs. So I'm gonna go see what it said to mean now. Aha. There we go. It said we re parsed Every listing looked for words, uh, immediately around each number flagged values with hour per hour.

[00:15:33] Avery Smith: If those are multiplied by 2080. Okay, great. For numbers below 500 that did not look hourly, we assume the K was missing. And so like 1,001 20 became 120,000. Okay, that makes a lot of sense. Got it. K. Let's take a look and then here is what our histogram looks like. I still think something's going wrong because the number of data analyst job listings that pay like this is like $10,000.

[00:15:56] Avery Smith: Seems a quite low. Interesting here. I don't [00:16:00] know exactly. I would actually trust this one a little bit more and I'm not sure exactly what was going on with Julius. Now, once again, to be fair, I'm not giving it very specific instructions. I'm just giving it very vague instructions and letting it do most of the thinking, analyze job salary trends.

[00:16:15] Avery Smith: I should probably make a little bit better of a prompt that would actually tell the ai this is what column to look for. This is what to do in this situation. Look for hourly, look for, you know, annually, those types of things. And I'm not doing that in this case. So if I gave it a better prompt, maybe it would do a little bit better.

[00:16:30] Avery Smith: But in this case, I'm actually going to trust chat GPT just a little bit more in this particular time that it, that it generated this chart. Now I will say, when I first did this analysis before recording this video, this is what chat GPT came out with, and you'll see that it had the exact same issue.

[00:16:44] Avery Smith: With a lot of these hourly jobs being seen as annually, and that's just really not the case. So you do have to be careful what's going on when you're using these AI tools. Remember, they're not a replacement for you. They're an aide for help you to do your analysis a little bit faster, but they really should still have you there checking all of their [00:17:00] work.

[00:17:00] Avery Smith: Alright, next question. One of the things we're really interested in is how many senior jobs are there versus how many junior jobs are there? So we can just ask another vague question. Let's say investigate the impact. Experience level, how many junior jobs are there and how many senior jobs are there?

[00:17:19] Avery Smith: And once again, I'm making this really vague and it's gonna be quite hard. 'cause sometimes senior is in the actual title, or juniors in the actual title. Sometimes it's not. It's gonna mean in the description. And other times it might be inside of like a one, a two, a three, or a four. This is what chat PT says.

[00:17:35] Avery Smith: It says, here's the breakdown of the experience level, junior 24, mid 9 8 2, and senior four three. Uh, let's go ahead and ask the exact same question to Julius AI over here. One of the things I really like about Julius AI is you can also switch to any model you'd like at any time. So, for example, we've been using the Julius default right now, which basically Julius decides what model is best for you.

[00:17:56] Avery Smith: I'm gonna go check out, maybe let's go to Claude four on it [00:18:00] because I've heard really good things about it and test it out for this question. All right. Julius finished running it did it quite a bit for us, which is one of the things I really like about Julius is it does more than than you ask it to.

[00:18:11] Avery Smith: You don't have to like. Ask it a hundred things to get a hundred things done, like it automatically does these types of things for you, basically. So it looks like we have junior at 31, senior at 360 2 and other at 10 51. We have the salaries broken down, which is really interesting. That senior is somehow lower than junior, which I didn't really expect.

[00:18:30] Avery Smith: And then we have some of the titles right here. So we first tag each post by scanning its title for a common thing like senior, senior lead, principal, staff, and then junior ones like junior, junior entry and associate. Another thing that we looked into is other. So there is 31, junior level 3, 62 of the senior and a thousand that did not really a hundred percent say senior roles, surprisingly.

[00:18:51] Avery Smith: And it says the word, surprisingly here, average a bit lower because a handful of non-US, ah, non-US and low salary titles dilute the mean medians are higher and more [00:19:00] in line with expectations. See, not only did it tell me that this was surprising, but it explained why it happened, and I think that's really interesting for a spot check.

[00:19:07] Avery Smith: The last table shows the first five rows of our annotated data, and I love that it's checking. It's letting me check the data for it. So that is really interesting right here for me to look at this data. So, very good job by Julius here. I think Julius kind of won this one. I found this really interesting and if I wanted to share it with you guys, I have two options on chat, GPT.

[00:19:26] Avery Smith: You of course can just share any thread. And make it to the public. Right on Julius, you can do the exact same thing by clicking share and making this thread public. But I also can do team sharing, which is really cool that I could share, you know, just within my organization, within my team. It doesn't have to be public, but my teammates could see the same analysis, which I think is a great feature that Julius has.

[00:19:44] Avery Smith: Now. Any part of a good data analysis is of course, looking at what no values or missing data you have. And to do that, I really like a Python library that's called Missing. No. So I'm going to ask chat GPT in this case to, can you use Missing No on this data set and [00:20:00] see what it says? It's basically gonna tell me what percentage of my data is missing for each one of the columns.

[00:20:04] Avery Smith: What you'll see that chat GPT says is the missing. No library isn't currently installed in this environment, so I can't use it here, but I can still help with things like pandas and map plot lip. So would you like me to help create a heat map or a bar chart showing which columns have missing values and how much is missing?

[00:20:18] Avery Smith: The interesting thing is I'm going to ask the same thing to Julius, and Julius is able to, even though that that might not be the most used package of all time, it's able to install the package in the environment and actually use it. This basically means that whatever Python package you can use in the wild, it's in the real life on your local machine you can use inside of Julius as well.

[00:20:37] Avery Smith: That's not true of Chachi PT. Basically, Julius gives you higher control. Your environment. Not only can you do things like PIP install Python packages that aren't currently installed on the environment, you can actually go to the settings right here, go to expand resources and take a look at what container you're using, how much ram, how much CPU, and basically it's going to, you can go to expand resources and see what type of container you're [00:21:00] using.

[00:21:00] Avery Smith: So I'm currently in a memory boost right now with 32 gigabytes of ram, which is super judging my analysis with large files because I have a lot of power in this ram right here. Julius also gives you a higher CPU than chat GPT gives you, so your analysis should be a lot faster on Julius. Alright, and there you go.

[00:21:18] Avery Smith: Missing Node. Just ran and created this chart. My category column is the only one that has missing data, so I can see that really easily right here. In chat GPT, I'd have to ask it to do a little bit extra stuff. So in this particular case, I think Julius won this example. Now, since this is a test, this is an experiment.

[00:21:34] Avery Smith: I do wanna ask a very vague question. I wanna say, tell me an interesting trend in this dataset. I'm just gonna go let chat GPT, go through this dataset and see what it thinks is interesting to tell me. Here's an interesting trend. Remote data jobs have consistently appeared alongside non-REM remote ones, and their volume holds strong month to month.

[00:21:52] Avery Smith: This suggests that remote data roles remain a significant portion of the job market, not just a temporal shift. Want to explore which companies or states posts the [00:22:00] most job remote jobs. So this shows me, if you can see, this shows me a chart of remote jobs, this yellow versus the non rere remote jobs red.

[00:22:08] Avery Smith: And honestly, I don't really get what it's saying here. It's just saying that they've increased both since, you know, April. I guess they were kind of the same back in January and February. I'm not exactly sure what this is telling me and how much I actually like it. Let's go ahead and ask Julius the exact same vague question right here and see what it comes up with.

[00:22:28] Avery Smith: Okay. Interestingly enough here, Julius kinda came up with. The same thing across the period captured in the file. The proportion of data related jobs advertised as fully remote rises month over month. How this was derived, what we see, why it matters. Very interesting that it both did the, uh, remote jobs.

[00:22:45] Avery Smith: Now I will say that this graph is not the best. It is creating way too much of a x uh, line chart here. I haven't been running this job board for longer than a year, so really only going back to maybe November of 2024 would've been the right move. That data's not even in this particular data set. So, [00:23:00] uh, not the best graph of all time, if I'm being honest here.

[00:23:02] Avery Smith: But it's interesting that they both chose the same thing. When I ran this before I started recording, this is what chat GPT gave me, basically a word cloud art of the most common keywords in job titles, which is United States Data Analyst, senior Remote Canada, Washington, Amazon Intelligence, California.

[00:23:20] Avery Smith: This is not very interesting, I don't think this is a very interesting chart, and I think it did, uh, very poorly. When I originally asked Julius ai, this is what it created for me, it showed me that most of the job titles inside of this data set were financial analysts and then data specialists, and then marketing analysts, so on and so forth.

[00:23:36] Avery Smith: And then it showed me, you know, out of these different categories like Tableau and data specialists and data visualization, what percent of those jobs are remote? And so it was really interesting to look at Tableau, the only 22% of those jobs. Were remote. Uh, and data visualization was 16 something like financial analyst was only seven.

[00:23:52] Avery Smith: So it looks like if you're in data viz, you have a higher chance of working remote than if you would be as a financial analyst. I thought this was very interesting. In this particular [00:24:00] case, I thought Julius did a lot better job than chat GPT did. But for some reason it decided to do some different analysis this time.

[00:24:07] Avery Smith: Another really neat thing about Julius is not only does it have threads, kinda like this chat interface, but it has this notebooks interface as well, which is really cool. The Notebooks interface allows you to do more repeated analysis if you're going to do this multiple times. So for example, maybe I'll go ahead and connect to my job, my data job listing CSV here, and then you can basically tell it to identify top job locations.

[00:24:29] Avery Smith: I'll say identify top jobs locations right here. And it's going to go ahead and create the code to do that. And then the cool thing is I can run that code whenever I want. Maybe I wanna do it every single month. Maybe send it out to my newsletter or send it out, post it on LinkedIn so all of you guys can see.

[00:24:43] Avery Smith: That can be easily repeated by just clicking this with the updated data. Now next month I can just come back in and hit Run All, and it would just do that. You can also export this as a CSV or export it inside of Google Drive, which is another cool thing is not only can you upload things like a CSV, but there's all different sorts [00:25:00] of connections that you can do.

[00:25:01] Avery Smith: For instance, you can connect to Google Drive, BigQuery, snowflake, or Postgres. Now going back to my thread or my chat conversation here, I can look at everything I did inside of Julius, inside of this outline right here, which basically creates a step-by-step list of everything I did. So, for example, this is when I was creating the trend analysis right here.

[00:25:21] Avery Smith: This is when I was doing the missing. No part of my data set right here. Really clear history of everything that I've done. Also, if I wanted to, I could have switched from Python and actually done this analysis in R versus chat. GPT only knows Python at this point. So which tool should you be using? Well, I think it really depends on your task and what you're trying to do.

[00:25:40] Avery Smith: Obviously, chat GPT is kinda like a Swiss Army knife. It can literally do anything. It can create images, it can write screenplays. It can do data analysis. Julius AI is more like a hatchet or something that you're going to chop down a tree with, right? It's something very specific that it does very well. I probably wouldn't write a movie script inside of Julius AI 'cause it's really focused [00:26:00] on data analysis.

[00:26:01] Avery Smith: One of my biggest concerns when I use AI tools for data analysis is just knowing what's going on. And I do think Julius does a better job of actually explaining what's going on. It gives you these little explanations afterwards on every single code it, it pops out. It gives you this code explainer. It gives you the data explorer to see all of the different data frames and tables It's created so you can actually visually explore the data, and for the majority of the time, it does more than I ask it to, which I really enjoy.

[00:26:26] Avery Smith: I don't really like having to babysit and be like, okay, now do this step, do this step. Julius AI is a lot more ambitious in trying to do all of the analysis you could possibly want in one go. It also is a little bit more customizable for you and your team. We talked about how you can share it with your team.

[00:26:41] Avery Smith: How you can create your own chart themes or choose your chart theme for consistent graphs every time you create a graph or how you could ramp up your RAM and have just faster analysis in general. So overall, I think if I'm going to be analyzing data, I'm probably going to be using Julius AI because I think it just does a little bit more, a little bit more quickly [00:27:00] than chat GPT.

[00:27:00] Avery Smith: But let me know what you think. I'd love for you to try the same experiment, grab a data set, create a chat GPT account, create a Julius account, and contrast and compare. Which one do you like more? If you need help finding a data set, you wanna check out this episode right here, and then let me know in the comments how your experiment went.

[00:27:18] Avery Smith: As always, thanks for watching.