YT YouTube2Text

YouTube2Text → Transcripts → AI For Data Analysis In 21 Minutes

AI For Data Analysis In 21 Minutes — Transcript

by Tina Huang · 5,098 words · 730 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00I learned how to do data analysis with
0:01AI for you. I guess we can call it vibe
0:04analyzing. But really though, I took 11
0:07courses on this topic. What can I say? I
0:09used to be a data scientist at Meta. I
0:10love data. I use data every single day.
0:12I'm going to save you the time and money
0:14that I spent buying these courses and
0:15give you the cliffos version of what I
0:17learned. As per usual, there'll be
0:19little quizzes throughout this video.
0:20So, pay attention. All right, let's go.
0:22A portion of this video is sponsored by
0:24LTX2. The outline of today's video is
0:26first I'm going to cover when is it
0:28useful to use AI for data analysis. Then
0:31we'll talk about the dig framework for
0:32how to approach analysis. But of course
0:34to make it all concrete we need some
0:36examples. So I'll then be showing you
0:37lots of examples. And finally I'll
0:39[music] explain how to take this even
0:40further and take your data analysis and
0:42build it out into dashboards or even AI
0:44applications. [music] I do want to make
0:45a note that a lot of the courses and
0:46examples are focused on using chatbt as
0:49the tool for data analysis. But that is
0:51not the case that you have to use
0:52chatbt. In fact, you can switch it off
0:54at Gemini and Claude would work the same
0:56way. In fact, sometimes they would
0:57actually work better. So, don't feel
0:58like you need to be married to a single
0:59tool. And I'll actually call out if
1:01there is a tool that I think would work
1:03even better. Okay, let's start off with
1:06when we should be considering using AI
1:08for data analysis. Well, from the course
1:10Chachi PT advanced data [music]
1:11analysis, Dr. Jules White, the
1:13instructor from Vanderbilt University
1:15has an acronym for this called achieve.
1:17He explains that there are five
1:19different areas that is useful for using
1:21AI in data analysis. Aiding human
1:23coordination, cutting out tedious task,
1:26help provide a safety net for humans,
1:28inspire better productivity and problem
1:30solving, and enable great ideas to scale
1:33faster, achieve. Aiding human
1:35coordination refers to helping people
1:37work better with each other because
1:39people actually tend to be quite messy,
1:41you know, and there's a lot of
1:42miscommunications. There's a lot of like
1:44back and forth between people. So,
1:46there's a lot of room for improvement
1:47here that AI can help with. Say, for
1:49example, you're in a meeting with a
1:50bunch of people and you have this
1:51meeting transcript. You can actually put
1:53it into AI and say, "Actal assistant.
1:56Read the following meeting transcript
1:57and provide me a summary of the key
1:59points of discussion." This is just an
2:00example. I'm sure you can think of a lot
2:02of other scenarios where there is a
2:04bunch of data that can be analyzed such
2:06that you can provide more clarity for
2:08humans. The second part of the framework
2:09is to cut out tedious task. Just as that
2:12suggests, it's best to let AI to be able
2:14to do things that are very repetitive
2:15and boring for people. For example, say
2:17you're hosting a workshop and you ask
2:19people to sign up for the workshop and
2:21provide different types of information
2:23like what their name is, what their
2:24occupation is, which department that
2:26they're in, what their interests are.
2:27Instead of having to go through this and
2:29manually analyze it, you can tell the AI
2:31that this is the list of people that
2:32registered for my workshop on prompt
2:34engineering and chatbt. Describe the
2:36data in this file. By the way, asking AI
2:38to describe data is is best practices
2:40which we'll cover a little bit later.
2:42But yes, so the AI will be like, okay,
2:44like you know, this file contains this
2:46type of information. It's a CSV file and
2:47it contains like timestamp, name, the
2:49email, the department, um this is a
2:52university workshop, how they're using
2:53chatbt and their tools already, the role
2:55that they hold, etc. You might notice
2:57that people are filling their department
2:58names in a lot of different variations.
3:00So you can actually ask the AI, there
3:02seems to be a lot of overlap between
3:03departments with alternate spellings.
3:05Can you list out all the departments and
3:06then do some intelligent grouping of
3:08them? And then you can ask it to create
3:09a bar chart showing the total number of
3:11registrations per department. This is
3:13the kind of data cleaning and
3:14visualization that is pretty mundane and
3:16AI is able to do this much more quickly.
3:19Third part of the framework is to help
3:20provide a safety net for humans. You see
3:23people often say like oh AI has a lot of
3:26hallucinations and that is true. AI does
3:28hallucinate but people hallucinate too.
3:31People make a lot of mistakes, like some
3:33really really dumb mistakes. I make dumb
3:35mistakes like literally constantly.
3:36Wrote my name wrong on a form yesterday
3:38for example. So that is why having AI as
3:41a backup is actually a really great
3:43idea. Say for example, you're on a
3:44business trip and you need to ask for a
3:46reimbursement. So you need to like
3:48generate some invoice thingy and then
3:49make sure you have all the fields
3:50covered. If you're anything like me, I
3:52am not very detail- oriented. I probably
3:54will make a really dumb mistake. So, as
3:55a safety net, you can actually upload
3:57this invoice into the AI along with the
4:00business expense policy and ask read
4:02each page of the attached business
4:03expense policy and see if the attached
4:05receipt complies with it. So many other
4:07examples of this. Every time you need to
4:08submit like an insurance claim, you need
4:10to like read some sort of document uh
4:12looking at like travel policy whatever
4:14like yeah so many examples of this. The
4:16next part of the framework for when to
4:18use AI for data analysis is the IEV
4:21which is inspire better problem solving
4:23and creativity. This is another thing.
4:25People always feel like AI is going to
4:27make people less creative and is going
4:29to take over creative things, but that
4:31is not the case. It's all about asking
4:33the right questions. Say, for example,
4:34you have a PowerPoint presentation that
4:36is really, really important. You can
4:37actually upload those slides into AI.
4:39Ask them to quickly summarize it and
4:41then ask it to act as a skeptic of
4:43everything I say in this presentation
4:45and find flaws, my assumptions,
4:46assertions, and other key points and
4:48then generate 10 hard questions for me.
4:50This is a way for you to actually force
4:53yourself to think about the questions
4:54that you could potentially be asked and
4:56come up with better ways and better
4:57solutions for answering them. People are
4:59actually very rigid creatures. We tend
5:01to think in a very specific way and it's
5:04very hard for us to actually expand past
5:06that. So using AI as a tool to help us
5:08expand our creativity is is actually
5:10really really helpful. And finally the
5:12last part of the framework for when you
5:14should be using AI for data analysis uh
5:16is the E, which is enable great ideas to
5:18scale faster. Let's go back to that
5:20workshop example. You're doing a
5:22workshop on prompt engineering and you
5:24have people coming from all types of
5:25different backgrounds um who are
5:27interested in all types of different
5:28things and all types of different
5:29levels. After you give the AI the signup
5:31form and analyze all the data about your
5:33participants, you want to create a cheat
5:35sheet for each of them that is most
5:37relevant to them. What you can actually
5:38do is ask the AI to map each attendee
5:40with their specific domain of interest
5:42and then generate a column called ideas
5:45that includes the corresponding
5:46idea/prompt
5:48um to put for their cheat sheet. This
5:49way now your CSV file for each attendee
5:52also contains a very specific prompt
5:54idea that is specific for that attendee.
5:57And then after the workshop, you can
5:59actually send them an email with their
6:00specific little cheat sheet. I'm sure
6:02you can see prior to AI, this would have
6:04been so hard to do if you have more than
6:05just like 10 attendees to be able to
6:07come up with like a custom cheat sheet
6:09for each person. So, whenever you're
6:10thinking about if you should use AI to
6:12do a certain analysis, you can think
6:13back to this acronym. Of course, I
6:15haven't yet covered exactly how it is
6:17that you should be approached to these
6:18analysis, which is what I'm going to be
6:19covering next.
6:23But first, let's do a little quiz.
6:25Please put your answers in the comments.
6:26This portion of the video is sponsored
6:28by LTX2, the new AI video engine for
6:30creative workflows. And this one
6:32honestly blew me away. What stood out to
6:34me isn't just the quality, it's how LTX2
6:37can finally tell a story. And if you
6:39can't guess by my career choice,
6:41storytelling is is what I live for. Most
6:43AI video models just give you short
6:45looping clips. A few seconds that look
6:47great but don't really say much. LTX2
6:50though can generate up to 15 seconds of
6:52continuous video with synchronized audio
6:54which means full monologues.
6:56>> Super villains also have feelings.
6:59>> Dialogues or even short scenes with
7:02music and natural sounds. [music]
7:07It's smooth, coherent, and feels
7:10cinematic. The kind of storytelling
7:12range that's missing from [music] AI
7:13video tools. As someone who's been
7:14creating content for over 5 years now,
7:17this one really impresses me. You can
7:18build short narratives, generate B-roll
7:20and snippets, or even cinematic
7:22transitions, all with a single prompt.
7:24You can now try out LTX2 to create your
7:26own AI videos. The link [music] is in
7:27the description. Thank you so much LTX2
7:29for sponsoring this portion of the
7:30video. Now, back to the video. From the
7:32course, ChatBT Plus Excel, master data,
7:34make decisions, and tell stories.
7:36[music] There is a nice little acronym
7:38for how it is that you should be
7:39approaching data analysis [music] using
7:41AI called DIG, which stands for
7:44description, introspection, and goal
7:46setting. By the way, if you do have a
7:47little bit of background in data, like
7:49data science or data analysis, um it's
7:51basically the same as EDA, exploratory
7:53data analysis, but specifically for
7:55using AI. The first step of dig, which
7:57is describe, it's a way for you and the
7:59AI to explore the data together. [music]
8:01This step is very important because it
8:02helps both you and AI gain a familiarity
8:04for the data and to also notice if
8:06there's any issues with that data. This
8:08is going to help a lot with
8:09hallucinations and issues [music] down
8:11the road. So very similar to normal EDA
8:13processes, after you upload like a
8:15spreadsheet or whatever data it is that
8:17[music] you give to the AI, like let's
8:19just say like a spreadsheet in this
8:20case, you would ask the AI to list out
8:22the columns in the attached spreadsheet
8:24and show me a sample of the data in each
8:26column. For example, if your spreadsheet
8:28contains data about different roles um
8:30and different salaries, [music] the AI
8:32would be able to output and say, "Here
8:34are the columns for your spreadsheet
8:35along with a sample of the data for each
8:36[music] column." So the column name you
8:38could have like salary ID, job ID, max
8:40salary, med salary which is median
8:43salary, min salary, pay period, currency
8:45and compensation type. You can already
8:47notice that under max salary and min
8:49salary you have nan which is not
8:51available. This is important to note
8:52because hallucinations tend to happen
8:54when you have things like missing data
8:55or incorrectly formatted data. So if you
8:58see something like this, the first thing
8:59you actually want to do is to confirm
9:01that is it just that DAI [music] is not
9:03parsing your data correctly. So you
9:04actually want to go in and see like is
9:06it actually not available or is there
9:08like a parsing problem? And if there
9:09actually is a parsing problem, you want
9:11to then tell the AI, hey, you are
9:13parsing this incorrectly. This is
9:14actually how you should be parsing it.
9:16Or maybe the data is just not available.
9:17Then you just want to make a note of
9:19this for the future. We will get back to
9:21that. [music] But first, you actually
9:22want to do a few more random samples.
9:24Just ask your AI like take a couple more
9:26random samples of the data for each
9:28column. Make sure you understand the
9:29format and type of information in each
9:30column. What we're basically doing here
9:32is verifying that the data is being
9:34parsed correctly and the AI has correct
9:37understanding of each of these columns.
9:39You can even ask it what do you think
9:41each of these columns represent and it
9:42might tell you that salary ID appears to
9:44be a unique identifier for each entry.
9:46Job ID is likely unique identifier for
9:48jobs or position max salary, min salary
9:49and med salary not all entries have
9:51complete salary data etc etc. So this is
9:53a way for you to validate that the AI
9:55understands what's happening also that
9:57you understand what's happening. The
9:58best way to think about this is that
9:59your AI is a very competent but still
10:01very junior developer or data scientist
10:03or data analyst. So you need to make
10:04sure it understands what is the data
10:06that it's actually receiving. Otherwise
10:07any analysis that you do on top of this
10:09could potentially be wrong. So after you
10:11do this, you want to move on to the next
10:12step of the dig framework which is
10:13introspection. This is when you want the
10:15AI to start looking at the data that it
10:18finished describing [music] and think
10:19about the patterns and relationships
10:20that exist in the data. This is also
10:22another great way to catch any
10:24misconceptions that the AI may have.
10:26Notices were just being like very
10:27skeptical all the time. Very important.
10:29You can just ask tell me some
10:30interesting questions that could be
10:31answered with this data set and why they
10:33would be interesting. And it might come
10:34with some questions like is there a
10:36relationship between compensation type
10:37for example base salary and variability
10:39in salary ranges. It's saying why it's
10:41interesting is that understanding
10:42whether certain compensation types like
10:44bonuses or equity are more likely
10:46associated with higher salary ranges can
10:47help employees and employers make more
10:49informed decisions about how they
10:50structure pay packages. Here's a
10:52question that came up with that could be
10:54a red flag. The question is, are there
10:55any noticeable patterns in salary data
10:57for different currencies if additional
10:58currencies exist? So, when you see this,
11:00you want to think to yourself, oh, like,
11:01is there actually other currencies that
11:03exist? And you might want to double
11:04check yourself to see if if there's
11:06actually data that is not USD. In this
11:08case, there actually are no other
11:09currencies. So, that's when you want to
11:11tell the AI all currencies are actually
11:13listed in USD. [music] So, it knows that
11:15information moving forward. If you catch
11:16your AI asking questions like these or
11:18you're catch your AI like asking
11:20questions that you know cannot be
11:22answered by the data set and ask you to
11:24generate some more questions that can be
11:25answered from the data set. In this way
11:27you're really helping the AI and
11:28yourself make sure that you really
11:30understand what's happening in the data
11:31set. It's also actually quite a good
11:33exercise because sometimes AI will come
11:34up with different things and different
11:36analyses that you might have not have
11:37thought of doing. I know this might seem
11:39a little tedious and you just want to
11:40skip to like making graphs and charts
11:41and doing analysis, right? But do not
11:43skip [music] these steps, okay? Trust
11:44me. Because like that's the thing with
11:46data. It's one of those things where if
11:48you mess up, it will just propagate
11:49throughout your entire analysis. So it
11:51is really worth the effort to actually
11:52make sure that everything is understood
11:54correctly. And this is actually the same
11:56with humans too. It's not like an AI
11:57problem. Even when I was a data
11:59scientist, [music] I spent a significant
12:01amount of time doing exploratory data
12:02analysis, making sure that I understood
12:04exactly what was happening in data and
12:05clarifying all that information as well
12:07because I knew if I just like rush into
12:10things, I'll probably end up making a
12:11mistake and be very embarrassed when
12:13someone catches my mistake or worst case
12:14scenario get fired because I made a
12:16really dumb mistake and then, you know,
12:17lost the company a lot of money or
12:18something like that. Anyways, so after
12:20you do this, the third step of the
12:21[music] framework is goal setting. It's
12:23very important that your AI understands
12:24what it is that you're trying to
12:26achieve. Like if you just went like
12:27analyze this data, your AI is going to
12:29be like what what the heck like you know
12:30what does that even mean? What am I
12:31supposed to analyze? What's the result?
12:33So it's the same. You got to be like
12:34very clear about what the goal actually
12:35[music] is. So you could tell the AI, my
12:38goal is to answer a couple of these
12:39questions, the questions that you know
12:41it it [music] generated previously and
12:43turn them into a really exciting
12:44interesting report to post on LinkedIn.
12:46This really helps provide context
12:48because then your AI is able to do this
12:50analysis and then is able to do things
12:52like give you LinkedIn ideas how it is
12:54that you can put it together [music] in
12:55the form of LinkedIn. This is going to
12:56be very different if you were actually
12:58analyzing this data [music]
13:00in order to, you know, like do something
13:02very serious like generate like a report
13:03for your boss. This is also just part of
13:05good prompt engineering practices. So,
13:07if you do feel like you want to brush up
13:09your prompt engineering a little bit to
13:10be able to be more clear about what it
13:12is that you want, I do recommend that
13:13you check out a video that I have over
13:14here which covers the foundations of how
13:16to do good prompt engineering. So, check
13:17it out over here. Anyways, whenever you
13:19have a data set that you want AI to help
13:20you analyze, it would be really helpful
13:22for you to go through this dig
13:24framework. It is a really great
13:25foundation and you can build on top of
13:26this as well. [music] Now all the stuff
13:28that we talked about earlier uh is
13:30pretty standard for if you're doing any
13:31type of data analysis using like Excel,
13:34Python, SQL or whatever. It's just like
13:36maybe more convenient doing in a
13:37conversational fashion with AI. But
13:39there are certain things that you can do
13:40with AI that would be extremely
13:42difficult for you to do just using these
13:44traditional tools. Like for example, if
13:45you're job hunting right now and you
13:47have access to this data set, you could
13:48be thinking like, oh, like I'm looking
13:50for a job that is between like 50 to
13:52$80,000 based on the East Coast and
13:55specifically works with [music] wood. I
13:58don't know, something like that, right?
13:59So in this data, there is no specific
14:02section that's like works with
14:03wood/notwork with wood, you know, or
14:05[music] like materials that you're
14:06working with. And it also doesn't
14:08specify like is it east coast or west
14:09coast. It's just like the location like
14:11Chicago, [music] right? But because of
14:12Genai's capabilities, you're able to
14:14like filter through this data um in a
14:17way that's far more intelligent to be
14:19able to find the rules [music] that you
14:21could be potentially interested in. This
14:23would be so hard to do if you didn't
14:24have Genai. And later on in the video, I
14:26have a lot more examples which I'll show
14:28you uh like Genai specific really cool
14:31data analysis things that you can do.
14:32There's also one more thing that I
14:33thought was really cool in this module
14:35of the course, which is the idea of
14:36traceability and replication. I thought
14:38this was like super clever because a
14:40major issue that people face when doing
14:42traditional data analysis is that they
14:44would like come up with some sort of
14:46thing and then it would be stuck in like
14:47a Jupyter notebook or like whatever and
14:49it would actually be very difficult for
14:51other people to reproduce that analysis.
14:52[music] But with AI, you can actually
14:54ask AI to come up with a traceability
14:56document that allows other people to be
14:58able to perform the same analysis to
15:00validate the results. You [music] can
15:02ask let's create a traceability document
15:04to make sure that others can one know
15:05what data was used two how the analysis
15:08was performed and three threats to
15:09validity. We want a guide for someone
15:11else to be able to replicate and know
15:13the limitations of the analysis. You can
15:15you can save this traceability
15:16information as like a readme.md. Uh
15:18don't worry if you don't know what that
15:19is. It's just very common for software
15:21engineers to be able to store it this
15:23way but you can kind of store it like a
15:24word document whatever doesn't actually
15:26matter. And then for each analysis and
15:27visualization, you can actually ask it
15:29to write a [music] single Python script
15:31that performs the full analysis to
15:33produce the visualization and the
15:34results. I think it's a really really
15:35clever idea and really smart thing to do
15:37if you're doing any type of data
15:39analysis using AI.
15:42All right, time for our next little
15:43quiz. Please answer the questions on
15:45screen in the comments. Yay. Okay, we
15:48can move on to some examples. I'm really
15:50excited for this section. First example
15:52is super simple but is actually really
15:53helpful is it's pretty much like any
15:56type of small document you can just
15:58directly upload do dig on it and then
16:00have it proceed to analyze that
16:02information and transform it in whatever
16:03way like if you have structure data like
16:05say you have a CSV form that has like
16:07all the different types of inventory
16:08throughout the past few months you can
16:10ask it to filter it for different types
16:11of inventories you can ask it what are
16:13the trends in the inventories over time
16:15are there certain items that are
16:16becoming more popular certain items that
16:18are becoming less popular maybe you want
16:19to like remove those if there's too much
16:21of that in the inventory. You can even
16:22ask to come up with a predictive model
16:24to see what are the inventories that you
16:26should be stocking for the next few
16:27months so that you're able to optimize
16:29[music] the amount of inventory so you
16:31don't have too much and you also don't
16:32have too little. You can also do
16:33visualizations both static
16:35visualizations. You might want to make a
16:36bar [music] chart that's ranking all the
16:38different types of inventory that you
16:39have. You might want to make a time
16:40series graph over time. What are the
16:42changes inventory? And you can also make
16:44dashboards as well, interactive
16:45dashboards. for this specific one. Cloud
16:47as of [music] the filming of this video
16:49um is a lot better than the other models
16:51to create interactive dashboards. It
16:53also usually is the best at writing the
16:55code as well in order to do the analysis
16:57and it tends to hallucinate less. Again,
16:59this is at the time of this filming,
17:00[music] so I don't know if this is going
17:01to change in the future, but just FYI,
17:03AI data analysis with different media
17:05forms is also a really cool application.
17:07For example, you can have like a video
17:08and you can ask it to extract 10 [music]
17:10frames from this video evenly spaced out
17:12from 1 second apart. Then you can ask it
17:14to take these images, resize the images,
17:16make it like 300 pixels wide and say
17:18convert it to grayscale and increase
17:20contrast by 30%. You can ask it to do
17:22things like combining the images to
17:24animated GIFs that flip the to the next
17:27image at 1second intervals. You can ask
17:28you to turn the images into PowerPoint
17:30presentations and then catalog all of
17:32the images into a CSV file with the name
17:35of the image and the name of the movie
17:36file that the image was extracted from
17:38and the operations applied to it. So
17:39really really cool that you can
17:41manipulate uh multimedia using AI. Prior
17:44to AI, this would have been so difficult
17:46to do. Another example of data analysis
17:48using AI is by automating things using
17:50zip files. Zip files are very, very
17:52convenient and they're amazing because
17:54not only are you able to put a lot of
17:56different files into a single zip file,
17:58you're also able to maintain folder
18:00hierarchy in the zip files themselves,
18:02which means that you can actually zip
18:04together a bunch of different files
18:05together and ask the AI to mass analyze
18:08them all together. So you can have
18:09multiple Excel files that you're telling
18:10it to combine and search and do whatever
18:12with it. Then afterwards, you can build
18:14all back together and then send it back
18:16to you. Also, Grave, if you need help
18:17organizing different files, you can ask
18:19the AI, one, I want you to help figure
18:21out what is in them by opening and
18:22reading each one to create a summary.
18:24Two, I want you to propose a folder
18:26structure that would better organize the
18:27files. Three, I want you to propose
18:28better naming for each file using just A
18:30to Z and 0 to9, keeping the extensions.
18:33And four, when you have all of this
18:34done, show me your proposed folder
18:35structure and name. Then it's going to
18:37do that. And then finally, once you're
18:38happy with it, you can ask her to zip
18:40everything up again and then send it
18:41back to you and and voila, everything is
18:43well organized and beautiful. And the
18:44final example I'm going to show from the
18:46course is a little bit more advanced,
18:47but so cool. This is when you can
18:49actually turn conversations into
18:52software programs. Let me explain. So,
18:54say for example, you have like a
18:55sequence of analyses that you did,
18:56right? Like for example, maybe you have
18:57like some sort of movie and then you ask
18:59it to get like 10 frames from this movie
19:01spaced 1 second apart. uh maximized it,
19:04I don't know, like did some photo
19:06manipulations on it um and then combine
19:08them together and then generate some
19:10descriptions for it and then put them
19:11all together into a CSV [music] file.
19:13You can then ask the AI, turn this
19:16process into a Python program that I can
19:18download and run on my computer and
19:20provide the path to the documents as
19:22command line arguments. Zip up the
19:24program for me to download and then it
19:26can literally go and actually like write
19:28a script that performs all of [music]
19:30these different steps and then put them
19:32all together in an executable program.
19:35That is so cool. You can literally do
19:37this for any sequence of analyses that
19:40you do. You can just like automate it
19:41like that. At least for me, that blows
19:44my mind. That is so cool.
19:48[music] All right. All right. I can
19:49literally go on forever, but I'm gonna
19:50stop for this section for now and I'm
19:53going to put the next little quiz onto
19:55the screen. Please answer these
19:56questions and put them in the comments.
19:58Okay. So, I wanted to include this final
20:00section because I wanted to make sure
20:02that you understand that just doing the
20:04analysis using AI, you don't need to
20:06stop there. There's actually so much
20:07more that you can you can do on top of
20:09that. From the examples that we already
20:11seen, we [music] can take this data
20:12analysis and then use it to generate
20:14emails, use it to make like social
20:16posts, use it to generate reports,
20:18PowerPoint slides, even build software
20:20programs and dashboards. But that's not
20:22all. You can even build full-on
20:24applications based upon these analyses.
20:26And no, you don't actually need to know
20:28how to code. You can just use pipe
20:29coding. Say you've analyzed a lot of
20:31traffic data. You can actually make this
20:32into an application that analyzes
20:34real-time traffic data and then like I
20:37don't know gives alerts to people. uh
20:39where generates reports based upon
20:40traffic incidents. You can have
20:42application that's able to take videos
20:44and blur out people's faces or like
20:46different identifications within the
20:48video. Here's an example of an
20:49investment research AI agent uh that
20:51people who join our AI agents boot camp
20:53will build that has an entire database
20:55with information about investments and
20:58it has this interface where the user is
21:00able to ask it specific questions and it
21:02will analyze that data to generate
21:04certain types of responses,
21:05conversations and reports. Yeah, there
21:07is so much that you can do. Data truly
21:09is power and being able to analyze and
21:11harness that power by using AI just
21:14opens up so many possibilities. All
21:16right, [music] I'm going to stop here.
21:17If you do want to dive into how to
21:19actually build out these like
21:20applications, agents, and things like
21:21that, I'll link a few videos in the
21:22description that you can check out that
21:24goes into a lot more detail about how to
21:25do this. But for this video, I'm going
21:27[music] to end it for now. I really hope
21:28that this was a very helpful video for
21:30you and you have lots of ideas for how
21:32to analyze your data now using [music]
21:34AI. Vibe data analysis. As promised,
21:37here is the final little assessment.
21:39Please answer the questions on screen
21:40and put them into the comments. Thank
21:42you so much for watching until the end
21:44of this video and I will see you in the
21:45next video or live stream.

About this transcript

This page contains the full transcript of AI For Data Analysis In 21 Minutes by Tina Huang, generated from the public captions YouTube serves with the video. The transcript has 5,098 words across 730 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.