How I use ChatGPT as a Data Scientist?

in the future there are going to be two types of  people one people who know how to use AI smartly   two people whose job will be automated by AI  I definitely don't want to be in the second   bucket and neither do you so in this video I  am going to share how I have been adopting AI   specifically Chat GPT as part of my daily job  as a data scientist so you can learn from it   the tips and hacks that I'm going to share can  be applied to any jaw family including coders   data scientists data analysts software Engineers  so keep watching for those who don't know Chad   GPT is an AI bot which basically talks to you  like a human and I've been using it as a tutor   and a personal assistant chat GPT and its  competitor like Bard are not going anywhere   so it's about time that we get smarter than all  of these chat Bots a quick disclaimer don't put   any of the work information that is intellectual  into chat gbt check your companies and schools   guideline before using chargpt it goes without  saying be responsible when using charge GPT   so in this video I'm going to share four ways you  can use chat GPT as a data scientist number one   coding for coding I've been using it primarily for  two use cases one is to write new code and second   is to debug so for writing new code chat GPT does  a really really good job for writing simple code   for simple questions it is not that great with  the complex problem and more logical thinking   but it does a really good job of giving you simple  code Snippets that you can use in your work for   example let's say I'm working on a project where  I have to create an A B test here I want to create   a sampling group of control and treatment groups  two groups with A and B and I want to have equal   sample size in both groups with randomization so  here I can ask chargpt to write a code for me that   will help me do stratified sampling create sample  A and B group stratified on Revenue so basically   here it started writing the code for me it's using  the train and test split which is pretty common in   machine learning data sets so here in this example  I know what exactly looking for I understand how a   b tests work and then I just get extra help from  chat DPT to write the code obviously I can write   this code on my own but I want to use that time  somewhere else so I can use chargeability here to   kind of help me write this code faster the second  area in coding is debugging so I've been using it   for debugging purposes let's say if I'm running  into an issue with my code where I'm getting an   error I am trying to figure out what is the issue  typically I would go to my co-worker and have them   look at my code alternatively I can go to chatgpt  and save the embarrassment and basically ask the   AI if what is wrong with it fine bug in this  python code for linear regression so this is a   code that I had already written and if you look  at the code closely you can find the bug in the   last three lines this is a very simple test of  course in what I've noticed is that chat repeat   it does really well with simple questions but it  doesn't do a great job with complex but in this   case it's able to find that issue and it's able to  to give me the correct code second I've been using   it to prepare business communication a big part  of the data science job family is communication   and storytelling that includes writing documents  creating slides explaining to stakeholders asking   the right set of questions doing analysis and  then explaining your analysis in words in simple   English words and this is where I've been using  chargeability and it's been really really helpful   in that sense for example a lot of the times I  have to as a data scientist I have to work with   business stakeholders most of those people are  non-tech audience so a lot of the times I have   to translate that technical language into non-tech  language here I can spend time on my own to figure   that out but I can also use chat GPD to speed  up that time that is going to take me and ask   it to simplify it for me for example one of the  models that I use quite a bit is causal impact for   experimentation space specifically on experiments  that cannot be done with the control and a and   b design so here I use causal impact quite a bit  but a lot of the times I struggle with explaining   it to stakeholders so here I use chargpt to  simplify what is causal impact in a very simple   language for those who don't know causal impact  is basically a pre and post statistical model for   understanding time series data for understanding  the difference in Time series data in the pre and   post period and it's basically a model in r that  you can use so here let's say I'll ask chargpt   explain causal impact model in simple language  create a summary for plot one two and three that   spits out here it did a wonderful job explaining  what a causal impact does and I'm pretty sure if   I were to share this communication with my  business stakeholders they would be able to   understand it and not only that if you are new to  causal impact and you're trying to understand and   you need somebody to guide you you can actually  use chant repeat to kind of like look at the   simple explanation so simplify Concepts has been  a big use case for me with chargpt and you can   use it for learning new Concepts or you can use  it for simplifying existing Concepts that you   know or your communication with stakeholders  and business people of course as we get more   and more surrounded by AI we do not want to be  too dependent on AI that we lose the side of the   brain where we actually do The Logical thinking  because as AI progresses a side of our brain where   logical thinking happens is going to become more  and more important because shared GPT tool and   any similar models out there are going to be able  to do simple tasks very quickly and that's where   our cognitive thinking our logical thinking our  analytical thinking is going to become more and   more important one of the platforms that enables  me to think more logically is brilliant.org   who is also sponsoring this portion of the video  brilliant.org is an awesome tool for learning stem   subjects interactively in fun Hands-On way they  have several learning paths including data Science   Foundation which covers probability fundamentals  and builds up to using data in predictions   experiments and neural network the problems are  super interactive and structured in a way that   help me understand difficult Concepts in a logical  way it is a great way way to keep your mind sharp   especially in the world of AI as the information  Gap continues to close to try brilliant.org free   for 30 days visit brilliant.org Sunday's college  or click Link in the description the first 200   people to click will get a 20 off Brilliance  annual membership give it a try you might like   the interactive style of learning which is one of  my favorite ways to learn new things the Third Way   that I've been using chat GPT as a data scientist  is to help me analyze data let's say I'm working   on some new data set that I don't fully understand  I can spend a few hours doing exploratory data   analysis short for Eda or I can ask chargpt to do  it for me in less than five minutes for example   I have this data set that is from my personal  project that I've used in the past so I basically   it's a bright share data and I'm gonna copy paste  it and I'm going to put it into chat GPT and ask   it to explain the data for me so here as you can  see like it's doing a good job explaining it's   going into detail for each column the second thing  I asked it is to give me descriptive stats and   here it's generating that for me as well again  be smart about it you can rely on that GPT for   basic stuff but you still need to fully understand  what the data does what are the different column   means you still need to have the basic business  understanding behind the data set to be able   to actually use it it's super powerful tool for  summarizing your data quickly for example if you   want to do some descriptive analysis on your data  you can do that let's say if I want to find the   mean value of column act column Revenue I can do  that I can also input a data into charge GPD and   ask it to create visualization I can also ask it  to create tables for me from the data sets I give   it so there are so many use cases and I should  probably cover it in a separate video what are the   different ways to analyze data with charge gbt or  any of the similar tools including Bard the fourth   way I've been using cha GPT as a data scientist  is for interviewing and I'm going to tell you   like this is literally literally literally this is  my favorite part about working with Chad Deputy is   the interview prep there are so many use cases  that I can talk here but I'll share a few uh   right now and no I do not mean that use it during  the interview I mean use it to prepare for the   interview so it's like before the interview so  like don't that's cheating if you are using Ted   Deputy during your interview it's cheating so  don't do that I do not support that so some of   the use cases that I've used it for interviewing  purposes or interview prep purposes uh one is for   coding so let's say I'm practicing on lead code  or strata scratch and I have a free version there   are certain questions that Solutions are only  accessible through the paid version so here what   I do is I actually put the question in in jet GPT  and ask it to give me a solution so for example I   practice coding on this website specifically SQL  and python questions I have a free membership   and there are only a set of questions that are  available for free the other set of questions   are available you can view them but you can  not see the solution so here can't believe I'm   saying this but I actually used I didn't I should  have bought the membership but I didn't buy the   membership but I actually used those questions and  put it into jazz GPD and get asked it to give me   a solution and it worked out perfectly for me in  the future I will consider getting a membership   please don't come after me but this this was great  the second bit that I've been using it to simplify   Concepts so here I've used Chad GPT to basically  review the fundamentals and take notes example   here one of the one of the questions that I Was  preparing for an answer for in when I was doing   interview preps is in classification regression  there's the concept of precision versus recall   and this is one of actually one of the favorite  questions in machine learning interviews where the   interviewers like to us like how do you explain  precision versus recall or they will give you a   problem and then they will ask you to like what's  the Precision and recall here of course there's   like some business context behind it but like  the basic fundamentals of what precision and   recall is so here I actually use chat GPT to  do so many variations because I realized that   I I can mix that up in interview setting and I  wanted to get a very dumbed down version no shame   in it I actually did ask Chad Deputy to explain  precision and recall to a child and then later   I also found some YouTube videos that were like  super helpful in explaining but this is just to   give you an example that I have used it for like  super basic fundamental concepts in statistics and   machine learning in data analysis data science to  basically to review the foundations besides what   I've shared so far there's so many other ways  to use chat GPT in your day-to-day job at work   and outside work including writing documentation  writing emails summarizing meeting notes writing   animal performance reviews writing peer feedback  automating Excel sheets using VBA code I cover   most of these use cases in this video so watch  this next let me know in comments how you have   been using chat GPT have you tried anything from  the list that I mentioned or any other creative   hacks that you would like to share thanks for  watching I'll see you in the next video bye

Channel: Sundas Khalid

Views: 17,666

Keywords: data science, data scientist, self-taugh data scientist, tech jobs, big tech, Machine Learning, Python, data science projects, data science tutorials, data science jobs, AI, big data, data analyst, data analytics, business analyst, sundas khalid, women of color in tech, people of color in tech, inspiring stories, data science at google, data science 2023, chatgpt, bard, generative ai, data science roadmap 2023, data analysis with chatgpt, learn to code with chatgpt

Id: VmdJ3qgyBDE

Length: 10min 52sec (652 seconds)

Published: Wed Feb 22 2023