ChatGPT Code Interpreter - Goodbye Data Analysts?

Key Takeaways

all right guys another week another new big thing in AI like can you give us a break like how are we supposed to keep up with all of this and well this video will be an attempt to get you up to speed on chat GPT code interpreter so it is currently in the alpha phase and it is a chat GPT plugin so I currently don't have access to this plugin yet unfortunately because usually I like my videos to be Hands-On with lots of demos that you can immediately try out yourself but unfortunately I'm on the wait list if you want to get on the waitlist as well you can go to this URL over here and then join the plugin wait list because there are already some plugins that you can leverage that you can use within chatgpt so here you can see some supported but if I scroll down now the next big thing that everyone's talking about that's popping up everywhere on social media on Twitter is code interpreter and I think for a good read reason because this is really interesting of course and especially from a data analytics and data science point of view so that is why I want to get into this in this video and while I don't have access yet what I want to do is I have some examples of other people that are using this already and I want to go over those examples and really explain like my thoughts my take on this and how I think it will affect data analysts work and data science work because things are going to change this is really a big thing and I feel like this really is the next phase the next level already of like interacting with large language models like open ai's chat GPT for example and why is that so these these large language models they kind of understand language in the sense that we can communicate with these models interact with these models using language we can type something ask a question and we will get back an answer in the form of text but now with code interpreter the model can use python and handle uploads and downloads so this means it becomes multi-model in the sense that we can work with different kind of file formats different kinds of information so what does this mean well it means that we are now not only limited to human language language that we understand but we can also leverage computer language in the form of code and digital files and that is how we operate that is how we interact with computers basically in the background it's all code that's performing some sort of operations on files resulting in output actions Etc so if you really think about like smart agents AGI applications like that need access to the digital realm basically they have to understand code digital files and code interpreter is like an attempt to make that work and now while we've already seen other attempts like for example baby AGI and auto GPT which I also have a video on the examples that I'm seeing now already from code interpreter are like so much better than anything else we've seen before and basically like in their own words open AI says like we would like our models to be able to use programming skills to provide a much more natural interface to most fundamental capabilities of our computers so I feel like this really is is the key like I've said it said that integration between understanding human language and on the other hand also understanding computer language so we the humans provide the agent or the AI with instructions information in the form of language but then the AI the computer also understands how to execute that on a computer to actually make it work and they say like having access to a very eager Junior programmer working at the speed of your fingertips can make completely new workflows effortless and efficient as well as openly benefits of programming to new audience and this is also something very interesting that I want to get into programming for new audiences because you don't have to be a coder anymore right now already but also especially in the future to to all kinds of crazy things with these models so what is it good at right now they say like initial user Studies have identified that code interpreter is good at solving mathematical problems both quantitative and qualitative and doing data analysis and visualization and this right here large part of your job as a data analyst and also like data scientist is doing data analysis and visualization and solving mathematical problems so what does this mean for data analysts and data scientist that is what I want to get into by running you through some examples and then really saying like Okay how is this going to affect the work we do so now let's get into some examples and I found this really good Twitter threat over here by Hassan Thor link will be in the description so you can check it out look at these examples on your own but he basically compiles a whole bunch of tweets where people are showing of how they are using cha GPT code interpreter right now so the first example over here like what can I do so basic video editing in chat gbt converting a uploaded gift to a longer MP4 version with a slow Zoom so you can see we have the original video which is a gif and then it turns into a video with a zoom so what is it doing over here it's converting a file and it's also making an edit and manipulation on that video file so what what does this mean for like the future like okay like you will have fully automated video editing software that can basically for example on given your style of content creation and and your types of edits and cuts can probably very quickly spin up a rough draft of an edit for you it's basically like interpreting the script basically first so understanding what people are saying in the video so then it's probably a transcription and then it understands okay this should be a good story and then it can cut it can create zoom in Zoom outs transitions like in a couple of years we will have fully automated video editing software that is basically what I think of when I see this okay next thing so we have segment music markets based on spreadsheet and come up with business strategies for each segment so there is a video over here basically a demo of a upload so it uploads a spreadsheet and then it says great you've uploaded the spreadsheet and then let's start by taking a look at the content so this is really how you inter who how you work with The Interpreter you upload a file and then here you can see it just reads the data and then you can ask questions about it it understands it and yeah it's so interesting from a data analysis point of view working like this and then giving your giving the model full access understanding to to your data and then that is of course like the next big thing that's that's going to come up like okay you upload your data so now it has access to all your data like can we do that with with company data and like right now with this Alpha phase and how open AI is set up probably not a good idea to share sensitive business information but I know like for example Microsoft is already working on isolated versions isolated clones of the GPT models so you can interact with them with sensitive information without sharing it with with the world so just keep that in mind as I go through these examples all right what's next analyze and summarize your music tastes from your Spotify playlist so here it's analyzing 300 300 hour Spotify favorites playlist with the code interpreter and what you can see right here is that it's coming up with all kinds of graphs over here so it's doing a principal component analysis it seems like and we have some some nice visualizations over here more plots and then here also the interpretation so the explained variance ratio and kumatliff explain fairness ratio from the PCA So This is highly technical stuff over here like understanding principle component analysis is pretty an advanced topic and jgbt code interpreter can help you with that alright what else all kinds of visualization charts and graphs so what are we looking at over here so this is then in the tweet from Ethan Malik and again like credits to all the authors of these tweets and if you want to stay up to date learn more I would suggest like following around checking out these these profiles because they already have access and can tell you probably a lot more but here you can see like these are like basic Seabourn mud plot lip kind of plots that you can create using Python and that data analysts and also data scientists do as large part of their work so not even sure like what this is but they're pretty fancy graphs all right and then here we have extracting colors from an image so what's going on over here so you upload an image and then so we can do the image and then you can say like extract the most what it says like top colors from the picture and make a color palette so this can help you with like Design coming up with like a design template the color palette for your your brand or your presentation or whatever and boom spits out the code over here too to help you do that and then it also executes the code so here you can see like actually how it's working so first it comes up with the code and then return so the palette.png and then here boom We Run the code now let's see like in the end let's let's have a look at the output file can you display the pellet PNG now boom there it is color palette so this is an example from Pietro's shirano and what else get country specific content breakdown by uploading a Netflix data set so this is a data set from kaggle it appears to be and then we get a report so we have a data set from kaggle pretty seems to be pretty large data set 12 columns and then we have a breakdown top 10 countries full report full breakdown and then here we can see how it's generating a gif like out of nothing basically the prompt is make a gif this size with falling green Matrix of letters assume no fronts 30 frames 5 FPS so video generation gift generation here we have a look at Bitcoin data analysis all kinds of charge to look at like the trend the seasonal component the residual component to do time series analysis and then what we have here turn your photos into a text file so we upload a PDF over here and it says like OCR and image for me and generate the text file so oh it's the PNG so not PDF but it's like an invoice it's processing then it does OCR optical character recognition to read everything from that image and now we have an output txt what does that look like boom there it is image to text alright so what does this mean for data analysts and data scientist and why the hell am I trending in the Netherlands like what okay I have no idea what this is sorry little Sidetrack but what does this mean like I think really like if if we look at all of this like our the work that we do will definitely change and looking at this and what it already can do to achieve the code interpreter I think really like data analytics jobs data analysis it's going to impact it's going to affect that like for sure and I think already pretty soon and I how I think this will play out is since for example like Microsoft is already like big friends with with open Ai and they announced that they will introduce co-pilot into the Office 365 stack so this will be in Excel so right now I like from working with tons of companies I know that like still the main way for most companies to do data analysis is just through Excel they have big spreadsheets company data financial data whatever data and they use that to do their analysis for them and they can do basic analysis with that but then when it gets too complicated they need help from either a data analyst or if they want to do some more predictive kind of stuff machine learning they need help from a data scientist that is typically how it works right now but if you interpret this kind of functionality into Excel you now enable end users basically who might not have the technical understanding of how to do the data analysis to come up with all kinds of insights and coming back to how GPT or oh sorry open AI was describing it like it offers a completely new workflow more effortless and efficient as well as open the benefits of programming to new audiences so this is really the thing here so it allows non-technical people that don't have a data analysis or programming background to do analysis for them and like we've seen already you can get quite complex with this now and that is there's the butt a huge a large part of data analysis of course is the interpretation making it specific to The Domain that you're working in the company that you're working in so of course you can't like replace a data analyst or a data scientist like with all at once like with a tool like this because the interpretation the reasoning behind it is still a large part of this and that in turn is also like right now my biggest S5 is for data analysts and even also for data scientists because like they will be next like probably next week you'll have chat GPT machine learning or whatever and it can do that as well so like I've said like domain knowledge understanding a field really well understanding a business really well like it is already really important but now there are also like plenty of jobs that just require like technical skills just like being able to do simple analysis simple work processing data but like like we're seeing right now like basic analysis basic Transformations the AI can already do this so it's now much more about okay we have this tool that can help us to do the work but now it's really your expertise as a professional to make sense of the data ask the right questions and make it so that it will come up with useful insights for the problem that you are helping your company with or helping your client with and also in general just like being a great team player and being able to communicate your findings effectively to the company to your team and again already like a skill that is already really important and can set you up for a good career but I think what we're going to see is it's kind of like a shift a transition where given these tools you can accomplish basically the same output more output with less people that's just a fact like this will supercharge your productivity so basically what will happen like teams companies can probably scale down in all kinds of areas like basically all the kind and this is not just like data analytics data science but like all the jobs basically that require a human to sit behind a computer and process some kind of information it will be speed up by AI there will be AIS there will be agents that can do that much more effectively much more efficiently so what that what will that mean either companies will like stay with like the same kind of like team and employees and just like 10x or maybe even 100x their output but like for a lot of like teams and roles that that might not even be necessary so what I think is is far more likely is that teams will scale down because whereas you first needed a team of like I don't know 10 people you can now do it with with one or two and what will naturally happen of course the ones like in the team like the a players the past players like they will keep their position and they will be super charged by these tools supercharged by these AIS so how do you set yourself up so you will not be replaced by these these AIS and in my opinion it really is like threefold so is one is really understand how to use these tools effectively and start experimenting with them like right now because otherwise wise you will be left behind because there are others within your field that are using these tools and they will 10x 100x your output so that's one understand this work with these kind of tools stay up to date and two is domain knowledge so try to fill in the gaps that these models cannot do yet and like I've said it's more of the the execution the the implementation so the transformation making the graphs but it's less about making it specific to your domain the interpretation so that is another like key area where where I should focus on if you want to like stay relevant and then third is overall like be a good team player and focus on communicating your results and finding effectively basically because humans just naturally want to work with like nice people people that communicate effectively are nice to deal with that that interpersonal contact these conversations those are the kind of things that AIS cannot replace and probably will never replace like human interaction and that is a large part of work like the socializing be helping each other out learning from each other that is still like a big part it will become less about the technical skills for like these kind of jobs whereas like right now you maybe need a data analyst or a data scientist who's just like really good at like one specific technical kind of thing but he's kind of like a jerk and I don't really like working with him but he's really good like like that guy will be out because the AI cannot can cover all the technical stuff and if there's another guy who's maybe a bit a little less technical but can still implement the technical things due to AI but he's really nice to work with and just overall helps you you can learn from him like he will get the job so like that that is my those are my thoughts on it so like I said learn to work with these tools gather a domain knowledge specific knowledge that is not captured right now like in these models not any capabilities and just overall be a good team player I think that will set you up for a future where you stay relevant in this new era of of AI basically but like also again these are all just like guesses my thoughts nobody knows how this is going to play out or of course like we've seen like in the past like jobs get replaced all the time but there also will be new jobs so overall I think if you are a data analyst or data scientist like being in this field like having the technical skills you you shouldn't worry at all like you you we are in a great spot right now like working with data is still amazing there's so much data there are still like companies that don't use data at all but like I've said probably the work will change maybe it will be more spread out so not so much focused on like large organizations huge teams but more so like like small little teams where every company eventually has some kind of like a AI engineer who manages the agents and works sets up the agent so all you get all kinds of new jobs and opportunities probably but it will likely shift like the days of just like plotting some graphs and then writing some python functions with matplotlib to like impress your boss or your clients like they will be over because they will be like hey yeah really cool that you can do that but I have like Excel over I can press like one button and I have that also with like an interpretation specific to my business so hey those are just my thoughts exciting stuff it's going really fast hard to keep up but I hope at least that this little ramble of mine has at least brought you up to speed on chechi BT code interpreter like I've said you can get on the wait list to try and get early access to this to already start experimenting with this see what it can do for you how it can help you with with your job and yeah I'm like really curious what next week will bring us in the world of AI the next big thing what will it be will it be Che GPT machine learning I don't know it's going really fast it's exciting but also a little overwhelming I would say hey that's it for now if you found this video helpful like please leave a like it really helps me out and also you let YouTube know that you want to see more content like this like to stay up to date on everything like Ai and data and also subscribe to the channel to make sure you don't miss any future videos that's it for now thank you for watching and and I'll see you in the next one [Music]