AI Meets Pandas: A First Look at PandasAI
Vložit
- čas přidán 25. 07. 2024
- 👉 Explore All My Excel Solutions: pythonandvba.com/solutions
𝗗𝗘𝗦𝗖𝗥𝗜𝗣𝗧𝗜𝗢𝗡
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
In this video, I explore PandasAI, a new Python library that works in tandem with Pandas to provide a more interactive data analysis experience. Follow along as I dive into a hands-on example, using a dataset to demonstrate how PandasAI answers direct queries about your dataframe.
Apart from demonstrating its usage, I also share my personal views on this new tool, discussing its pros, cons, and aspects like privacy, security, and costs. This video is an initial exploration of PandasAI - giving you a taste of what it can do, and how it can enhance your data analysis workflow with Python and Pandas.
Tune in, whether you're new to Python or an experienced data scientist, to gain a fresh perspective on data analysis with PandasAI.
🌍 𝗟𝗜𝗡𝗞𝗦:
▶ Jupyter Notebook from the video: github.com/Sven-Bo/PandasAI-F...
▶ PandasAI GitHub Repo: github.com/gventuri/pandas-ai
⭐ 𝗧𝗜𝗠𝗘𝗦𝗧𝗔𝗠𝗣𝗦:
0:00 - Intro
0:20 - Setting Up the Dataset
0:31 - Analyzing Data with Pandas
1:39 - Analyzing Data with PandasAI
3:20 - Sharing My Opinion on PandasAI
3:46 - Discussing Privacy, Security, and Costs
4:30 - Outro
𝗧𝗢𝗢𝗟𝗦 𝗔𝗡𝗗 𝗥𝗘𝗦𝗢𝗨𝗥𝗖𝗘𝗦
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
🆓【𝗙𝗥𝗘𝗘】Excel Add-in (𝗠𝘆𝗧𝗼𝗼𝗹𝗕𝗲𝗹𝘁): pythonandvba.com/mytoolbelt
📊 Dashboard Excel Add-In (𝗚𝗿𝗮𝗳𝗹𝘆): pythonandvba.com/grafly
🎨 Cartoon Charts Excel Add-In (𝗖𝘂𝘁𝗲𝗣𝗹𝗼𝘁𝘀): pythonandvba.com/cuteplots
🤪 Fun Emoji Excel Add-In (𝗘𝗺𝗼𝗷𝗶𝗳𝘆): pythonandvba.com/emojify
📑 Excel Templates: pythonandvba.com/go/excel-tem...
🎓 My Courses: pythonandvba.com/go/courses
📚 Books, Tools, and More: pythonandvba.com/resources
𝗖𝗢𝗡𝗡𝗘𝗖𝗧 𝗪𝗜𝗧𝗛 𝗠𝗘
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
🔗 LinkedIn: / sven-bosau
📸 Instagram: / codingisfun_official
💻 GitHub: github.com/Sven-Bo
💬 Discord: pythonandvba.com/discord
📬 Contact: pythonandvba.com/contact
☕ 𝗕𝘂𝘆 𝗺𝗲 𝗮 𝗰𝗼𝗳𝗳𝗲𝗲?
If you want to support this channel, you can buy me a coffee here:
▶ pythonandvba.com/coffee-donation
*What do you think of PandasAI? Let me know in the comments below!* ⤵
What ever you teach it’s very clear an understanding, keep it up . ❤
Nice application. It seems a paid subscription to OpenAI is needed to use the API?
@@JessChristensen Yes, that's correct
Great intro and thank you. Having worked with mission-critical corporate and life-critical healthcare data, someone has to have ownership and responsibility of analytical outputs. I can see how powerful this could be for getting some rapid analytical directions, but (at least for now), the absence of accountability and the high likelihood of errors is a show-stopper for critical work. I'm looking forward to seeing how it can function as a kind of junior assistant analyst. Cheers - Dave
Glad you liked it. Thanks for watching and taking the time to leave a comment!
This is awesome! We're pretty close to not having to code at all for data analysis
Makes me happy for those that can't code... but sad for someone like me that likes to code
Hey Luke, thanks for tuning in! Right now, I do enjoy using AI as a coding buddy, asking for snippets and insights as I go. You've hit the nail on the head, though. I can imagine that in the near future, coding might no longer be necessary in data analysis. We could see more powerful, AI-driven no-code tools doing the job. Let's see..
Thank you, I am already excited to go start using the PandasAI once on my desk😊😊
Glad you liked it. Thanks for watching and happy "coding" with PandasAI! :)
Thank you for this video !
This is an amazing library, however i was wondering if its possible to plot the "charts" it render in a streamlit app ?
Thanks for watching. I think that might be tricky, but I have not tried it out yet.
Noob question :) I am a Blender user for (old school) architecture, with PandasAI can this be used in the Blender's Python console to read an external database, read to a written instruction, and respond by creating a 3D model? I'm thinking the external database would include a schedule of rooms that would include dimensions and areas, relationships to other rooms, equipment list and so on. Does this make sense? Cheers Nigel
Hey Nigel, thanks for checking out the video! Right now, PandasAI is mainly used for data analysis tasks and works closely with Pandas DataFrames. So, it can't take instructions from an external database and convert them into a 3D model (at least not yet).
not sure what the use case for this is when you can just paste in the git hub link into the prompt bar and Chat will analyze and help with errors or write documentation or whatever.
Thanks for watching and throwing in your two cents.
Hi! How do I give this "memory"? I want to use it in a sample chatbot. It should be aware of the context of the conversation as it goes on, then only it will be able to understand user's queries
Thanks for watching. I don't think it's possible, but it's better to check the AI's documentation or ask your question in the GitHub repo: github.com/gventuri/pandas-ai
@@CodingIsFun yup makes sense!
Hi, great video. Thank you very much. Would it also work with Bard ?
Thanks for watching. Currently, Bard is not supported in PandasAI.
"But for now, I'd rather write my own pandas code because it lets me know how the data is processed.", I totally agree.
Thanks for watching! 👍
Great Video! Thanks. Careful with your API key!:)
Thanks for watching and the hint. I deleted already the API key before uploading the video 😉
hey Although I am using Paid API key, the API is taking a long time to respond. is there any chance, we can improve the Performance?
Thanks for watching. You might want to raise an issue in the respective PandasAI GitHub repo. Thanks!
With any dataframe i am running on this its giving me following error.. Even can't find solution on bing or chat gpt
"data type 'int64' not understood
" unfortunately i was not able to answer your question, because of the following error:
data type 'int64' not understood
"
Thanks for watching. Feel free to raise an GitHub issue here: github.com/gventuri/pandas-ai
I can't find solution to this 😢.. Tired of it
Do you think this is a good option for beginner python data analyst like me? Or rather I should focus on the mechanics of understanding the code first?
Nope, I think this would be the oposite. If you do not want to become a data analyst and you just want to have some easy questions solved, it could be a good idea. But data analysis is much more than this.
Honestly, I'm not entirely certain if there'll be a ton of coding required in data analysis down the line. I recently saw a demo of ChatGPT's code interpreter, and it practically wrote all the necessary Python code on its own. You simply ask a question about the uploaded data and ChatGPT handles the rest. That said, I still think that knowing how to code can be useful in a lot of areas, not just in data analysis. So, if coding is your thing, I'd totally recommend sticking with it.
@@CodingIsFun thats something that already was available in a certain grade with the possibility of employing queries with Pandas. I saw chatgpt and can answer simple questions, but not the complex ones, as well as you lose the control about what it is happening with the data, and the impossibility of verificating the result its giving you. They show it under controled situations, where you ask about small sets where the data is known.
@@CodingIsFun, thanks! Do you have recommendations for 2023 build up skills as a data analyst? My job is sort of dealing with that, but a lot of other things as well. I knew Excel, SPSS, and currently visualize with Looker studio and I started the SQL/Python libraries path. Not sure if this is the best way to go, but there is quite a lot of interesting stuff.
Appreciate the channel, btw, easily one of the most down to earth explanations on youtube.
I like the picture 🐼. And it's a short leap from here, but at this point it still needs a human to know what to ask, and to care what the results are. Plus, most places have little to zero technical analysts / tools, so if anything this would be another tool for their analysts - or have no effect at all... 🤷. What this could represent though, is the evolution of the AI's ability to give accurate info, versus "hallucinations" - even if used just to have it show how the answers were determined.
Thanks for watching! You've got a point. Even with tools like this, we still need people to ask the right questions.
Does it work well with large datasets?
Thanks for watching. I have not tried it out on a large dataset.
Is it possible to retrieve the intermediate_steps ?
Thanks for watching. That is currently not possible.
I would not use, because of the same reason: I want to control what I do, how I do it and everything involved in the process. Also, once you know the syntaxis, the classic Pandas in Python is faster, and easy to change. I want to control what I save in the variables, and to be sure about the results. If you automatize some lines, you already do not control how it reached the result, and you even do not know or understan the result. If I employ the .nlargest() function, I know exactly what Pandas will do. If I just ask the AI about the top 5 countries by GDP, I do not know what it was doing, if the result is plausible, and what the AI understoo as the top5. The main problem with the OpenAI tools for data analysis is that you can never be sure if the result is what you meant.
Other example: there is already time ago the posibility of using queries in Pandas, and as far as I know, it is still a not so popular option. And I think this is because of 2 reasons, the first one is that syntaxis does not improve in efectiveness or speed, and the second one is that you keep a less strict control over the operations.
Thanks for watching and sharing your opinion! 👍
excelente información, thank you, unfortunately here in Venezuela we are blocked to use openAI, but anyway I prefer the normal pandas
Thanks for watching. Glad you liked it. Oh, I didn't know that OpenAI isn't available in Venezuela 😕
yes, we are blocked, and not only openAI, but also Java, Mysql, and several more
I think I'm first. 🎉 Always a good day when Sven uploads.
*First comment!* 🎉🚀
I think it's a matter of trust. I"m not sure that we are on this level yet, that serious business decisions could be made after prompting the chat gpt.
Thanks for watching. Trust in AI is indeed a vital concern, especially in business contexts. Let's see where this AI journey takes us.
How do you get Job without any Degree? Can you please guide us 😊
Thanks for tuning in. Since I don't have firsthand experience, I can't offer a lot of advice here. But you should totally check out this video from Luke: czcams.com/video/vRbgun8ryVE/video.html. The stories from his guests are super inspiring. Good luck on your data journey! 🍀
Thanks for this one but I guess I will prefer your explanation on the Pandas library instead of AI 😊
That's great to hear, Ricardo. You know, sometimes I feel like I might not be needed anymore, given how quickly and clearly ChatGPT can generate and explain code. 😅
@@CodingIsFun I think aside from your teaching you're also offering journalism because each of these vids are suggestions for what tools to use to solve problems.
Ohh my god.. this is the future of coding.. no programmers are needed.
Thank you for watching the video and taking the time to leave a comment!
Coders are so smart that they even coded the coders out of coding
😅
An interesting tool, but the results need double-checking.
Totally agree
It will make coding obselte
That might be totally right, especially in data analysis. 😮
@@CodingIsFun so what should we do??