Python Pandas now has AI (GPT-4). It's good and could take your job

Thanks to Brilliant for sponsoring this video :-)

Pandas AI combines Pandas, the python data analysis package by Wes McKinney, with generative AI. It creates something that enables data analysis with very little code. Should we be worried?

and here

Learn Data Science (affiliate)

Learn Python with Giles

📚 My favourite python books for beginners (affiliate links)

🆓 Free Python Resource
(This is a great introduction to python)

⚙ My Gear

📱 Social Media

👌 SUBSCRIBE to ME!👌

#AI #python #dataanalysis

[zamplify]

I feel like this video is more about the editing than the subject.

[YugalJindle]

Chanel should be named “learn if you can while I interrupt”

[bc]

Here's the thing: do people hire you for analysis because they don't know how, or because they don't have time / desire to do it themselves? Most people don't know how: they don't think analytically in general, they don't know what to analyze, they don't know how to do it technically...

None of those things change just by you knowing being able to get it done faster with AI. You would need a complete societal transformation and radical education reform in order to get people to not need you to think analytically for them.

Beyond that, it's just like debating between Python and R, or any other language / tool stack: it will still be the analysts using the tools.

Another way I've put it is that for the foreseeable future, you still need humans to care what the results are. And if things change so much that humans are totally worked out of data-driven management entirely, you're really not going to care about the job anymore.

[mahendragl]

AI could take your job...
Casually askes everyone to learn math and stat from the sponser....😂

[AussieBIMGuru]

Honestly need to see how a nonprogrammer uses these tools to judge it properly. Currently programming experts are showing us the value of these tools, but the proof will be in seeing if low/no level programmers actually want to engage with these. E.g. most of them were lost at pip.

[springsm]

I wonder if something like pandasai could be applied to Wireshark/Nmap, and looking for any irregularities in network traffic.

[GwriPennar]

You could just use GPT-4 with Data Analysis to avoid the hefty API calls via Pandas AI and reduce costs. Then keep your code to be independent of OpenAI even though its availability is quite high but not 99 percent. Just a thought. But can see the power of using the API within a notebook.

[murphygreen]

Can AI keep you still while talking and make your screen with the code in focus?

[jhonshephard]

What this kind of indicates is, you need a lot more theory than practical code now. I have experienced this recently where I helped someone who was great at writing the code but didn't have as much theoretical knowledge. The LLM would not know what to investigate or what to do in which order for example when your model has much higher than expected accuracy, which after debugging can turn out to be your data is not labeled correctly. The LLM can't debug that but we did.

[tshwarelomatinketja]

Brilliant this video's sponsor ❌️
This channels sponsor ✅️

[davidk]

I believe this man is in the early stages of Parkinson's disease.

[purplepotato]

I've just made up my mind to try and learn Python now, and I'm sort of worried? I'm concerned that by the time I've got the basics down, there will be too many data analysts, and/or AI would be capable of running data for those who don't know coding, which would make the entire endeavor moot.

[myleslloyd]

2013-2022 is the last 10 whole years of data (if it includes 2013 data). The borough only showing Kens&Chels is, of course, an error.

[EdwardTilley]

You asked for just the top of the iceberg so I wouldn't worry too much, these tools end up taking longer than they save you today.

[estevaoyt]

Great content, I love your videos! So much fun in the: "Im youtuber, money is no a problem to me" 😂 Keep it up!🎉

[juleswombat]

So many of the python libraries have their own specific syntax for doing the same or similar things. The numpy syntax a bit different from pandas, which is different to pytorch etc, so we end up with a large confusing syntax to perform data analysis. LLM tools help as co pilots, but we end up losing our syntax memory. I am surprised it has taken so long for an AI/ML tool to arrive to inform us on the best model, technique to use for data modelling and AI, and encode it for us. I am glad I do not have to make a living from having experience in the use of these low level libraries.

[nicholasrobins]

Work data without code? most analysts I know don't know how to work with data. I had to explain what a median was the other day. Keep in mind this was over email, so they could have Googled it first! but yes this is a cool step forward.

[tiolavara]

So this is pretty much almost the same as desbribing your problem and dataset to chatGPT itself, but you do not learn coding in the process, would not say it is useless, but mehhhh. Also I am surprised it failed with the 10 years

[michaelnurse]

SciKitLearn? It is not 2011 anymore. Cmmn guys.

[sumithraj]

First person to comment, don’t take human jobs AI.