The Pro version of ChatGPT, called ChatGPT Plus (subscription at €20/month), has introduced a new beta feature that was highly anticipated (by me, in particular), called Advanced Data Analysis.
How to Activate Advanced Data Analysis?
Once logged into ChatGPT, click on your email address at the bottom left, go to "Beta Features": here you can enable "Browse With Bing" (ChatGPT will then be able to go online to retrieve up-to-date information), Plugins, and "Advanced Data Analysis", which we're discussing today.
Then, when you open a new chat, click on "GPT-4" and replace "Default" with "Advanced Data Analysis".
The most obvious visual difference is a small "+" that appears on the left, in your text area.
What is Advanced Data Analysis for?
It's the evolution of Code Interpreter, on steroids: instead of just reading code, we're moving towards a function capable of analyzing any type of data.
So, yes, code, but also images or other files. Once the function is enabled, for example, you can give an image to ChatGPT and ask related questions.
Another novelty: when the model feels it lacks information to understand what's going on, instead of making a very personal (and often wrong) interpretation, it will ask you questions, like "To make sure I understand correctly, let's clarify a few points."
UX
For instance, I uploaded a screenshot of a future tool we're currently working on, explaining my UX (User eXperience) concern. After a series of questions, ChatGPT managed to give me 2 very serious and relevant improvement suggestions out of the 3 it proposed.
Graph Generation
This is the function that, in my opinion, is the most powerful. I created a CSV file that summarized "fictional" sales of several products, which I uploaded. I asked the AI to make a sales graph for me, in "cumulative sum" mode, by week, with the best week in green and the worst week in red.
And magic happened. ChatGPT laid out a plan in 4 steps:
- Convert the data into a usable structure (in this case, a DataFrame).
- Process the data to get the cumulative sum by week.
- Identify the best and worst week.
- Create the linear graph with the specified colors.
And that's exactly what it did!
For each step, I can see the generated Python code (yummy), and we end up with a beautiful JPG that presents my graph.
And Everything Else...
It's honestly quite astonishing: I've already seen people photograph a crossword puzzle grid and give it to ChatGPT, which solved it in a few seconds. OK, technically, everything is feasible in Python, with squared libraries (I don't need to understand all the ins and outs to code that), but here, it's done in a few seconds without having any coding knowledge. Hats off!
Starting with a URL you provide, our AI analyzes the page and generates 50 relevant keywords, followed by 20 ideas per keyword - a total of 1000 content ideas. For each idea, we provide you with 3 development axes to assist you in starting your article writing.