No ratings yet.

A couple of months ago I started as a working student at a consultancy firm. I was asked to create a dashboard for our internal recruitment department. Since I had zero knowledge of PowerBI, I was given some time and options to learn by following a couple of online trainings. What I learned at a later point was that it is nearly impossible to completely learn how to make a PowerBI dashboard by some online training, since real-world data is never as flawless as the training data is. So a tip in advance: do not bet on a simple training before creating your first dashboard. 

After this dashboard, I was assigned to support a client where they were making a more complicated dashboard which covered financial data. The first internal dashboard was quite simple as it covered only 4-5 tables with less than a hundred rows and columns, but the financial dashboards covered tens of thousands of rows and a lot of columns and tables. With initially having zero knowledge of Power BI, I would now consider my skills as medium and this is partly thanks to ChatGPT. In my opinion, the tasks in which ChatGPT can support you can be divided into three categories. 

1. Finding things in Power BI

Power BI has a lot of features and options and as a newcomer it can be very hard to find the correct ones. By entering sufficient prompts into ChatGPT, you should be able to find any feature/option you want. In my opinion, the main difference with googling it, is that you do not need to visit several websites to finally find your answer. In the first screenshot, you can find an example. 

2. Finding and/or explaining DAX formulas

DAX is the ‘code language’ (it isn’t really programming code or something) which you use to create variables (which are necessary for building visualizations). As Power BI is a program by Microsoft, DAX has a lot of similarities to how you write functions in Excel. Although they are similar, they do – for some reason – have a lot of small differences. ChatGPT can be helpful to either find the relevant DAX formula you need or explain the differences between them. As seen in the screenshot, I asked ChatGPT to explain the difference between the 5 different DAX ‘count’ formulas. 

3. Formulating DAX formulas/input

Firstly, I had quite a lot of trouble by formulating the correct DAX formulas/input and my variables did not work. During the process I found out how to use ChatGPT appropriately. Since ChatGPT obviously is not able to write a 100% perfect DAX, I got used to entering abstract inputs and consequently adjusting them to my specific situation. I think this was the most useful application of leveraging ChatGPT when creating PowerBI dashboards. The last screenshot shows how it instructs me to properly use count- and filter- context when creating a variable.

No ratings yet.

Attending lectures is highly encouraged by professors, since it will give you a more direct type of education and will provide the chances to ask questions and have interactions. While I personally believe in the value of attending lectures, I do think that the knowledge gained during a lecture can be diverse and is dependent on a couple of things.  

One of those things is the ability to focus on the professor speaking while trying to make notes. Personally, I try to make notes during a lecture which capture the important things which are presented in the lecture slides in combination with what the professors say about them. This can be difficult since not everything on the slides is relevant and not all contexts given by a professor are necessarily relevant. On top of that comes the speed with which you need to make notes of the slides and notes of what the professor said. Sometimes, these two can go very fast and therefore I went searching for an AI tool to help with this.  

I wanted to find a tool which picks up the relevant information on every slide and notes it. This would decrease my time on typing the important stuff of the slides or copying certain images and increase my listening time. Although I went looking for such a tool, I have not yet been able to find one.  

I tried multiple tools like slidespeak.co, pdf.ai, and Google’s new Bard. Here is a short summary so you do not need to try them (at this moment in time) 

Slidespeak pitches that you can summarize PowerPoints or pdf and can ask questions to it but works very poorly. See the screenshot as an example. When asking about summaries or simple answers it does not provide answers.

 

I also tried pdf.ai but those similar problems.  What I also noticed is that the tools had problems with converting image-to-text. As almost all lectures include slides with a combination of both, this is a problem for my idea.  

Therefore, I tried Bard, which can convert image-to-text (Sha, 2023). Although it was not able to provide me with 100% reliable notes, it did a sufficiently better job than the other tools. I tried to find out what the boundary for complexity was in context of the combination with text and image and concluded that it was somewhere between the next two slides. It was able to correctly summarize the first one but did not capture the time effects on the second slide. 

Unfortunately, I must conclude that a tool which will summarize lecture notes reliably is not yet out there.  

References 

Sha, A., (2023). You can now upload images to Google Bard; here are some cool examples. Beebom. https://beebom.com/google-bard-upload-images-cool-examples/