Make learning more engaging, applicable and accessible with AI
Generative AI is changing learning by making it more accessible and personalised. Tools like FoondaMate, a chatbot on WhatsApp and Android, help students with homework and exams, offering 24/7 access to educational materials with an AI.
For data analysts, the demand for more skills continues to grow. However, balancing a full-time job, family, and whatever else life brings can make finding time to learn a challenge. This led me to explore what I could learn during my commute to the office.
Studying on the go presented several issues:
These challenges often were due to the format of training materials, which were mostly long text documents or videos. Yet, during my commute, I can easily engage in short text conversations and listen to audio content like podcasts.
This sparked the idea: can AI convert documents into text conversations, or even audio formats, like podcasts, to achieve more on-the-go learning?
Document to Dialogue
Advancements in the ChatGPT mobile app make this more accessible than ever. Rather than quizzing ChatGPT on a topic, you can upload a document to the app, which ensures you are using accurate content. Then, you can ask the AI to summarise the information or answer specific questions about it.
"Hey ChatGPT please can you review this document for some questions I have, please respond with concise answers, when you are ready reply 'ok' [Insert Document]"
For paid users of ChatGPT, enabling voice mode allows for conversations with the AI, receiving responses as audio. Free users can use the microphone to convert words into prompts, but responses will be as text. All conversations are stored as text in your chat history and can still be prompted with text inputs as usual.
I've found this helpful for breaking into a new topic. For example with dbt, early on I was struggling to remember what different files did. However, being able to upload the documents and start asking questions made the process much less time-consuming.
It helps to begin with what you already know to get more personalised answers. For example, when learning about Tableau Level of Detail (LOD) calculations, I asked:
"I know how to:
Based on this document, how can Level of Detail calculations help me improve my work?"
This approach provides a more personalised response that is highly tailored to help you learn quickly. Using voice commands this could be a typical conversation you have at home or while walking, which makes it highly accessible compared to more traditional learning methods.
Dialogue to Speech
Personal projects have been essential for me in cementing new skills or techniques. For example, when learning different Tableau functionalities, I built a dashboard for Tableau Public and revisited it to relearn how I did it.
But could the same be true for these conversations with AI? Could they be transformed into podcasts for others to learn from?
领英推荐
Podcasts are just conversations between hosts and guests, on a particular topic, with an element of entertainment. With a few modifications, you can make your discussion podcast-ready and much more engaging than turning the documentation as it is into audio.
For ChatGPT I went through the process of:
In the Tableau Level of Detail (LOD) Calculations episode the goal is to spark ideas for day-to-day use of LODs with Tableau. To achieve this the two speakers would walk through how they work and the differences between FIXED, INCLUDE and EXCLUDE, but crucially provide examples of where they can be used and what they build.
Building the script does present new challenges:
It can be difficult but does reinforce your knowledge and build on what you learnt at the start of this venture.
With the scripts ready, the next step was to convert them into speech. I chose ElevenLabs for this task due to its impressive text-to-speech quality, with a wide range of voices and styles. Ranging from dramatic crime documentary hosts, regional accents, and thankfully conversational speakers.
The free tier of ElevenLabs allows 10 minutes of audio in a month. So I could generate dialogue for my speakers in different voices. With the audio creation, it's not 100% consistent in pronunciation, so LOD was sometimes said as lod or L.O.D. Converting the words to their pronounced forms helps make this more reliable.
Once the audio clips were ready, I used DaVinci Resolve to stitch the audio files together into a podcast episode. Although there is a more streamlined podcast script feature in the paid version of ElevenLabs, I made do with the resources available.
In all, creating a six-minute audio clip took about half a day. However considering the alternative of finding and coordinating with someone to discuss the topic, I felt the time investment was justified.
Take a listen to this podcast episode
Response to the podcast has been positive. For anyone considering a similar project, this can be a great way to share what you've learned in a more accessible way and help others gain skills too.
Throughout the project, you will develop your understanding of the topic, and it lets you be curious about these subjects and how they will help achieve new goals in your current workplace. Converting this to a podcast adds a fun goal to aim for and certainly reinforces what you've learned, much like when you have to give training or teach a topic.
When it comes to learning, start small. This could be a short conversation on the way to work, or five minutes spent with a podcast recapping a topic. These are some ways generative AI can help keep us learning and with job demands. The key is to be motivated and stay curious to learn more.