Somewhere in my career someone once told me. “Work smart, not hard”. So with that in mind, what to do if you are good at automating processes and want to complete a certification?
Well read on and find out.
I wanted to complete, and hopefully soon certify in one of AWS’s certifications. I have made a ton of notes from the course. Almost rewriting it word by word. The problem was it was still a lot of information to take in.
I wanted a more compact version of the course. For one , to be able to remember what I learned in the beginning of the course and also to perhaps listen to it in my car.
So what to do? Well this is where the power of AI comes into play. Yes, that AI everyone and their second uncle from the “smart” side of the family is talking about. GPT, or more precise GPT3.5. I wanted GPT to summarize my notes for me and give me a more compact version. So I sat for hours and hours copying and pasting text right? No!. At the time of this article there is not API for ChatGPT, only for GPT3. (Most websites get this wrong). So I had to use GPT3 for the summarization. I did however ask ChatGPT to write the code for me. Well mostly. I had to help it along here and there and fix some of the bugs myself. Below is a short clip of the script doing it’s thing. Not much to look at, but seeing the text pop up in the terminal and then as a text file is kind of awesome.
Since I know GPT tends to sometime leave things out I decided to have it summarize each section of my notes twice. Once as a shortened paragraph and then as a list of just the facts. That way I can increase the chances of transferring all the information to the summarized version.
So now my notes are in a different format, so what? Well now that it’s a bit more condensed I thought it would be nice to listen to it in the car. I mean I so spend about 3 hours driving to and from work, when I’m not working from home that it. So I might as well make good use of it. So I wanted to use AWS’s Polly to convert the txt into speech (TTS). I have used a number to TTS software in the past and as just a short while ago this was the best TTS I could find. Also it’s easy to integrate with it, unlike some other online TTS sites.
How did I do this. Well I used GPT again of course. It still only got me about 80% of the way there. And after trying to point it to the right direction, and failing, I decided to finish the code myself. Good to know I still have a job, for now. (Play dramatic music in the background)
And finally after some tweaking to the script I managed to do a batch convert of all the files GPT summarized for me.
Have a listen to one sample below:
So that is it right? Well as someone said once, “Everything doing is worth overdoing!”. So following that line of thinking I ask my partner to record a couple of sample audio clips for me. I then used the services of a relatively new company called Elevenlabs to train a model based on her voice. Basically cloning her voice. Although I must admit the cloned voice does not really sound like her. Probably because the base model was training on mostly American accents. Also at the moment the costs of using this method could start to get rather expensive real quick based on how many files I have to convert “back” to speech. But it was a fun experiment none the less. Imagining my partner reading my notes back to me while I drive.
Feel free to have listen to one sample I made “her” read.
So now it’s time to see if the approach will work. Or if I will fall asleep at the wheel listening to my noted read back to me. Only time will tell. I am of course not only relaying on my notes and are going though multiple courses and doing a couple of mock exams while I’m at it. So hopefully if all goes well I can update this page showing my certification one of these days.
Wish me luck!