I’ve used ChatGPT for quite awhile as a paying customer.  Here are some of the things I use ChatGPT for daily:

• Google Sheets formula troubleshooting
• Google App Scripts creation
• Python scripting
• Italian conversation
• Asking questions about Italian language usage
• general translations
• working on business marketing ideas
• Lots more

Things other than “scripting” usually cause me no issues.  They are mostly subjective things like “is this word usage weird” or “can you help me make this word usage less weird” or “I have these 5 ingredients that need to be used up today, can we figure out a recipe”.

When you get into actual programming though, that’s where ChatGPT just goes belly up!

Today I described three different kinds of script writers to ChatGPT:

1. People who generally understand scripting, but are busy and just want some help working through an issue, and who can take somewhat straightforward, but maybe incomplete instructions and roll with them.
2. People who are tech savvy enough, but who have never really had traditional scripting education who kind of hack their way through things but are interested in the “why” because the next time they might not need to ask if they figure out how to do something.
3. People who at least understand what a script is, but it’s all a foreign language to them, and they just need a finished script to do the work, and don’t really have a great desire to understand “why”.

I fit into the #2 group above. I am a bit of a programming hack (as my software engineer husband would say). I futz about with commands until I get one to work, and then I say “hallelujah”, implement the script, and remember the parts that are important, or create a document cheat sheet that reminds me of how to do a specific function for another time.

For #1, it’s totally acceptable to receive information like “Here’s the fix, replace these two variable and resave”. That person will be able to look at your changes, and figure out how to make those changes.

For #3, you would need to just create a new, edited script, including all of the changes necessary, with instructions to try it out, and let you know if it worked. If it didn’t work, you would need to present the entire script again, so that the #3 person could just replace the script and try again.

Those are the two extremes.

I’m a #2. Fairly simple changes, I can figure out for myself, but today I’ve spent over an hour trying to edit my script by using instructions ChatGPT gave me (that would have worked for a #1 but are confusing for a #2), figuring out where the changes belong in the flow of the script, and then trying to fix spaces vs. tabs!  I mean, it gave me instructions to replace “XYZ with ABC” and XYZ didn’t even exist in my script – in other words it was just “making sh*t” up that would work in some context, but had nothing to do with the script we were working on.

So, I decided that I needed to be clearer with my Chatbot and let her (mine is a she) know that rather than giving me vague instructions on how to replace a certain section of the script, she would simply present me with a brand new script with the changes necessary, and let me know what she had changed and why.

My issue today regarded a script I have that downloads Swiss German and Swiss Italian podcasts to a folder in PCloud drive, so that I can listen to them in my iPhone app of choice “CloudBeats”.  Awhile back, we had worked through this script, perfecting it so that it creates the file names with:

yyyymmdd <name of episode>

that makes them sort properly in the app.

Today I worked on installing “whisper” so that I could create transcripts of the Italian podcasts.  I find that if I listen to an Italian podcast, then read the transcript, then listen again, I have a better overall comprehension experience than just trying to listen without a transcript.  So after I got whisper working properly, before I kicked it off in a cron job to automatically transcribe every new Italian podcast, I needed to address an issue where I get duplicate files!  Any filename with an accented character seems to download again, as though it didn’t already exist on the disk.

So we started working on fixing this problem.

The first thing ChatGPT suggested was that I edit a couple of sections. I asked for clarification as to exactly what/where I needed to edit, and then got errors that I had a mix of tabs and spaces in my python script!  Really?  We can discuss python’s inability to deal with a mixture of tabs and spaces at another time, but rather than just giving me a new, functional script, ChapGPT tried to troubleshoot my tabs/spaces issue for almost an hour!

Finally I asked for a new script incorporating the important changes.

ChatGPT had my original script.  Shouldn’t this be easier for her than for me?  Apparently not. The new script totally changed the output directory, and the filename structure – d’uh!  So there were of course no duplicates, because the files were being downloaded into a new fresh directory that had no existing files!

Then, my files had all been formatted with the date in yyyymmdd format at the beginning of the filename, so that they would sort properly.  That part of the script had been totally removed as well, so it didn’t matter that they were in a new folder.  They would all have had different file names anyway.

It was clear that rather than using the existing script, the bot had just taken the most immediate need – dealing with the extended characters – and had just created a new script entirely, omitting all of the other naming functions in the script.

I mentioned once almost two years ago when I started using this “assistant”, that if she had been a paid employee she would have been fired for incompetence and general insubordination!

It seems obvious to me that our robot overlords are simply not ready to effectively take over just yet.  And then it dawned on me that it’s just like our current situation in the US with Trump and Elon Musk. ChatGPT doesn’t have to be efficient in it’s destruction of our accumulated knowledge. It just has to persistently wear us down in order to be effective in our subjugation.

I, for one, am not gonna take it from ChatGPT OR Trump/Musk.  I’ve used my brain and I’ve fixed my silly script.  Maybe if we all would just use our collective brains, we could fix this bigger problem as well!