Chat GPT told me some porkies, but it does say it can make mistakes…

Photo by Alex Knight on Unsplash

All this AI malarky, where is it taking us?

I have a real enthusiasm for AI at the moment. I’ve set up a GPT account (no affiliation btw) & use it to fast research many many tasks that used to take me hours to find the same answers on search engines.

Those are the plus points of AI, and at present, despite the bad press & scaremongering in certain social circles, this way of shortcutting our research methods does have a real benefit. Of course, with the benefits do come some detrimental effects.

One, for example, is the reduced cognitive need for tangental thinking. The way my brain works is pretty interesting (to myself at least haha), and I find AI sort of removes the need for abstract thinking when doing internet research. The AI models rapidly scour all available sources (ahh yes but what if they miss some important singular piece of data). See, my mind just found a 2nd reason as I was trying to explain the first issue with AI. I couldn’t think that up again if I wasn’t typing this out, so the mind is a complex & sometimes unpredictable thing. That’s why the mind is different. I’m not saying AI isn’t improving rapidly, but the quirks of our minds will always outdo anything that is using pre written information for its output.

Thirdly (as the second reason already found itself), and in my mind the most important at present, is the fact that the latest AI models (at time of writing) can actually gaslight the human in some instances. I had read about this happening a few months ago in a newspaper article, but I now have first hand experience. In fact I stupidly followed the AI model’s info a little too literally, & bought a photo enhancing software for a trial month, when all along the AI had the ability to do batch photo edits.

All I was after was to enhance some photos for inclusion in a book I wanted to print. The AI model allowed singular photo enhancement, but this had to be done live in the chat window. The AI told me to add my batch of photos into a folder, upload it, & it would return the folder complete with my photos enhanced ready for my book. I followed the advice, & after a long pause, I was asked to reduce the number from 100 photos to 50 as the capacity was too much. I did so, it failed again, & it asked me to reduce the photos to 25. Same again, it crashed, it said drop to 10. It crashed yet again, & said it was unable to process more than 1 photo at a time. I then asked it if it could automate this process to allow my 100 pics to be enhanced one by one. It said currently, no. It then suggested a well known software package that would give similar results. By this time I had wasted a couple of days in my spare time, so I decided on trialling the suggested software. The result was disappointing. The AI photo enhancement had already proved to be absolutely amazing, but the software was nothing like as good.

I started to ask AI in different ways why it had misled me, & it kept apologising, saying it had in fact misled me, & that currently it was unable to do the tasks it had in fact agreed to do. I then found out that I could in fact do batch AI photo enhancement using the OpenAI portal using python scripts. I was able to do this (again with the help of the original AI model) & this also quickly wrote a code to allow me to do this from my PC. The main difference was that it charged for each edited photo, but considering the excellent quality it offered, I decided it was worth a punt.

It was ironic that the AI model didn’t tell me that its own backend system was able to do the task, and pointed me to 3rd party software. This shows there’s still a way to go, but at the same time it is highly useful, time saving, & unrivalled at certain tasks. If I was marketing this system I would certainly make sure it recommended its own engine to someone asking how to do a task like this.

I can say, however, that in the last 8 weeks I must have saved a few days worth of time by having the AI system in place. I use it for work issues, planning, weighing up decisions that are complex to do when they have many variables. In fact with a mind like mine that can sometimes overcomplicate decision making, having a system that will give positive & negative points to guide me is more reliable than running it by someone who’s scrolling on their phone & only half listening. See my other article on this subject here

Thanks for reading!