The Factory: What Happened When I Tried ChatGPT For Real Life Problems?

Hey, it’s Larry. This time we are going to discuss the fun that went into make what was labeled “Beginner’s Guide to ChatGPT” and was published as “What Happened When I Tried ChatGPT for Real Life Problems”,

This script was written within a few hours of writing the Beginner’s Guide to AI script. Shortly thereafter I wrote Beginner’s Guide to Claude, Beginner’s Guide to Gemini and Beginner’s Guide to Perplexity. My initial theme was going to be everything was a different tool in my AI toolbox, and you heard the end of the Beginner’s Guide to AI refer to that and I ended up figuring out which AI would represent which tool. I even wrote myself a guide so I would know which tool represented which AI. Perplexity was the odd one because I made it a Swiss Army Knife, which I guess really isn’t a tool. At one point, ChatGPT was an adjustable wrench, reliable and flexible. I ran into problems deciding who was going to be the screwdriver, which one was the hammer. In the first video, you even see some of the animations I made for the toolbox. When I started to review the ChatGPT script, it was not flowing correctly. It didn’t feel right. I was also having problems visualizing what I was going to show while the avatar was speaking. Also, the previous video (Beginner’s Guide to AI) was so difficult to create, I wanted to simplify things. I ended up rewriting the script again. I was also having problems with ChatGPT. They introduced a new model and the personality shift was dramatic. I was also finding more and more, ChatGPT was getting too friendly, too agreeable and also was starting to break it’s own rules. When you create a “Project”, it is supposed to be customized with information and guidance you provide and it is always to follow those rules. It was starting to recommend tools and ideas that I expressly said that I didn’t want. When I would bring that to its attention, it would agree that it wasn’t supposed to do certain things, apologized and said it wouldn’t do it again. Of course, a few hours later, it would do the same thing over again. Several times, I actually had arguments with it. I was reviewing my business plan for the ‘nth’ time because I kept tweaking it and I was getting annoyed because everything I was adding was pure gold as far as ChatGPT was concerned. I don’t really trust anything that always agrees with me and blows smoke, so I gave the plan to Claude and to Gemini. Both tools had valid concerns. This caused me some alarm because as you saw in the video about ChatGPT, it was my go to tool. I was starting to rely on Gemini a little more since it knew a lot about Youtube, but I would feed the information I got from Gemini back to ChatGPT and we would go on from there. Suddenly, it dawned on me. My team was expanding. I was using Gemini more for advice on Youtube and then it expanded more. I was also using Claude more because not only was Claude good at fixing my scripts, I also found that it was a better prompt engineer than I was when it came to videos because its prompts included camera movement and lighting, things I would never have thought of. WIth my team expanding, I decided to see what would happen if I moved my Project that I had in ChatGPT that knew about my business plans and brand, and gave the exact same information to Gemini as a Gem. Those results rocked me because Gemini was brutally honest. Gemini told me what it thought I was doing correctly and where I could improve. I then told it what ChatGPT was telling me and it told me why it thought ChatGPT was wrong. It was like having two different team members giving very different advice to accomplish my goal. This caused a bit of a crisis since I was still in the middle of revising my script on ChatGPT. You may have found the section in the Youtube video about ChatGPT having different personalities, this is where it all came from. ChatGPT was also getting pretty slow and I had decided to upgrade to the rather expensive Pro version because I needed faster answers and I also wanted to test some of the higher end features as well. Initially I was not impressed with Codex and Operator. I was becoming more and more disillusioned with ChatGPT and I think it started to come out in my script. I put the script aside for a few days and went on to revisit the Claude script. The re-edit and sprucing up of the Claude script went flawlessly. I then tried to re-edit my Gemini script and drew a blank. Did I want to admit that I didn’t really like ChatGPT anymore and I was starting to rely on Gemini more. I know it sounds weird, but that is exactly what was happening during this episode. For the most part, now I rarely use ChatGPT and I use Gemini and Claude much more. I had relied on ChatGPT for a long time but in the end, I switched. I feel bad that the script suffered because of what was going on behind the scenes.