ChatGPT vs. custom GPTs - ChatGPT Tutorial

- [Instructor] Now, when should we use ChatGPT versus a custom GPT? To understand that we need to understand the strengths and weaknesses of each interface. Here in the traditional ChatGPT interface, we have a very flexible general purpose assistant that can help us with any sort of task from math, coding, writing, research, and so on. We can write to the assistant by asking anything that we want, like write me a one paragraph summary of large language models for a 6-year-old. And the model, after a little bit of time, will stream out an answer that hopefully addresses the requirements of our request. Now, this flexible interface makes it for a super powerful helper for any sort of task that you might encounter in a computer. It gives us a few functionalities that can be extremely useful. For example, we can click over here and add data from other apps like Google Drive or OneDrive. We can add photos and files. We can include tools like Deep Research to generate long deep research reports. We can create images that are tailored through a very custom and detailed request. We can search the web for information and we can create a canvas that will split our screen into two, allowing us to write and prototype apps or essays and so on. So the ChatGPT interface is extremely flexible. However, if we find ourselves rewriting a lot of custom and detailed prompts in our conversations with ChatGPT, that's the signal that maybe using a custom GPT might be a good idea. For example, in here on my top left, I'm going to open up my Italian tutor custom GPT. Now this custom GPT was created with instructions to help me practice and study Italian. So if I say hello, welcome to this course about custom GPTs, and I send this without any other information to this custom GPT, what happens is that the custom GPT will have my instructions that were predefined into its configuration, kick in and produce a response that addresses the structure that I instructed this custom GPT to perform. So as you can see, we see the initial translation and then a couple of bullet points with explanations about this sentence in Italian. So makes it for a great tutor on Italian and I didn't have to write a lot of information to get exactly what I was looking for. If I wanted to get the same output with ChatGPT, I would've had to write a very long prompt with a lot of detailed information about what I wanted the output to look like. So custom GPTs are great for super specialized tasks where what we're looking for is a model that can learn from specific instructions and custom documents to generate answers and providers with help for automations and things like that that perhaps will be too cumbersome to perform with the general ChatGPT interface.