Writing Better Prompts with ChatGPT

So, as far as prompts, one way to help improve your prompts, especially if you're not getting what you want, is start with your prompt that you have, and you can ask ChatGPT to help you write a better prompt. You can say, to ChatGPT, you can say, act as a prompt engineer, review this prompt and help me optimize it. Ask me any questions required to get the results you need for this optimization.

You can do that one chat, and then it's going to give you something, and then you can go create a new chat and then type in whatever you're trying to do, because in one chat, you're just trying to have it help you to write a good prompt. And then in a new chat, you're going to use whatever that prompt is to do whatever you're trying to do. So, if a prompt just isn't working, this is a method that you could try.

I kind of alluded to this earlier about, like, asking it for help to do things, but you could just say, context of, you know, pretend you're a prompt engineer, and then look at this and, like, ask me questions to do the best job you can do. And then it might ask questions and help you to write a better prompt. The goal of this is to write a good prompt with the help of ChatGPT.

Very interesting to think that the thing you're trying to ask questions, you're saying, help me to ask you a good question. Like, help me to write a good prompt to get you to do what I want you to do. But you don't know all of what it wants to know, sometimes.

So, if you say, how can I help you, like, ask me information, it can help to complete the prompt and make it a more full, complete prompt. So, I think it helps to just think of this as, it's an assistant, and sometimes it's okay to ask for help doing something. You can say, I'm not getting the results I want.

AI Classes & Bootcamps

• Live & Hands-on
• In NYC or Online
• Learn From Experts
• Free Retake
• Small Class Sizes
• 1-on-1 Bonus Training

Named a Top Bootcamp by Forbes, Fortune & Time Out

Learn More

How can we do this? And maybe it can give you some ideas. Talk to it. Sometimes it writes too much.

And, of course, we saw earlier, we could say to give it some specific thing of, like, how many words, how many sentences. Also, one funny thing that I saw somebody, they said, no yapping. Like, they just, they said, you know, write this, and no yapping.

Like, and it was just a short, two little words to just say, like, don't kind of go off track and, like, just be to the point, be concise. But no yapping can be sometimes used to be said to be concise. As far as better prompts, I showed you earlier about the little eye icon to get the prompt that ChatGPT sends to DALL-E.

So, I showed you that earlier. Now, once you have created a chat, sometimes you want to share that with somebody else so they can see it. Just so they can see it.

So, there's a share button up here. And if you share this, you can create a link to this chat. No.

So, let's say I create a new incognito window here where they're not even logged in. And this goes and shows them that. Now, this is just for them to see the results.

They can't do anything in here. Although they could continue the conversation in their own thing, which kind of copies it over to their account. So, they want to continue doing whatever they want to, to ask it more questions, to refine things further.

But when you make this link, it took a snapshot of what you had at that moment to say, here, you can take a look at this. Anything you add to that conversation afterwards will not be included with the link because it's a link to how it was at that time. You're just saying, hey, I want you to see this conversation that I had because maybe I want you to give me some input as to which version you like.

Or, you know, you could send somebody the results and they can copy and paste out of these. But it's just a way to say, hey, I've got this stuff and I want you to see it. If you go into your settings, this is where you can see, for example, under data controls, you can see your shared links that you've created.

So, if you ever want to go back and get rid of that link, you could delete it or you could copy and go back to the original source chat. But if you delete it, now the link is gone. So, the link doesn't work anymore.

So, it's one way you can share things if you want. It's not a collaboration. They can't add to it or change it.

They're just a way for them to see it. Memory. So, we saw memory earlier.

What's the deal with memory? So, memory is a paid feature. And the idea is that as you're chatting, certain things will trigger it to say, oh, that's a memory. I'm going to remember that.

It deems whatever it thinks is memorable and tells you that it's updated memory. Or you can also say something to it to say, like, I'm a writer specializing in SEO web content. And if you just talk to it, like, I didn't say it to do anything.

I'm just telling them about myself. Those are things that will trigger a memory. And so, the idea of memories is that they're going to be used across your chats from now on.

So, as you create new chats, it'll remember those things. So, you don't have to keep typing in every single time saying, I'm a writer specializing in SEO web content. You could say that in every single chat, but you'd have to say that in every single chat.

Memories just save you some typing because it's remembering all those things across chats. If you don't want it to remember something, because maybe it accidentally said memory updated, and you're like, no, no, no, don't remember that. Like, that was just for this one-off thing.

You can just go in there and tell it to forget it. Or if something changes, you can tell it to forget it. So, for example, let's say you say, I own a hair salon.

So, I'm going to create a new chat here, and I'm going to say, I own a hair salon. So, that's going to trigger a memory and say, oh, okay, I'm going to remember that in the future. Again, this will only work in paid accounts.

Oh, that's wonderful. Let me know if you'd like some help with anything related to your salon. And also, I could say, I have an 8-year-old daughter who loves cats, which in my case is true, because I do.

I have an 8-year-old daughter who loves cats. There's nothing really for it to do other than create a memory. Oh, that's sweet.

She loves cats. Let me know anything. So, now, okay, it's remembering that.

I can always go in, and I can check out my settings in the data controls. This is where I tell, you know, whether we're going to do training. I manage my shared links.

And, oh, sorry, personalization is where we do memories. So, I can either turn memory off. If I don't want memory on, I could turn it off if that bothers you.

Or if I want to leave it on, but I want to go back and manage, we can take things on and off. So, like, maybe here, I don't want to remember that, so I'm going to delete that. But I can keep the other ones.

If you want to get rid of them all, you can clear all your memories. So, it knows that I have a hair salon. It knows that I have a daughter who loves 8-year-old cats.

Oh, sorry, I have an 8-year-old daughter who loves cats. So, let's say, okay, so, yeah, so, later, I'm going to say, create a birthday card for my daughter. That's going to be another chat.

So, now it knows about my daughter, right? So, create a birthday card for my daughter. It knows she loves cats. So, do you think we're going to see some cats? Notice it has 8 instead of birthday.

8. Happy 8th birthday. Now, the only thing is. It should be 9. Exactly.

Because if she doesn't have a birthday, it should be smart enough to realize she's going to be 9. But it's not quite smart enough for that. But it didn't get the cats part, right? And actually, the text is getting better, but it's still not good. It's an 8th leg.

Actually, that happy 8th birthday, although it should be like 8th birthday. Actually, that I can read two words is actually good. Like, they're slowly getting better.

But, like, also, this is kind of a little weird. Like, the lines are a little bit smoochy, kind of weird. But, like, also, it's a photo of a card rather than, like, the actual card that I could print.

So, I'd have to refine this. But I was very generic. I wonder if I said for her upcoming birthday.

I wonder if it could figure that out, that it's going to be her 9th birthday. That would be smart. Although it should have figured that out.

Okay. So, it wrote a card. So, I'll say create an image of a birthday card I can use.

That's where you have to. If you go in assuming that it knows what you're thinking, that's the problem. We need to be very specific in what we're saying.

Okay. Actually, it didn't put the 9. Oh, no, it did put 8. 8H, not 8th. And there's no way to correct it.

Change the 8 to a 9th. So, I could highlight this and say change this to 9th. Let's see if it gets it right.

That's actually a really good text for Chatship TV. They do keep working on it. Text was complete gibberish in the beginning.

Like, you couldn't read it at all. It's actually not bad. Still a little smushy.

There's always these little smudges. I don't know what it is about the smudges that you notice every once in a while. Like, that's an AI thing.

No, they just keep up. Okay. But hey, at least it doesn't say 8 anymore.

But, yeah, it's a start. But notice the memory part, right? So, that was kind of the point, is that it was remembering things. Also, give me a social media marketing plan for this coming month.

Tailor to your hair salon. Because they know that I have a hair salon. So, they know I want to market my business because I own a hair salon.

So, that's just to show the memory part working, right? So, that way I don't have to keep saying it. Now, you might be like, I don't want to pay that much. I could just say I own a hair salon.

I mean, you could. You certainly could. Memories do save you a little bit, but, you know.

Also, memories could be things like I prefer to have things written in a certain style. Those could be other types of memories as well. But there might be a better way, instead of doing them as memories, there might be a better way of doing that.

And that is, oh, and by the way, there's temporary chats. So, when you create a new chat, you can do a temporary chat. And a temporary chat, it's kind of like incognito in a browser.

So, those chats are not saved in your history. Those chats do not use your memories. They do not create memories.

So, they're kind of like a blank slate. They say forget about all the memories. Don't create memories.

They're also not used to train AI models. So, even if you say normally to train the models, those don't do anything. So, they're like, they're going to go away.

Technically, they say they might keep a copy for their days, just for safety purposes, I guess. To, like, review accounts for doing bad things, I guess. But after that, they're gone.

So, they're not saved in your accounts. People can't go back to them and see those. So, that's temporary chats.

But custom instructions. Custom instructions are for free and paid users. And in some ways, they're kind of similar to memories.

But you have to more explicitly go and set them. So, with a custom instruction, you have some specific, what they call custom instructions. What would you like ChatGPT to know about you to provide better responses? So, this is information.

It's kind of like memories. But this predates memories. So, this feature is available for all users, free and paid.

And you have to go and type this stuff in. Memories is, they try to make it more convenient for people. Like, as you were just normally conversing with ChatGPT, memories kind of get automatically created.

Whereas, custom instructions, you have to go out of your way to type in these custom instructions. But everybody gets these. So, you could put in, I own a hair salon.

I have an eight-year-old daughter. Or whatever. So, these are things that it should know about you so that it can provide better responses.

Then here, how would you like ChatGPT to respond? You could say things like, I like SEO-friendly content that has headings that are easy to scan. I like short paragraphs and frequent use of bulleted lists or numbered lists. I want to target adult learners.

So, use adult, you know, professional language. Or I like to use, you know, I'm an elementary school teacher. So, target, you know, third grade reading level.

And then it will always use those across all your chats. So, a lot like memories in many ways. But memories are kind of automatically created on the fly.

Whereas these, you have to go out of your way to kind of go in and set these. So, you can go in to customize ChatGPT. Everybody can do this, free or paid.

And you can put in your custom instructions of what it should know about you. So, you're describing yourself. What do you do? Where are you? What are your goals? Like, things that it should know so that when it's writing its responses, it has more context about you and what you do.

And then, how do you want it to craft its responses? So, how long or short? How casual or formal? How do you want it to write it? Those kind of things. And once you set these, it will be for everything. Yes.

All your future chats. Yeah, all your future chats. So, enable for new chats.

So, all your new chats. It won't go back for your previous chats. But your new chats that you create and free and paid accounts will use these custom instructions.

So, it's kind of like a permanent memory. And you get to say how you want it to respond. So, you don't have to keep saying this stuff over and over and over again every single time.