Episode Transcript
[00:00:00] Foreign.
[00:00:05] Welcome to ChatGPT Curious, a podcast for people who are, well, curious about ChatGPT. I'm, um, your host, Dr. Shantae Cofield, also known as the Maestro, and I created this show to explore what ChatGPT actually is. Really, though, are the files in the computer, how to use it, and what it might mean for how we think, work, create, and move through life. Whether you're skeptical, intrigued, or already experimenting, you're in the right place. All that I ask is that you stay curious. All right, let's get into it.
[00:00:38] Hello, hello, hello, my curious people, and welcome to episode 23 of Chat GPT Curious. I am your grateful host, the Maestro, and today we are talking about using Chat GPT's personalization settings to get better responses. Y', all. This. This is a dope thing. So I up. Well, not totally, but kind of, because this portion of ChatGPT is not new. Uh, but I'm really going to focus on this custom instructions section. It's not new. I have just been skipping over it, and I recently learned that if I fill it out, I get way better responses, AKA responses in a style that I much prefer. So in an effort to inject some humanness into this episode, before we get into the. The details and all this, the tech stuff, let me start off by saying Merry Christmas to those of you who celebrate, and happy holiday season to those who celebrate something else. And Happy Thursday to those of you who don't celebrate anything. All right, no judgment. Do your thing. Love this for you. Christmas is not a big deal here. Thanksgiving is a big holiday. Uh, so Lex and I will be chilling on Christmas, probably go to the beach. That's like a thing. If I'm in California, I'm gonna go to the beach because other people can't. And I don't do it to say haha to them, but I'm like, hey, I want to take advantage of this thing that I know other people would love as well, and I love it, so I'm gonna go do it. So if the weather cooperates, go to the beach, maybe go on a bike ride. I don't know. We'll see what we're doing today. Um, but more importantly, it is my brother's birthday, one of my brothers, Justin. So a very happy birthday to him. He doesn't listen to this podcast, at least I don't think he does, but happy birthday nonetheless. Uh, Lex and I just got back from vacation. Vacation in Playa del Carmen, and it was dope. And during that time, I was actually watching some AI Vibe coding videos from one of the self PA courses that I'm taking. And this particular course, I've mentioned this name before, it's from my friend Khee. And, um, it's marketed towards folks in the finance world. But the technical components are universally, universally applicable. That was a weird word. I kind of swallowed that. Universally applicable.
[00:02:44] Uh, so I was. I love this stuff, y'. All. You know, I'm. I'm all in. But I was cruising through some of the earlier lessons because it was like background stuff and let's talk about talking about LLMs and how they work and things like that. So I was cruising through them. But one of the things he brought up that I have not done is using the custom instruction fields in the personalization section of ChatGPT.
[00:03:05] So way back in episode nine, right, we're on episode 23. Way back, way, way, way back, episode nine, I actually outlined a bunch of the personalization features, went through basically all of them, but I 100% skipped over the custom instructions field, namely because I don't feel like free type questions. I don't really like free type questions. And for whatever reason, I told myself that what I put in there would be ignored. I was wrong. So I will link that in the show notes, um, if you listen to that episode. But we're going to talk about this personalization section again, and we're going to focus on the custom instructions section because that shit makes a difference. So to access the personalization section of ChatGPT, you're just going to click on your name at the bottom of that menu that's on the left side of the screen. From there, you will just click on the word personalization and a new window will pop up. You can do this on your phone as well, right? The settings are universal to your account, so you click on personalization and a new window pops up.
[00:04:06] Once that window pops up, you got a few options. The first option is a dropdown menu and it lets you select the base style and tone. There are eight different options. There's the default. Professional, friendly, candid, quirky, efficient, nerdy, cynical.
[00:04:20] I have candid. But feel free to use whatever you want. There's explainers for each of the different personality types. The next section is what I want to highlight and what I have been skipping. Custom instructions.
[00:04:32] What has been objectively, very significantly impactful for me after changing this is using this area to tell Chat how I want it to respond. Again, this is a free type space, so you can put whatever you want in there, but you are limited to 1500 characters.
[00:04:51] To generate the text that goes in there, I had ChatGPT interview me. The original prompt I used was, I'd like you to interview me one question at a time. I always specify that because they'd be giving me 100 questions and I'm like, yo, it's too much. I'd like you to interview, interview me one question at a time so you can generate custom instructions that I will paste into the personalization section in ChatGPT.
[00:05:15] I would suggest that you start with that prompt or a similar prompt if you have no idea what you're wanting ChatGPT to focus on in terms of instructions and guidelines and interacting with you.
[00:05:26] So I put that in first. And what I quickly found was as, uh, it started interviewing me, I did in fact know what I wanted in terms of instructions, and that was instructions to help with brevity, decreasing hallucinizations. Wow. Hallucinizations. Hallucination. That was a hallucination right there. Decreasing hallucinations, minimizing sycophancy. So sycophancy is like being a yes man. Like, oh, yes, yes, you're the best. This is so smart. I would like minimize that. I don't need all that. And, uh, I wanted to avoid the initial fluff that it generates when providing answers, because even with the candid response, it still be like, that's a great question. Here's an answer with no fluff. And I'm like, you just gave me a bunch of fluff to tell me you're not gonna give me fluff. I don't, I don't need that. So I re prompted ChatGPT and I said, these questions are not helpful. Mainly, I want you to interview me so you can create instructions related to brevity, avoiding and minimizing hallucinations, sycophancy, and just giving me responses without the initial fluff. Like, that's a great question, maestro. Here's the answer without any fluff.
[00:06:26] I want you folks to notice you can speak to this thing, type to this thing like a person, right? Uh, you can have typos. You can just use regular language. Yes. When it gives you an answer. It's doing math to get that answ. But the ways in which you can communicate with it. That's fucking magic. You just talk with it and be like, yo, I fucking hate this. Change it up. And here's what I don't like. And it knows what you're, what you're saying, and it's going to give you different answers.
[00:06:49] So after I gave that second prompt.
[00:06:51] Uh, it reframed the questions and.
[00:06:55] Or I should say, after I reframed it, the questions that it gave me were much better.
[00:07:00] And it actually gave me both the question and bulleted list options for the answers. Amazing.
[00:07:06] Of note, I didn't prompt it to respond in that style, but you can, you can do that if it's not giving you that. If it's just like giving you a question and that's it, you can be like, hey, give me a question and then suggest answers in a bulleted list. Right? You can do whatever fuck you want. Um, but it gave me the answers in a bulleted list and I was like, this is great. So I could just pick from them or I could say all of these things or none of these things. So that was super fast, super helpful.
[00:07:29] So after it asked me all the questions and I gave it my responses, it generated a paste ready custom instructions, but I learned they were too long. I just copied them and pasted them and I was like, this is too long. It says there's a character limit of 1500. So I went back and I asked it to regenerate. I was like, regenerate this without losing any of the requirements that we established. And I took that thing and I pasted that into the custom instruction fields. And honestly, folks, I am loving the change. Like, it's so good. It is so helpful. The responses are streamlined. They're to the point. It flags when it's making inferences or when it's kind of like reasoning, right? If it's not, like, doesn't. I don't say if it doesn't. It never knows things, right? Jeffrey doesn't know things. But when it's. When it's inferring things, it says that and it flags it. It is not agreeable just to be agreeable, which is so dope. It will actually disagree with me. And I'm just like, this is what I want. This is how I want it to be used.
[00:08:25] I love it, absolutely love it. How you set it up will be up to you. But I can confirm that what goes in the custom instructions box does in fact matter and will 100% change ChatGPT's behavior. Okay, so that's like the most important thing I want you to take away from this episode. But I'm going to go through all the personalization settings and just so you have, you know, the lay of the land. So as for the remaining personalization settings, the next field underneath, uh, that custom instructions box is.
[00:08:58] It's like Three free type fields and it's like nickname, occupation, and more about you, meaning any interest, values, preferences to keep in mind. I actually, actually have those blank. You think I'd learn based on my recent revelation of how much of a game changer filling in custom, you know, the custom instruction box was. But Chat GPT already knows these things about me based on the fact that I've used it for like three years. So I don't, I don't feel the need to write that in there, but I think that it would be super helpful right out of the box. So if you're just new, if you're new to ChatGPT and you wanted to get to know you faster, definitely fill those out. I, um. The next section in the personalization settings is, um. I spoke about this in episode nine is Memory.
[00:09:40] Inside of memory, it has a manage button, right? There's a little button that says Manage on it where you can view a list of all of the things that chatgpt remembers about you. And in that list, uh, you can delete anything you don't want Chat to remember. It is, in my opinion, really cool to go through this and be like, oh, you picked that up. You picked that up. There's certain things that you're like, I did not want you to remember that. And I see why you bring this up sometimes. Like, if you ever, I don't know, are using this to help your clients in any way, then it will bring it. Sometimes it remembers those things and then it may kind of, you know, like meld them into the answers it gives you. And you're like, why do you keep bringing this up? And it's because of that. So you can go into memory and just delete that memory.
[00:10:17] Uh, the flip side of this is to add things to the memory. Whenever you're chatting with Chat GPT, just type add to memory. And it will, if you want it to remember something about you, and it will add it to that saved memory bank. And if you want to delete it, you can, um, but that's how you add things to, to that.
[00:10:34] Underneath that, within, uh, the memory section, there's toggle fields and they have. Each of those toggle fields has explanations I have for me personally, just in case you're wondering, my settings, I have reference save memories on, I have reference browser memories off, and I have reference chat history on. So you can make your own decisions with that. But that's how I have it. The main thing I want you to take away from that is going in and the Manage section and like auditing what it remembers about you.
[00:10:59] Really, really helpful.
[00:11:01] There are after this, two more sections in the personalization settings area. Record mode and Advanced Record mode. It's a single toggle on, off, uh, setting, and the Advanced section is an accordion section. So when you click it, a handful of options will show up that are toggle on and off as well. Each of the options for both Record Mode and the Advanced section has an explanation of the feature. So you can go and read that and make your own decisions. I personally have all of them toggled on. O n on. Right. And uh, that. That's the entirety of the personalization settings. Yes, you can change the voice for voice mode. I went over this in episode nine, but, um, that's in general settings. It's not in the personalization section. Um, but to me it is well worth it to spend some time in that personalization section. Uh, especially the custom instructions and managing memory. This episode, I'm making it intentionally shorter. So hopefully you can take the time that you save and go and do that like it will make a difference. It absolutely makes for a better user experience. And I think that it's only going to continue to improve, um, because OpenAI and, uh, what is a fucking weirdo, um, Sam Altman, they basically stated it, he announced a code red, what, two weeks ago when Gemini released their Gemini 3. 3. 3.0.
[00:12:21] And he was just like, oh, shit. And so basically when that happened, they made a commitment to improving memory and personalization, trying to really focus on that because they're attempting to maintain market share and really differentiate themselves from the other LLMs and OpenAI or ChatGPT. Their focus is the consumer. You have things like Claude or Anthropic. Right. They're going to enterprise. They're really focusing on coding more than anything.
[00:12:45] Uh, so we have ChatGPT here, OpenAI and they're like, yo, our stick. Our whole thing was going to the average consumer, now that there's other models that are as good with the tasks that people ask of them, how can we differentiate? How can we keep these people? And it is by focusing on the user experience, focusing on memory, uh, and personalization. So I think it's only going to continue to improve. Um, so I would say definitely for right now start off or for right now, definitely go and spend some time in that section. It doesn't take that long and it will make a big difference.
[00:13:17] Okay, so how I use ChatGPT this week? Each. Each episode. Wow. Had a little stumble there. Let me back it up. Each episode, if you don't know, I include a section where I briefly discuss how I use ChatGPT that week. This week, I use Chat GPT to help me redesign my workout split. Um, I'm pretty sure that I've mentioned this use case before in an early EP earlier episode, but I leaned on it again and I actually had it do way more this time. So last time it was largely just a sounding board and I was like, hey, I think about doing this exercise. What do you think about the order of this? Like, there's more so about. More so about kind of looking at exercise selection. This time I actually had Chat come up with the program. I was like, hey, make this. Here's my specifications, my requirements. And then we edited it together.
[00:14:02] Worth noting and tying back into the point of this episode is that at one point I prompted CHAT to remember that I am an advanced lifter and I have a doctorate in physical therapy. This is helpful because of how it will communicate with me and the exercises it will choose. Uh, uh, so it's worth saying, like, hey, add this to memory. Um, um, but that is how I use it this week and I was very happy with, with the results. So if you want to use it that way, go ahead, give it a try. Or not. I'm just here to share how I used it. All right, that, my friends, is all for today. Hopefully you found this episode helpful and if you did this time, I'm going to ask, consider sharing it with someone who you, who you know, is curious about ChatGPT. I think this is a, uh, helpful one in my opinion.
[00:14:42] And I, I want people to have good experiences with things. And this definitely, you know, changing the personalization settings and those custom instructions, it definitely changes your experience. So would love for as many folks to know about it as possible.
[00:14:55] Don't forget, don't forget I have a companion newsletter, the Curious Companion, that drops every Thursday. That is basically the podcast episode in text format. So if you prefer to read or you just want that written record, join the newsletter a fam. You can head to chat GPT curious.com forward/newsletter or just check out the link in the show notes. As always, my friends, endlessly, endlessly, one more time. Endlessly appreciative for every single one of you. Until we chat again next Thursday, stay curious.
