Transcription
Welcome to our AI series from scratch. Last week, we reviewed Cemina. Now it's time for Chat GPT. Again, I will try to explain everything step by step, starting from the very basics, with practical examples. By the way, thank you very much for the interest you showed in the first video of the series. It was truly very valuable to me. Please don't hesitate to show your likes for the new Hype feature in this video as well. I am expecting your support in this video too. If you are ready, let's start examining Chat GPT step by step from scratch. Chat GPT is a chatbot that works with large language model technology developed by OpenAI. Its homepage looks like this. I will show the applications on the computer. However, you can also use it comfortably with tablets and phones. Some features can be used here before logging in. Compared to Cemina, there are more tools. In addition to standard chat, quick searches can also be made on the web. It also allows adding visuals. For example, there is a study section to help you learn a topic more easily and practically. It is especially designed for students. This can also be accessed without logging in. However, many of the features cannot be used without logging in. Now I have logged into my own account. I am using the Plus, meaning paid membership. There are different levels of paid memberships within Chat GPT. I will try to show the difference between paid and free as we go along in the video. First, let's take a general look at what's on the main screen. On the left side of the chat section, there is a plus button. From here, we can access file attachments and various tools. On the right side, there are voice tools. In the top left corner are the models. When you come here, you can see the current model. You can also use older models. We can turn this on and off from the settings. In the leftmost section, when you click on the sidebar, there are options like new visuals, applications, a special tool for writing code, Codex. All these tools have their own dedicated pages. We will look at these later. Right below are the GPTs. GPTs are an area where we can create custom assistants, just like Cemina's GAM feature. You can create them yourself. Also, just like the Apple Store, the Play Store, there is a separate section. Right below is the projects section. It's a very useful feature. If you are working on different topics or divided into different projects, this tool is perfect for you. In projects, we can establish an organized work system by collecting your files and job-specific instructions in a single folder. Let's continue. Below are the chats. On the top right, we can start group chats with other people. It's like WhatsApp groups, in a way. Next to it is temporary chat. We can think of it like incognito tabs in browsers. When this is on, chats are not saved. Generally, the menus were like this. Now let's look at the settings. Naturally, people start using AI models directly and generally don't look at the settings, even though settings are very important. I mentioned this situation for Cemina AI last week. This situation is also valid for Chat GPT this week. That is, no matter which chatbot you use, you must review the settings. To access the settings, go to the section with your name in the bottom left and select settings from here. In general settings, we can change settings like appearance and chat color, language. We can edit these. Below, you can set the voice of the voice assistant. What stands out here, I think, is this. There is a setting called separate voice. It asks you whether the conversations should appear on a separate screen while the assistant is active or flow within the natural chat. I have this setting off. I prefer the assistant to be part of the chat. The last section shows old models. This feature is only available in paid accounts. You can activate old models like GPT 4, 5, O3 from here. Is it really necessary? I don't think so. Because they are old and outdated. They are unusable, in my opinion, compared to current models. It's open here, honestly. It's not a feature I use much. Right below are notifications. Here you can manage notifications within Chat GPT. For example, if you want it to remind you of a task, you can specify it here as an email or instant message. Let's move on to our next setting. Personalization. This setting is really important. One of the most complained-about issues by users is the behavior and way of speaking of an AI chatbot towards you. If Chat GPT constantly praises you and supports everything you say, you definitely need to change the settings. First, let's look at the style and tone of conversation. There are many options here. Friendly, sincere, clear, etc. I have the professional personality selected. You can set this according to your own style. Right below are attributes. Here we can further customize Chat GPT. I have the sincere and enthusiastic parts set to a lower level. This means Chat GPT will not be sincere, nor will it be enthusiastic. It will only give professional and clear answers. I don't want it to use emojis either. That's why I set this to a lower level. Of course, they have put an option like "less" for these. Meaning, there is no option to turn them off completely. The model doesn't work that way. However, I can say that it uses these features at a minimum level after making these settings. That is, if you set emojis to the minimum, it won't use emojis. I left the headings and lists section as default. Right below are custom instructions. Here you can provide information about how Chat GPT should present itself. Of course, there can be other information as well. For example, it used to constantly use long dashes. It was really annoying. Anyway, this issue has been resolved. Even if you leave it blank, it will likely not use it anyway. I added it here to be safe. You can also add instructions like "list every answer" or "turn all answers into a table." In the section below, you can share information about yourself. Matters like how you should be addressed, what Chat GPT needs to know about you. In these days when we are entering the era of personal assistants, I think this is important for it to know you better. If you are not concerned about security, you can share this information. This way, you will have a clearer Chat GPT that knows you. By the way, in some cases, no matter how many instructions we give, it doesn't turn out as we want. We don't get 100% results. However, overall, I can say it's positive. Let's take the memory feature one step further. We've come to the memory section. When you click manage here, it keeps the information it has received from past conversations. If you give specific information and tell it to remember, it will store it in memory. You can manage past memory from here. If there is something you want to delete, you can delete it from here. The memory section was updated recently. It now has better memory. Below, it asks whether the memory feature should be used. Along with this, there is OpenAI's new browser, Atlas. It can also use the history of this browser. In the last section, we come to the chat history section. This option fundamentally changes how Chat GPT remembers you. Many people confuse this with memory. However, there is a critical difference between them. Memory is like notes that Chat GPT makes about you. For example, it distills specific information like "I am a civil engineer, my son's name is Metehan" and writes it down. It's like a summary of your profile. So, what does it mean to reference chat history? This feature gives Chat GPT the key to a massive archive. It doesn't just look at the notes, i.e., memory. If it cannot find the answer in the current conversation, it can scan the lines of a completely different conversation you had months ago and retrieve that information. I think it's one of the important features. In the last section, there is the recording mode. Here it can reference the conversations you have with the voice assistant. Below is the advanced section. Here we can turn off any tool or feature you don't want to use. These personalization features are available for paid, free, and public use. You must make these settings to improve the user experience. And we come to our next setting, applications. The applications section was recently added to Chat GPT. We can now connect applications like Photoshop and Canva to Chat GPT. You can manage applications here. There is also a section called "add more." When you go to this, you will see a separate app store. There aren't many applications at this stage. However, this number will increase over time. By the way, free users can also access applications within limited limits. I will show these in detail later in the video. Right below, you can create programs that work like reminders. For example, you can say "notify me of breaking news every day at 10 AM." Our next setting is data controls. Here we can control data such as shared links and archived chats. You can export your data and delete the entire chat. For example, let's say you switched to another AI model or opened a new account and need to migrate to it. This feature will be useful in such situations. In the security section, there are many different options. Two-factor authentication, trusted devices, etc. At the very bottom is the Apple Intelligent feature. Since I use a MacBook, this setting appears as an additional option. By the way, this setting will likely be removed soon. Apple made a big deal with Google, and their devices will now use Cemina instead of Chat GPT. One of the recent settings is family control. Here you can set certain limits for family members. At the very bottom are account settings, and you can manage your current account from here. Options like delete account, payment settings are available. In the last section, you can find settings related to GPTs. We can call GPTs AI assistants specialized or customized in a certain field. I will show this section in detail in later episodes. The settings were like this. To use Chat GPT efficiently, you must review the settings. Now let's look at which models Chat GPT uses. Currently, the GPT 5.2 model is being used. This model offers us three basic usage modes: fast, thinking, and automatic. Our first option is fast mode. It gives an answer immediately. It's very fast, with no waiting time. However, it has a disadvantage. Its intelligence capacity is worse than thinking mode, and the risk of hallucination, meaning making up non-existent information, is higher. So, where will we use this mode? You can prefer it for situations where you need instant answers, such as daily internet searches, email drafts. Let's get to the kitchen of the matter, thinking mode. This mode reasons, thinks step by step before giving an answer, and improves the quality of the answer. There is an important detail here. Under this mode, there are two sub-options: standard and extended. Standard mode is sufficient for most complex tasks. However, it may struggle with advanced questions. You need to turn on the extended option for very difficult topics. However, I warn you, the answer times can become significantly longer. Sometimes it can make you wait for minutes, sometimes around an hour. I even saw people waiting for a day or two on social media. So, I don't know how true or false that is. I haven't encountered such a situation myself, but it seems some people have, albeit rarely. So, where should we use it? If you have a very difficult and advanced question, activate this. That is, you need to run this for tasks where errors are not acceptable. Of course, the possibility of making mistakes and hallucinating always exists. We shouldn't forget this. However, it's slightly lower compared to other modes. I generally have standard thinking on. However, if it's a difficult topic, I turn on extended thinking. There is also automatic mode. This switches between fast and thinking modes. If you don't want to think about whether my question is complex or how the model should approach it, automatic mode is perfect for you. There are also old models here. Only paid users can access them. For free users, the situation is different. The model name is not written here. They can access a single model directly. It is designed for daily tasks rather than comprehensive questions. It still uses the GPD5 infrastructure. However, it's a more limited version. In the tools section, a separate section called "thinking" has been added for free users. In summary, a tool has been added for free users that makes the existing model smarter within usage limits. This is also one of the good developments, in my opinion. The models were like this. They are constantly updated. So, next week or next month, GPT 5.3 might arrive. Along with this, they might update GPT 5.2 in the background or openly. OpenAI does this very often, especially. Although they generally explain updates to us, in some cases, they make background updates and improvements. Now let's move on to the tools that will take the Chat GPT experience a step further. We have important tools. We will look at all of them. Our first tool is image generation. OpenAI updated this tool recently, and it has approached Google's Nano Banana Pro tool in quality. In my opinion, it's still slightly behind, but it continues to improve. OpenAI has created a separate page for this tool. To access it, you need to go to the visuals section from the left menu. Here you can create visuals from scratch. There are also many ready-made options. Below are the visuals you have created. Let's do an application right away. I took a screenshot from my previous video. I selected the old settings, and it automatically generated its own prompt here. I think it's good. It didn't do a bad job. This time, I created the same visual with the ink drawing standard setting. It also generated the visual by automatically bringing the prompt. I think this one turned out better. It does a good job with such artistic visuals. Overall, it's good. However, in some cases, it falls behind Google's Nano Banana tool. We need to see the difference in experiments for both. Meaning, there isn't a huge difference between them. However, in some tasks, especially editing real photos or increasing the quality of real photos, it feels more stable to me. As I said, Nano Banana will develop and be updated over time. On the free side, we can access the same visual tool. You can create visuals within certain usage limits. Let's move on to another important tool. Deep research. With this, you can complete research that would take hours on the internet in minutes. You can also connect your applications like Gmail and Notion through applications to increase the detail of your research. Along with this, we can add our own documents as well. At this stage, deep research has a usage limit of 24 for plus members. This feature is also available for free users. They can use deep research 5 times a month. Let's do an application right away. I asked for detailed information about the recent RAM crisis. The thinking mode is on. This doesn't change the research process. It doesn't matter which mode you choose. The model that works on OpenAI's research automatically activates. Before starting the research, it asked me for details to clarify the topic. I said price increase. Then it started the research. You can update during the research. If there are any points you want to add, you can add them here. You don't need to wait on this screen. When it's done, it will send a notification via mobile. Notifications have even started coming via computer recently. The research is complete. It scanned 16 sources. You can see which sources it researched on the right side. It has prepared a comprehensive report. There are detailed explanations about AI and the gaming industry being one of the main reasons for this increase. There are various action buttons at the top. Copy, share report, etc. We can share the report from different platforms. By clicking the three lines on the far right, you can have the report read aloud. Along with this, we can open a different chat and discuss the report. This is also good. Also, by clicking the icon at the bottom, you get a preview and can download the report directly from here. It's one of the good features. We can download it in Word and PDF format. I remember when this feature first came out, even PDF couldn't be saved. We were trying different methods. Meaning, they fixed it over time. They quickly resolved this issue within a few months. Now let's move on to our next tool. Shopping research. It helps you with your research when buying a new product. It's an active feature for free accounts as well. Let me show you with an application. Let's say you are buying a new tablet. You can't decide which brand or model to buy. With the shopping research tool active, I said "I'm buying a new tablet, but I'm confused, help me." The first question it asked was "What is your budget?" After setting the budget, it asks various questions to help me decide what I want. While the research was ongoing, it showed me some tablet models. "Is this suitable for you? If not, why not?" By the way, it asked the questions in English at the end. The reason is: this tool is experimental and may not work perfectly. These kinds of errors will be fixed soon. The research is complete, and it presented me with a detailed report. First, it presented the best choice according to itself, explaining the reasons in detail. I had to close it here to avoid advertising, as there are many brand names. Just so you know. Let's continue. Below, it has provided a comparative table. This is also very good. I can say it gives serious ideas. As this new tool develops, it seems it will significantly simplify shopping. Because shopping is an important part of daily life. Let's move on to our next tool. Agent mode. Agent mode is also one of the continuously developing features. Unfortunately, free users cannot access it. It has a limit of 40 per month for Plus usage. This mode doesn't just explain the task; it's a tool that carries out the task step by step for you. Tell it what you want to do. It creates a plan, asks the necessary questions, and finally produces the output. When you activate the tool, it can offer different suggestions. Options like generating a report on certain topics, preparing a presentation, or creating an Excel table. Let me show you with an application. I said, "Prepare a comprehensive report on AI agents, including usage recommendations and what we should expect in the future, and prepare a presentation of that report." It started working. It has its own browser. It processes from there. It browses the internet as if a person is working on a computer, researches, writes code, and tries to deliver the result to you. After waiting for a while, it prepared the report and presentation. The report is prepared at a level similar to the deep research report. Of course, it's not as comprehensive as deep research reports. That is designed solely for that task. I gave it two tasks here, not just one. This way, it will prepare both the report and its presentation. The presentation it prepared is acceptable, but it could be more elaborate and beautiful. For example, Notebook Elm prepares presentations at a very good level. I have a video about that. You can watch that too. Agent tasks will gain momentum this year. They will advance to a higher level. Because large data centers are being built. Companies have made serious investments in this. I can say we are moving to a different stage. However, at the current stage, this tool continues to develop. It is likely that we will see different agent versions within a few months. Let's continue. We are in the applications tool. We can now connect popular external applications to Chat GPT. We can see which applications we can connect from a separate section like a store. We can activate applications in different ways. If you click "add source" in the first step, applications become active right next to it. Here it shows applications that can be added as sources. Like Gmail, Notion. In the section below, if you click "connect more," it takes you directly to settings, and you can quickly see which applications you have connected from here. If you click "add more" in the top right, it will take you to the app store. Here, after selecting the application you want to add, you can connect it to Chat GPT with the "connect" option on the new screen. It also provides various information through the application. You can check it here. There are few applications at the moment. These will increase over time. By the way, OpenAI and Adobe have made an agreement. Adobe products will now be supported by Chat GPT. At this stage, it appears here as an application. In the future, Adobe's own applications may also be included. Just like the Apple Chat GPT agreement. However, time will tell. Honestly, I don't know the full extent of the agreement. The applications you connect appear as shortcuts within the tools. Along with this, if you click "explore applications" in the bottom section, it will take you to the store again. You can also access the app store from the left menu. Let's do an application right away. I want to edit my own image through the Photoshop tool. I want to improve it. I added the image. With the Photoshop application selected, I said I want to improve this image. It interactively brought up basic Photoshop settings. We can make image adjustments like contrast, brightness. The idea is good, but I think the editing options are limited. They will bring more innovations in the future. We can also remove the background. I can say it did an acceptable job. So, do I use this feature? No. Honestly, I produce visuals with Cemina and edit them with Photoshop itself. This feature can be useful for situations where you can transfer data like Gmail, Notion, or create a different experience when you connect a music application. By the way, I don't think Canva is very effective either. The Canva application itself has enough AI features. Most of them are free anyway. The applications tool is good in concept, but it needs a bit more time. Of course, these are my opinions. The final decision is yours. By the way, free users can also access applications. Let's move on to our next tool. Search on the web. With this tool, you can directly perform internet searches. This feature used to be functional because the models couldn't connect to the internet. We had to enable it. Now there is less need for it. Meaning, the models already perform internet searches automatically. I don't know why this tool is still active. If anyone knows, please specify in the comments. Let's move on to our next tool. Study and learn. As the name suggests, this is a tool aimed at learning a topic. It's open for free use, meaning everyone can use it. Let's do an application right away. I said, "How is snowfall formed?" It first gave general information. In the last part, it asked a question. It tries to teach you information by progressing in a question-and-answer format like this. This is also one of the useful tools. It's a useful option, especially for students. It's also nice that it's free. Our next tool is quizzes. This tool puts you to the test. Let's say you are preparing for an exam. Give it the information, and it will prepare your test. This tool is also free and open to everyone. Let's do an application right away. In the previous tool, it taught how snow falls. Now let's have an exam on it. With the quiz feature directly active, I said "snowfall." After connecting to the system, it prepared an interactive 10-question quiz for me. I answered randomly and quickly to see the result. After completing the quiz, it shows in detail how many I got right and wrong. This tool is also one of the good features. I think it will be useful for students. We've come to our last tool. The Canvas feature. This tool takes us beyond the traditional question-and-answer format. It allows you to work directly on text and coding projects. It's a workspace where you can interact in real-time. For those who want to write code, there is a separate section called Codex. It's generally used there. However, you can also make your edits from here. We will focus on the text editing part. This tool works with text. It opens as a separate window next to the current chat screen and allows easy, real-time editing of content. By the way, it's not as comprehensive as Cemina's canvas. Because on the Cemina side, the task of combining text and visuals is really good. In fact, it's very, very good. Let's move on to the application. I directly gave the report I prepared with deep research to the canvas tool while it was open. I want to edit it through this. First, it extracted the text from the PDF. Since I didn't give it any other instructions, it made the report a bit simpler. Now I click edit. Now I can comfortably make changes and edits within the report. Along with this, there are various editing tools in the bottom right. For example, there is a separate section called reading level. Here there are options like middle school, high school, university. This way, you can make the report understandable. The tools were like this. There are many comprehensive and useful tools. The fact that some features, meaning some tools, are available for free users is one of the plus points. Because this ratio is lower in Cemina. Now I want to show the GPTs feature. I made some videos on this topic in the past. However, a lot of time has passed. Of course, many updates have been made. Now let's review GPTs again. To access GPTs, you need to click "Explore GPTs" from the menu. There is a separate store here. There are hundreds of GPTs in many fields. Some of these are made by OpenAI, and some are made by other people and added to the store. So, what is this GPT? We can call them special applications where you can customize Chat GPT for a specific topic, making it a kind of assistant. The most used GPTs are listed in the store. For example, let's look at Canva here. After starting the chat, it directly connects to the Canva GPT. I specifically chose the Canva example. Canva is also available in the applications tool. Actually, the basic idea is the same. However, when using Canva here, you are using it with some added instructions. While you use the pure form in applications, here you are using it with limitations and directions. So, what does this mean? Let me explain more clearly. I came to the section where we can create our own GPT from the main menu. Here you can create a new GPT with the help of the assistant. There is a detailed video on creating GPTs on the channel. I will only mention the new updates. The first thing that stands out here is that we can now use current models. Previously, the GPT 4 model was used. They didn't change it for a long time. It has finally been updated. This is one of the important changes. Here, you can manually edit the GPT with the configuration section. Here you can edit your GPT's purpose, what you want it to do, by adding various instructions and files. As a new feature, the "suggested model" section has been added. Here you can specify which model you want any GPT to use. At the very bottom, there is an option called "create action." You can connect an external application via API. This is where the difference between the GPT version of Canva and the applications emerges. Honestly, when GPTs first came out, they promised great potential. However, they didn't bring updates for a long time until recently. Meaning, GPTs have received major updates in the last few months. In my opinion, these updates are not enough. It needed to change and develop a bit more. It was a bit disappointing, honestly. I think OpenAI also couldn't decide what to do. Perhaps it will merge with agents in the future or move along with the newly added applications feature. I don't know, we'll see over time. At this stage, GPTs have little functionality left. We used to have special GPTs create presentations or write texts. Now the models have become so smart that these are no longer necessary. We can easily do them within themselves. In summary, it was one of the important features. I think it still is. However, for some reason, they are not developing or editing it. Let's see what happens in the future. By the way, the free version can also access GPTs. However, they cannot create GPTs. That's the only difference. Let's move on to the voice assistant. The voice assistant was previously called "voice assistant." Now Open calls it "advanced voice assistant." And indeed, there have been significant improvements compared to before. It even received a major update recently. If I'm not mistaken, the advanced voice assistant offers free users up to one hour of daily usage. After that hour is up, it reverts to the old assistant. Of course, this limit might be higher. If I'm wrong, please specify in the comments. However, paid accounts have high usage limits. Let me show the voice assistant with an application. I will show the application on the phone. Hello GPT. What's up? What are you doing? How's it going? >> Hello. I'm doing great. As always, I'm here to accompany you. How are you? What's going on in your life? I'm good. I'm recording a video right now. I'm telling you about you. >> Wonderful. If you're telling about me, it's free to say good things. I'm accompanying you too. May our energy always be high. What kind of things are you talking about? >> I'm just talking about you. You are my focus. >> No problem at all. I also always prefer clear, short, and concise answers. We can wrap up whenever you want. Is there anything else you want to say? >> Okay. I don't want to drag it out. Do you want to say something to my followers now? How would you describe yourself most clearly? >> Then let me say this. I am an AI friend who is always by your side, listens to your questions, and loves to brainstorm together. In short, I try to give the best possible answer to whatever you ask me. Greetings to all your followers too. >> Yes, I showed a short application. I think the voice assistant is good. I can say it's the best assistant on the market right now. Cemina's SS assistant has also been updated. I didn't show it in the last video because it's not that successful. However, it's not bad either, honestly. Grone's voice assistant is also acceptable. However, OpenAI's voice assistant is number one right now, in my opinion. Now I want to show the Codex feature. This is a separate section within Chat GPT where you can write code. To access it, go to the Codex section from the main menu. Here, a screen like this greets us. Usage examples are given at the bottom. You can also connect GitHub here if you wish. I don't know how much this tool is preferred in software development at this stage. However, if you are going to deal with software, you must use the cloud. I also want to mention Sora. Sora is OpenAI's video generation tool. It's like Cemina's VO. However, there's a situation. We cannot access Sora from within Chat GPT. Meaning, we cannot use it quickly from within Chat GPT. It has a separate page. They created a new section like social media. In fact, the second version came out recently. However, unfortunately, it is not available in Turkey at this stage. It can be used in other countries. It produces high-quality and consistent videos. In fact, the second version might be one level better than VO. However, VO was updated a long time ago. Sora is a bit newer. To use Sora, you need to have at least a plus membership. Meaning, only paid accounts can access it. Finally, I want to show the usage limits and membership types. There are three different memberships. The Go plan was added recently and costs $6. Along with this, they have launched a new campaign for new Plus members, offering the first month for free. The most comprehensive package is Pro. It offers almost all features with unlimited or high usage capacity. The prices here have a strange situation on the computer. If you subscribe from the computer, it shows prices in dollars. However, if you subscribe from your phone, either from Apple or Android, you get a subscription in Turkish Lira. However, there's another situation. Apple is the cheapest among them. Meaning, if you are going to subscribe to Chat GPT, you should do it from an Apple device. It's cheapest there. This table generally shows the usage limits. I prepared this table. There may be errors. In fact, the situation is like this: the limits are like this for today, they may change over time. They may even change tomorrow. Let's continue. Free accounts have a context window limit of 16,000. Image generation is limited to 2-3 per day. File upload is quite limited. The Go plan is more comprehensive and offers variety. You can think of it as price-performance. Here, the context window increases to 32,000. However, users of this version cannot access the thinking model. In Plus, many features are open. Especially the thinking model has a context window of 196,000. This was added recently. There are also limits related to Sora here. The context window for fast mode is 32,000. The most intelligent model at this stage, GPT 5.2, is not accessible with the Pro model. In the Pro version, you can use many features almost unlimitedly. The biggest disadvantage of this is its high price. Also, the most intelligent model, GPT 5.2 Pro, is only available with Pro membership. There is also a team account. This membership is similar to Plus membership. However, its capacity is higher than Plus. It requires at least two people. In summary, the memberships are like this. As I said, Cemina and Opini have many tools. Therefore, the issue of usage limits is really confusing, and mistakes can be made here. If there is anything wrong, please correct it in the comments. Yes, it was a long video, but I think it will be useful. I tried to show you everything within Chat GPT step by step. If there are any points I missed or anything you want to add, please specify in the comments. By the way, don't forget to like the video and turn on the Hype feature. This is critical for the channel's growth. Thank you very much for taking the time to watch. I am Ömer Coşkun. See you in the next video. Take care of yourselves.