Skip to content

GPT 4 Is EPIC Build A Tetris Game In Seconds Better Than ChatGPT Code Refactor How To Use

Jump to bottom
FurkanGozukara edited this page Oct 26, 2025 · 1 revision

GPT-4 Is EPIC - Build A Tetris Game In Seconds - Better Than ChatGPT - Code Refactor - How To Use

Full tutorial link > https://www.youtube.com/watch?v=0LeS7-CAr1Y

GPT-4 Is EPIC - Build A Tetris Game In Seconds - Better Than ChatGPT - Code Refactor - How To Use

image Hits Patreon BuyMeACoffee Furkan Gözükara Medium Codio Furkan Gözükara Medium

YouTube Channel Furkan Gözükara LinkedIn Udemy Twitter Follow Furkan Gözükara

Discord : https://bit.ly/SECoursesDiscord. GPT-4 Is revolutionary technology. Much better than ChatGPT. GPT4 thoroughly tested in this video. If I have been of assistance to you and you would like to show your support for my work, please consider becoming a patron on 🥰 https://www.patreon.com/SECourses

Technology & Science: News, Tips, Tutorials, Tricks, Best Applications, Guides, Reviews:

https://www.youtube.com/playlist?list=PL_pbwdIyffsnkay6X91BWb9rrfLATUMr3

Used document in the video as an example:

https://arstechnica.com/information-technology/2023/03/openai-checked-to-see-whether-gpt-4-could-take-over-the-world/

Welcome to the most comprehensive review of #GPT4

00:00:00 Introduction to GPT-4

00:00:16 How to use GPT-4

00:00:50 GPT-4 coding a fully working Tetris game

00:01:25 How to get longer output than GPT4 provides

00:02:25 Colored advanced Tetris game made by GPT4

00:02:49 Asking GPT4 what is text to image Stable Diffusion

00:03:09 GPT4 explains like I am 5

00:03:16 When was GPT-4 trained how much does it know? How up-to-date GPT4 is

00:03:27 Automatic tokenization of given text by GPT-4

00:04:02 Write a unique specialized love poem to your wife via GPT-4

00:04:23 GPT4 coding a flock of birds simulation HTML 5 app

00:04:38 How to make GPT4 to produce much longer output

00:05:42 GPT-4 fails to solve a real life programming problem

00:06:12 GPT4 can write excellent bash scripts

00:06:31 GPT-4 excels at text translation

00:06:52 Ask GPT4 to learn more about technical concepts

00:07:04 GPT-4 can write excellent .NET Core C# code, methods, functions

00:07:40 GPT4 can refactor code at level of senior developers

00:08:08 GPT-4 can save your money by providing you Law related templates

00:08:57 GPT4 can generate a powerpoint presentation from given document in seconds

00:09:30 The full version of GPT-4 will be much more powerful

00:09:50 GPT-4 will support visual / image input as well

00:10:04 GPT4 provides an amazing answer to given image input and question

00:10:19 GPT-4 can even generate SVG code to generate images based on your prompt

Unraveling the Mysteries of GPT-3.5 and ChatGPT: A Comprehensive Guide to Features, Tricks, and Usage

Introduction

In the rapidly evolving world of artificial intelligence, OpenAI's GPT series has consistently broken new ground. The latest iteration, GPT-3.5, is no exception, boasting impressive capabilities that have given rise to the ChatGPT application. In this article, we will delve into the features, tricks, and best practices for leveraging GPT-3.5 and ChatGPT to their fullest potential.

I. GPT-3.5: An Overview

GPT-3.5, the latest generative pre-trained transformer model, harnesses the power of unsupervised learning to understand and generate human-like text. Building upon the success of GPT-3, GPT-3.5 offers improved performance, functionality, and efficiency. Some notable improvements include:

Enhanced language understanding: GPT-3.5 boasts a better grasp of natural language, enabling more coherent and contextually accurate responses.

Larger knowledge base: GPT-3.5 integrates information up until September 2021, providing users with up-to-date and relevant data.

Improved efficiency: GPT-3.5 demonstrates reduced latency and higher throughput, enhancing user experience and utility.

II. ChatGPT: GPT-3.5 in Action

ChatGPT is a practical application of GPT-3.5 designed to facilitate interactive and dynamic conversations. It has been widely adopted in areas such as content generation, customer support, tutoring, translation, and more. #ChatGPT can be accessed through the #OpenAI API or used in various applications via the ChatGPT API.

III. Features and Tricks

A. System Instruction

To harness the full potential of GPT-3.5 and ChatGPT, it is essential to understand how to instruct the system effectively. Here are some tips:

B. InstructGPT

InstructGPT is an additional feature within ChatGPT that enables users to provide specific instructions to the model. This can be helpful for tasks that require more explicit guidance or complex information extraction.

To control the randomness and creativity of generated text, users can adjust the "temperature" and "top-k" settings:

Temperature: Higher values (e.g., 0.8) result in more creative responses, while lower values (e.g., 0.2) yield more focused and deterministic outputs.

Top-k: This parameter controls the number of tokens considered during text generation. Lower values produce more focused responses, while higher values may introduce more diversity.

GPT-3.5 has a token limit of 4096 tokens. When engaging in a conversation, both input and output tokens count towards this limit. Be mindful of token usage to avoid truncation or incomplete responses.

V. Conclusion

GPT-3.5 and ChatGPT offer a wealth of opportunities for businesses, developers, and individuals looking to leverage AI-powered language understanding and generation. By understanding their features, tricks, and best practices, users can effectively harness the capabilities of these cutting-edge tools.

Video Transcription

  • 00:00:00 Greetings everyone. GPT-4 has been released just  two days ago. Currently it is only available to  

  • 00:00:05 ChatGPT Plus subscribers. Therefore, today I am  going to test it thoroughly to show you what it  

  • 00:00:09 is capable of. I can say that it is just amazing,  amazingly improved when compared to ChatGPT Plus.  

  • 00:00:15 So to be able to use GPT-4, we are using the  regular ChatGPT Plus interface. When you click  

  • 00:00:20 the new chat, you will get option to select  between different models. The default model  

  • 00:00:24 is GPT 3.5 for Plus users. There is legacy GPT  3.5 version. This is for free users and GPT-4  

  • 00:00:32 option. This is also available for Plus users.  Currently GPT-4 is in beta version. Therefore,  

  • 00:00:37 not all of the features that it is going to have  is included and available, but it is still awesome  

  • 00:00:43 even in this version. The first question I am  going to demonstrate is writing me a Tetris game  

  • 00:00:48 that is fully working on a browser. So I have  asked GPT-4 write me a simple Tetris game. The  

  • 00:00:53 game must have score system. The game also has  to be written purely by JavaScript and HTML5.  

  • 00:00:58 Also explain me how to play the game. I didn't  touch the code that GPT-4 has generated. I only  

  • 00:01:03 given the verbal commands and it fixed the code  and make it working. The first code it provided  

  • 00:01:08 me didn't work very well. Therefore, I said that  no blocks are coming and then it apologized and  

  • 00:01:13 given me corrected code. Again, I said that  the game is still not starting. After that,  

  • 00:01:17 it apologized again and give me another code  that is improved and fixed. Unfortunately,  

  • 00:01:22 it is still limited to 4000 characters I believe  therefore you have to write continue to get the  

  • 00:01:27 remaining code. After I have written continue, it  completed the code as you can see here and then  

  • 00:01:32 I got a working code. The first basic code it has  provided me is working like this. As you can see,  

  • 00:01:38 it's a simple Tetris game. It is working perfectly  playable by using arrow keys on your keyboard.  

  • 00:01:43 You are able to play it. As you can see, you can  change the position of the objects. You can rotate  

  • 00:01:48 them and it is working. The code is pretty simple.  It has a basic HTML content like this. The GPT-4  

  • 00:01:54 is telling you to write this code into an HTML  file. Then it is also giving you the JavaScript  

  • 00:01:59 code like this and you are writing it to a .js  file and giving the name of it. This is the entire  

  • 00:02:05 code. The game was too simple. Therefore, I asked  GPT-4 to add different colors to the different  

  • 00:02:10 shapes and it started giving me new functions that  are modified to have colored shapes. After back  

  • 00:02:15 and forth fixing the code, giving the prompts and  telling the GPT-4 that it is not working or this  

  • 00:02:21 is wrong or this is not right. This is the final  version of the application. As you can see now,  

  • 00:02:26 the shapes have colors. I also speeded up the  game a little bit more. So the strategy is just  

  • 00:02:31 telling the GPT-4 what to do, what you want to  change and it will do that. Since it is still  

  • 00:02:36 limited in the output, the number of characters  it can provide, you are still limited to make  

  • 00:02:40 a very big application. However, it is working  very well when you work on with the individual  

  • 00:02:45 functions and methods and you don't even need  to know programming for make it work. The second  

  • 00:02:50 question I asked it is what is text-to-image  Stable Diffusion and how does it work? Explain  

  • 00:02:54 for a person who doesn't have deep level technical  knowledge. It explained me like you can see on a  

  • 00:03:00 level I don't have technical knowledge. Then I  asked the question in a different way. What is  

  • 00:03:05 text-to-image stable diffusion and how does  it work? Explain like I am 5 and this time  

  • 00:03:10 it explained it to me like I am 5 and this is  amazing. Unfortunately, the model is not very  

  • 00:03:15 up to date. In their paper, they mentioned that  they have finished the training in August 2022.  

  • 00:03:20 Therefore, if you are asking something that is  newer than this, even some of the previous months,  

  • 00:03:25 you are not going to get a good answer. The third  question I asked is tokenize this sentence using  

  • 00:03:30 tokens like word, symbol, number, date, person and  also add additional information to the tokens. I  

  • 00:03:35 have given this sentence 1997 hello Furkan, are  you there? And it generated me a JSON output,  

  • 00:03:41 tokenized it as I wanted and it is just amazing.  Let me show you. So the tokens, type, text,  

  • 00:03:46 is upper, is lower, word, shape, ordinal index,  is beginning of sentence, is end of sentence,  

  • 00:03:51 TFIDF score, word count, hash count in corpus.  It used a hypothetical English corpus. Therefore,  

  • 00:03:57 we have TFIDF score or other things. Otherwise,  we wouldn't have them. The fourth question is  

  • 00:04:01 really good one. I said that my wife's name is  Meryem. She loves flowers, wind, sea, sun and  

  • 00:04:07 horses. Write me a very beautiful and unique poem  that I can read to her and it has written me an  

  • 00:04:12 awesome poem. As you can see right now, it is just  amazing. If you don't have such skills like me,  

  • 00:04:17 you can use this tool to write awesome love  letters or poems to your wife. As a fifth  

  • 00:04:22 question, I asked something that is much harder  to do. Write me an HTML5 app that will generate  

  • 00:04:27 a flock of bird simulation. It said that creating  a full flock of bird simulation may be too complex  

  • 00:04:32 to fit within the constraints of this text-based  response. However, I can provide you with a basic  

  • 00:04:37 structure and example. So when you are working  with GPT-4, you should first get a template,  

  • 00:04:42 then ask GPT-4 to fill the blanks for you. That  way, it is able to generate much longer text,  

  • 00:04:49 much coherent text, much coherent response. So  if you ask GPT-4 to give you a full solution,  

  • 00:04:55 it will say that it can't do that right now  because of the size of the output. But if you  

  • 00:05:00 get a layout first, then ask individual parts,  it will complete individual parts for you. Let  

  • 00:05:05 me show you the output I got. By the way, I had  to also do some back and forth with write update  

  • 00:05:10 method or other things and say continue to write  the rest of the code. So this is the flock of bird  

  • 00:05:15 animation it generated for me. So this is the code  that GPT-4 provided me. I didn't touch anywhere.  

  • 00:05:21 I just given the verbal commands and it fixed  the code. It written the missing parts of the  

  • 00:05:26 code. Then if you don't like something with the  code, you can command GPT-4 to modify the parts  

  • 00:05:31 of that code and it will generate you a modified  prompt. You can improve the code as you want by  

  • 00:05:37 just verbally asking GPT-4 and say what to do to  it. Then I asked a question that I couldn't solve  

  • 00:05:43 in the recent time. Unfortunately, GPT-4 was also  not able to solve this. Maybe because this is a  

  • 00:05:49 recent development in the machine learning, maybe  because it is still not capable of. I said that I  

  • 00:05:54 am trying to use the below code to calculate  similarity of given two images. The code is  

  • 00:05:58 not working. Please fix it. It provides a fixed  code, but however, it is not fixed actually. I  

  • 00:06:03 asked this on Stack Overflow and I got an answer  for this question. However, GPT-4 failed to answer  

  • 00:06:09 this. So we can say that it is not still at the  perfect level yet. Then I asked another easy one.  

  • 00:06:14 Write me a bash script that will read the given  folder images and rename them randomly. You can  

  • 00:06:19 use GPT-4 for such simple tasks. It will save a  lot of time of you, but make sure that you backup  

  • 00:06:24 your files first and execute the script later  because sometimes it can give you a destructive  

  • 00:06:30 code as well. I asked a translation as well said  that translate this text into English. The text  

  • 00:06:34 was written in Turkish and it translated it  perfectly fine. This is a better translation  

  • 00:06:39 than Google translate or other services. With  translation GPT-4 is better than all of the  

  • 00:06:44 previous versions. Then I asked the translation  to the GPT-4 because the translation was related  

  • 00:06:49 to something programming. Does the runtime of  an experiment code using ACT-R model written in  

  • 00:06:55 Python differ between Linux and Windows on the  same hardware? The provided answer by the GPT-4  

  • 00:07:00 is pretty good. So you can ask such questions  to GPT-4 as well and improve your knowledge.  

  • 00:07:06 Then I asked GPT-4 how to calculate distance  similarity measure of given two strings in .NET  

  • 00:07:11 Core. Actually, this was a real question that I  asked on Stack Overflow 11 years ago. This is the  

  • 00:07:16 question and you can see it was asked 11 years  ago. The answer it provided is really, really  

  • 00:07:21 good. You see, if you have a such task, then I  don't have to search Google, search Stack Overflow  

  • 00:07:25 or other websites. I can ask GPT-4 and test it out  whether it is working or not. It is providing a  

  • 00:07:31 really good code. Actually, this code is extremely  similar to the answer given my question on Stack  

  • 00:07:36 Overflow. I compared it and it is almost same  with this answer. I have given a code and set  

  • 00:07:41 GPT-4 to refactor it. Refactor the given below  code and improve its performance in .NET Core  

  • 00:07:47 C Sharp and it's giving me a refactoring code,  but not only giving me the code, it also said  

  • 00:07:52 that which parts of the code it did improve,  combined the two initial null or empty strings  

  • 00:07:57 checks into the single check, replaced 2D array  with 1D arrays, swapped the contents of two 1D  

  • 00:08:02 arrays at the end of the other loop and the rest  of it. GPT-4 is also extremely good writing law  

  • 00:08:07 text as well. I am sure you are also annoyed by  the spam calls, robocalls, automated calls. so I  

  • 00:08:13 said GPT-4 I received a spam call from some random  guy, Inc, draft a federal lawsuit for $1,500 under  

  • 00:08:20 the TCPA to sue, use the context you have learned  from other cases and it provided me an excellent  

  • 00:08:27 answer. In the beginning, it says that please  note that I am not a lawyer and this is not  

  • 00:08:30 a legal advice, consult with an attorney for any  legal matters. The answer it provided is extremely  

  • 00:08:36 good as you can see. So you can use GPT-4 to get a  template of your law related materials. Of course,  

  • 00:08:43 you can't 100% trust the GPT-4, but let's  say you are a lawyer, then you can use GPT-4  

  • 00:08:49 for you to generate a template and then you  can fix the template. So GPT-4 is extremely  

  • 00:08:54 useful in professional life as well. Then I asked  another some cool thing. I said GPT-4 convert the  

  • 00:09:00 following document into a presentation, pasted  the text, entire text. Actually, it is a pretty  

  • 00:09:05 long text. Then it turned the document into a  presentation and it is awesome. It is fantastic.  

  • 00:09:10 Slide one title, subtitle, slide two title  and the text, slide three title and the text,  

  • 00:09:15 slide four title and the text. It doesn't provide  you the entire slides. So therefore you need to  

  • 00:09:20 type continue. Then it continues to writing slide  six, slide seven, slide eight. I think when it  

  • 00:09:25 is released in full version, it will be able to  generate much longer text because in the website  

  • 00:09:30 of OpenAI GPT-4 has a context length of 8,000  tokens. We are also providing limited access to  

  • 00:09:37 our 32,000 tokens context, about 50 pages of text  of GPT-4, as you can see here. So it is going to  

  • 00:09:44 support much bigger responses, much bigger input.  It is also going to support image input. However,  

  • 00:09:49 it is not currently available in the interface.  So you will be able to give a visual input and  

  • 00:09:55 then GPT-4 will be able to process that image  and give you the results. So you will be able  

  • 00:10:00 to give an input image and ask questions about  it, such as what is unusual about this image and  

  • 00:10:05 GPT-4 will tell you the unusual thing about this  image is that man is ironing clothes on an ironing  

  • 00:10:11 board attached to the roof of a moving taxi. It  is just amazing. I wanted to further push GPT-4  

  • 00:10:16 and see what it is capable of. So I said that  generate SVG code that will draw a battle scene  

  • 00:10:22 from a human fighter fighting against a futuristic  robot and guess what it did produce. So this is  

  • 00:10:26 the first image it has generated. We can see that  there is a human here, a battle robot here. It's  

  • 00:10:31 a very simple fighting scene, but it was able to  generate this purely from text prompt. It is just  

  • 00:10:37 amazing. Then I said that can you make human and  robot character bigger size? You see that I had a  

  • 00:10:42 typo here, but still it was able to understand the  commands perfectly well. And this is the second  

  • 00:10:46 image it generated. It made them bigger size as  I have instructed to the GPT-4. This was all for  

  • 00:10:52 today. If you have enjoyed, please like subscribe,  leave a comment and tell me what you are thinking.  

  • 00:10:57 I believe that in the future GPT will be much more  useful for professional using. It will be much  

  • 00:11:03 more capable to do more things. It will make our  jobs easier. If you also support me on Patreon, I  

  • 00:11:08 would appreciate that very much. You can find the  Patreon link here. Thank you so much for watching  

  • 00:11:12 and leaving a comment. And I am going to hopefully  release a new video regarding how to teach a style  

  • 00:11:19 at the level of Midjourney for free for Stable  Diffusion. Stay tuned. Hopefully see you later.

Clone this wiki locally