Basic commands required to become a Chat GPT4 super user.

First of all, we will explain the concepts that we need to know to become a super user of Chat GPT. Chat GPT is a large language model trained on open artificial intelligence and is used for natural language generation and comprehension tasks. The model is used in various applications such as interactive chatbots, machine translation, question-answering systems, and text generation.

Chat GPT is trained on a large-scale dataset and thus possesses rich knowledge about language. It provides the capability to generate and customize various language models.

This article provides a detailed explanation of how to become a super user using Chat GPT and introduces methods to optimize language generation and comprehension tasks using the various features of the model.

An appropriate prompt is required to perform language generation tasks using Chat GPT. A prompt is the input sentence provided to Chat GPT to generate responses. The following three basic rules are highly recommended when writing a prompt.

Task description: Provide an explanation of the task for better understanding.

Example: Provide an example of the task to help Chat GPT understand what type of responses to generate.

Prompt: The sentence that Chat GPT must generate based on the description of task and examples.

For example, let's assume that we are performing a task to translate from English to Spanish. The task description is "Translate English to Spanish", and examples are “The word for table in Spanish is Mesa” and “The word for apple in Spanish is manzana.” These inform Chat GPT about what type of response to generate.

Once we know how to use Chat GPT, the next thing we need to understand is the prompt(command). We can use various prompting techniques without any knowledge of new programming technologies or machine learning algorithms. Prompting allows us to control the type and amount of information provided to the model and helps the model find the answers we want more easily. Let's take a look at seven prompts we need to be aware of to become a super user.

There are seven major types of prompts to command ChatGPT.

1. Standard Prompting

2. Role Prompting

3. Adding Personality to Your Prompts

4. Generating Knowledge

5. Chain of Thought Prompting

6. Prefix Tuning

7. GPT-3's In-Built Prompts

1. Standard prompting

Standard prompting is the most commonly used prompting method in Chat GPT. This method provides the model with questions and examples to help it easily find the answer we want. First, we write a description of the task to solve with some examples and a prompt that needs to be completed by Chat GPT, including omitted text the model needs to generate. By doing so, the model understands the context and output format required to perform the task.

Task description: Extract airport codes from this text.

Examples:

1. I want to fly from Los Angeles to Miami. The airport code is LAXNMIA.

2. I want to fly from Nashville to Kansas City. Please write airport codes for the airport.

Prompt: I want to fly from Orlando to Boston.

we can instruct Chat GPT to generate correct output format and perform the task using this method.

2.. Role prompting

Users can customize the behavior of ChatGPT by setting roles when it does not produce desired results. For example, if you want to practice job interview as a recruiting manager, you can run interview simulations for various positions such as software engineering or data science by having ChatGPT simulates the the role of a recruiting manager. ChatGPT asks about educational background, job experience, and other things that should be mentioned in an interview. Then, it can be customized to fit one's needs, regardless of the roles and used as a language tutor to practice foreign languages like Spanish or be transformed into a movie critic to analyze movies. Furthermore, you can gain inspiration from repositories where you can find several examples of using prompts such as 'act as a standard comedian.’

3. Personality prompting

Personality prompting adds a special tone, professional field, and other unique elements to the text generated by ChatGPT by including style and technical language in the prompts, and as a result, it evokes a uniqueness such ass formality and author's field in the generated text. For example, a prompt such as "Write the topic in the style of an expert with 10 years of experience in Field X" can specify the author's style, formality, and field to ChatGPT. To customize the output of ChatGPT further, however, prompts should be adjusted by adding technical language. For example, a generic prompt such as "Write a blog post of no more than 500 words about why AI will not replace humans" is highly likely to produce generic outputs.

4. knowledge-prompt

Useful information on the topic can be generated using the knowledge generation feature before generating the final response. For example, you can ask ChatGPT to generate five facts about why AI will not replace humans once the prompt which is "Write the blog post using the generated five facts” produces certain outputs. The outputs include useful information produced before generating the text and it leads to better results when it comes to writing a 500-word blog post.

This approach grants factual knowledge to the model in order to generate new information, encouraging the model to utilize existing knowledge during the process of generating new information.

Likewise, knowledge-prompt can be used to create general knowledge. It leads to an answer to the question, "What is the highest mountain in the world?” by inputting information such as where the highest mountain in the world is located, how high it is, what geographical features it has, and how climbers ascend the mountain, etc.

Therefore, the prompt is a very useful tool for generating knowledge. However, if the input information is ambiguous or imprecise, the information generated by the model can be of a poor-quality. It is important to enter declarative information as much as possible and review it to ensure that the input information is accurate and precise when it comes to using the prompt.

5. Chain of thought prompting

Unlike the conventional prompting approach, a chain of thought prompting encourages the model to generate intermediate inferences before providing the final answer to the problem. In other words, the model describes the reasoning process rather than giving a direct answer to the question. So, why is inference important? Previous research has shown that explaining the reasoning behind inferences leads to more accurate results.

For example, think of the question, "What happens if I took an umbrella with me when I left my home?". It requires logical reasoning, thus using chain of thought prompting can help us achieve better results.

To be specific, the prompt can be formulated as follows if it is created based on the question, "What would happen if you took your umbrella when you left home?"

"What will happen if you take an umbrella with you when you leave home? Make the following three inferences."

1. There is a possibility of rain.

2. Taking an umbrella could lead to predicting the future.

3. It could improve one's mood.

Now, we induce model to provide a final answer to the question after making these inferences. In this case, the model will generate an intermediate answer, "If you take an umbrella, there is a high probability of rain," and then provide the final answer, "If you take an umbrella, you will not get wet in the rain."

6. Prefix-tuning

Prefix-tuning which is one of the ChatGPT prompts is a way of instructing the model to solve a specific type of problem by adding a specific prefix at the beginning of the prompt. This approach helps the model understand various types of problems and solve them more accurately.

For instance, you can create a prompt, "A problem type to be solved is a math problem, and the model needs to find the sum of two numbers." In this case, the model will recognize that the subject of the problem is math and will be instructed to find the sum of two numbers.

On top of that, assume that there is a prompt, "A problem type to be solved is translation, and this sentence needs to be translated into Spanish." In this case, the model recognizes that the problem is a translation task and is instructed to translate the sentence into Spanish.

As a result, prefix-tuning provides clear instructions to the model and improves its accuracy.

7. GPT-3's in-built Prompts

GPT-3's in-built prompts refer to pre-defined prompt functions that are built into the GPT-3 model. These functions are easy to use without the need for separate code or configuration to perform certain tasks.

For instance, "Translate" prompt is one of the in-built prompts in GPT-3. It provides the function that automatically translates an input sentence.

Another example is "Answer to the question" prompt which is utilized to generate an answer to the question such as what the area of a circle with a radius of 5 is.

Also, "Writing" prompt is used to automatically generate contents on a given topic. Say that the topic “romance” is given, the prompt easily writes on it.

As mentioned above, GPT-3's in-built prompts can be effectively utilized in various fields. By using this feature, natural language processing tasks is easily and quickly performed, and as a result, it can save time and efforts.

The types of in-built prompts for Chat GPT-3 are as follows:

• Answer Questions

• Chat

• Code Completion

• Compose Email

• Creative Writing

• Explain

• Generate Poetry

• Language Translation

• Summarize Text

• Task-oriented Q&A

*This article has been generated by training CHAT-GPT4.