The AI industry is evolving rapidly, with major players such as OpenAI, Microsoft, Notion and Google constantly striving to maintain their competitive edge by launching products like Brad, Copilot and Notion AI. Among these, OpenAI’s ChatGPT has been a pivotal product that has sparked the widespread adoption of AI technology.

For those who are new to ChatGPT, there may be a knowledge gap of approximately five months since its release in November 2022. This gap may limit their ability to fully utilize the capabilities of an AI chatbot like ChatGPT. However, given the ongoing developments in AI technology and the popularity of ChatGPT, there are ample resources and opportunities available to help new users quickly catch up and make the most of this innovative tool. One of the main talking points has been the recent introduction of the term “Prompt Engineering”.

## Prompt Engineering

Prompt engineering is a processing concept within natural language (NLP) used to train models that focuses on identifying inputs that lead to desirable and useful outcomes. It involves developing prompts that are carefully crafted to guide the AI towards producing correct answers. By leveraging prompt engineering techniques, researchers and developers can create more intuitive and effective AI systems that can better understand and respond to natural language input.

Since the introduction of ChatGPT not only is it important for researchers and developers to train these models but also for its users. Since ChatGPT is a prompt-based chatbot its functionality is highly correlated with the input the users provide it. This means that there is a lot to gain from actually knowing how to ask the right questions to the AI when desiring a particular result.

## How should I formulate my questions?

The following section will summarize prompts for different uses relating to the optimization of studying and grasping difficult concepts in our fields of study such as Mathematics, Econometrics, Operations Research, Actuarial Studies and Programming.

Firstly, let us start with the basics, one useful thing to know is that Large Language Models (LLM) have been predominantly trained with an English dataset. This naturally means that one can (such as doing a Google search) better prompt the chatbot with English sentences when trying to look for a particular answer. One of the best things to do while using ChatGPT is to make use of its multiple chat windows. Each window can be used for a specific goal or topic. When starting up a chat, it is best to “prepare” the AI for what you will want to achieve. For example, when needing help to create plots for a project in R you could start by saying the following: *“I would like for you to act as an AI that helps with the creation of plots in R using the ggplot2 library, I will be providing you with R code and questions regarding the improvement of plots and would like a detailed answer.” *This process sets up the chatbot to what you would like to receive as an output but also lets it know what you could be inputting. When following this procedure, you will generate multiple chat windows which will specialize in different topics!

Next, it is important to know the limitations of ChatGPT. The chatbot based on GPT-3.5 has an input limit of around 8,000 words whereas GPT-4 has an input limit of 25,000 words. Both models also have a limit on words they can output, you might have already noticed it when the chatbot just randomly stops answering your question. Furthermore, there are some other issues within ChatGPT which mainly arise due to the nuance of the prompt asked, it might for example come up with some wrong computations or wrong answers in general, that is why for now, it is important to double-check its responses. You can best do this by asking the model itself to cite its sources.

With all that in mind let us start by showcasing prompts to help your day-to-day study sessions.

### Ask ChatGPT to explain something as if I were 5 years old.

If you are starting to understand a subject it is sometimes good to take a few steps back and get a really simple explanation of the important concepts. The following example could maybe illustrate the power of this a bit better:

*Me: *

*Please explain the OLS, Logit, GOLS models as if I were 5 years old. Please go into detail regarding all models.*

*ChatGPT: *

*Okay, let me try to explain OLS, Logit, and GOLS models in a simple way.*

*OLS is like drawing a line to show how two things are related. For example, if we want to know how much taller people are based on how much they weigh, we can draw a line to show that as people weigh more, they tend to be taller too.*

*Logit is like playing a game of “yes or no” questions. We use it when we want to know if something is one way or another way. For example, we can use it to figure out if a person is a boy or a girl based on their height, weight, and other things.*

*GOLS is like using a magic trick to make the line we draw in OLS even better. Sometimes the line we draw with OLS is not quite right, but GOLS can help fix it. It’s like using a special tool to make the line fit the data better, so we can understand how things are related even more accurately.*

For most of you, this might seem a quite simplistic answer, but that is because you already understand these concepts thoroughly. Imagine using this for more complex subjects which you know nothing of.

### Asking ChatGPT for step-by-step instructions for solving a problem.

Directing ChatGPT to a specific problem and telling it to provide you with step-by-step instructions is a powerful tool. The following example could be of help if you are starting with mathematical proofs.

*Me: *

*I have the following problem: *

*“Prove that the function f(x) = x^2+ (x+3)^3 is continuous using the epsilon-delta definition”. Please give me step-by-step instructions on how to solve this problem.*

*ChatGPT: *

*Sure, I’d be happy to help you prove that the function f(x) = x^2+ (x+3)^3 is continuous using the epsilon-delta definition.*

*To prove that a function is continuous using the epsilon-delta definition, we need to show that for any epsilon (ε) greater than 0, there exists a delta (δ) greater than 0 such that |f(x) – f(c)| < ε whenever |x – c| < δ. Here, c is a fixed point in the domain of the function.*

*Let’s begin:*

*Step 1: Choose a value for c*

*Let’s choose c = 2. This is an arbitrary value, but it will help us in our proof.*

*Step 2: Write the expression for |f(x) – f(c)|*

*|f(x) – f(c)| = |x^2+ (x+3)^3 – 35|*

*…*

I hope you get the gist of it. It is super powerful to redirect the chatbot to work in a structured way.

### Ask it to create flashcards or practice questions.

Now let us imagine you are studying for your exam but you already went through all your practice questions and want more. It is possible to let the chatbot create practice questions for you. Moreover, flashcards to study concepts are also something it can generate.

*Me: *

*Could you please create flashcards and practice questions regarding continuous functions and their proofs? Please add answers below*

*ChatGPT: *

*Sure, I’d be happy to create some flashcards and practice questions for you on continuous functions and their proofs. Here they are:*

*Flashcards:*

*Front: Definition of a Continuous Function*

*Back: A function f(x) is continuous at a point c in its domain if the limit of f(x) as x approaches c exists and is equal to f(c).*

*Front: Epsilon-Delta Definition of Continuity*

*Back: A function f(x) is continuous at a point c in its domain if for every ε > 0, there exists a δ > 0 such that |f(x) – f(c)| < ε whenever |x – c| < δ.*

*Front: Proving Continuity using the Epsilon-Delta Definition*

*Back: To prove that a function f(x) is continuous at a point c, we need to show that for every ε > 0, there exists a δ > 0 such that |f(x) – f(c)| < ε whenever |x – c| < δ.*

*Front: The Squeeze Theorem*

*Back: If f(x) ≤ g(x) ≤ h(x) for all x in some interval containing c, and if lim x→c f(x) = lim x→c h(x) = L, then lim x→c g(x) = L.*

*Practice Questions:*

*Prove that the function f(x) = 2x – 1 is continuous for all x.*

*Answer: Let c be any point in the domain of f(x). To prove that f(x) is continuous at c, we need to show that for every ε > 0, there exists a δ > 0 such that |f(x) – f(c)| < ε whenever |x – c| < δ.*

*Let ε > 0 be given. Then, we can choose δ = ε/2 so that if |x – c| < δ, then*

*|f(x) – f(c)| = |2x – 1 – (2c – 1)| = 2|x – c| < 2(ε/2) = ε.*

*Therefore, f(x) is continuous for all x.*

*…*

### Create mind maps and mnemonics.

One great example of how to use the chatbot’s full potential is by asking it to create mnemonics (ezelsbruggetjes) for things you must remember. This is a great way to remember things, here is an example that might be helpful for your future endeavours.

*Me: *

*Please create a mnemonic which will help me remember the assumptions made to be able to apply the OLS model.*

*ChatGPT:*

*Sure, here’s a mnemonic to help you remember the assumptions made to apply the OLS model:*

*“The FIVE N assumptions for OLS”*

*F – Functional form of the model must be linear in the parameters.*

*I – Independence: the observations must be independent of each other.*

*V – Variance: the variance of the errors must be constant across all values of the independent variable.*

*E – Expected value: the expected value of the errors must be zero.*

*N – Normality: the errors must be normally distributed.*

You should of course always double-check if these mnemonics are correct. The great thing about this is that you get a quick way of remembering important concepts easily.

### Ask ChatGPT to act like a specific profession.

*Me: *

*Act as if you were a professional English grammar checker*

I input the fourth paragraph of this article:

*Me: *

*Since the introduction of ChatGPT not only is it important for researchers and developers to train these models but also for its users. Since ChatGPT is a prompt-based chatbot its functionality is highly correlated with the input the users provide it. This means that there is a lot to gain from actually knowing how to ask the right questions to the AI when desiring a particular result. *

*ChatGPT: *

*The sentence is well-written and grammatically correct. However, there are some minor suggestions to enhance clarity and flow. Here’s a possible revision:*

*“Since the introduction of ChatGPT, it has become crucial not only for researchers and developers to train these models but also for users to learn how to interact effectively with them. As a prompt-based chatbot, ChatGPT’s performance is closely tied to the quality of user input. Therefore, understanding how to ask the right questions is essential in obtaining the desired results.”*

For sake of argument, I will not be changing the above paragraph, but now you can see how the usefulness of the chatbot extends to things you first did not think about.

Another great example for those starting/in process of doing interviews.

*Me: *

*Act as a Data Science position interviewer, provide me with questions I could expect*

*ChatGPT: *

*Sure! Here are some sample interview questions you could expect for a Data Science position:*

1. *What motivated you to pursue a career in data science, and how did you get started in the field?*

2. *Can you walk me through the typical steps you would take when tackling a new data problem, from data cleaning to modeling?*

3. *Describe a complex data analysis project you completed recently, including the tools and techniques you used and the results you obtained.*

4. *How do you handle missing data in a dataset, and what are some of the techniques you can use to impute missing values?*

*5. Tell me about a time when you had to work with a large and complex dataset. How did you ensure data quality and manage the data effectively?*

6. *What is your experience with machine learning, and how do you choose the best model for a given problem?*

7. *Explain the concept of overfitting, and what steps do you take to prevent it in a machine learning model?*

### You can ask ChatGPT to provide the best prompts for a specific topic, including understanding more about ChatGPT itself.

*Me: *

*What are great prompts I could ask you if I would like to understand Extreme Value Theory better?*

*ChatGPT: *

1. *What is Extreme Value Theory and how does it differ from classical statistics?*

2. *What are the key assumptions of Extreme Value Theory and how do they impact the analysis of extreme events?*

3. *How is Extreme Value Theory used in the modeling of rare events such as floods, earthquakes, and financial crashes?*

4. *Can you explain the concepts of block maxima and peaks-over-threshold methods in Extreme Value Theory?*

5. *How does the choice of distribution impact the estimation of extreme events in Extreme Value Theory?*

*…*

I hope the above examples gave you a better clue as to what you can achieve with ChatGPT.

The introduction of the internet was a significant milestone, and while some individuals and companies failed to adapt, others such as Amazon, Microsoft, and Google transformed themselves by leveraging its power. Similarly, as we embrace cutting-edge technologies like ChatGPT, it becomes increasingly essential for individuals to adopt them promptly. Failure to do so risks being left behind and seen as outdated by previous generations.