ChatGPT or CheatGPT? The impact of ChatGPT on education

On November 30, OpenAI launched its AI chatbot called ChatGPT. ChatGPT is the most important revolution we have ever had on the Internet, much more important than Blockchains and NFTs. The capacities of ChatGPT are phenomenal. It can do many things, such as writing poems, writing and summarising articles, writing and debugging codes, and solving puzzles and mathematical questions. In addition, the ChatGPT tool can also be used with other tools such as DALL.E 2 (https://openai.com/dall-e-2/) to generate drawings or the Philosopher AI to answer philosophical questions (https://philosopherai.com/). Some tools can also allow the creation of music based on our lyrics. For example, we may use ChatGPT to generate some lyrics and give them to another AI tool to develop our song.

If we look at the impacts of these tools pessimistically, the development of AI tools would mean the end of human intelligence and the triumph of human stupidity when only influential individuals can access these tools. My short experience with these tools proves to me, however, that these tools are more beneficial than harmful. I'm going to present different use cases in various fields. I will also show other cases when these tools can be misused. This article will be updated regularly based on new findings and information I will acquire about this topic.

I will start by presenting different use cases for which ChatGPT is useful as well as its limitations. Then, based on these limitations, I will give various proposals on how to adapt teaching in universities and schools to incorporate these tools rather than excluding them. Finally, I will answer the question of whether AI will outweigh HI (Human Intelligence).

• Writing Poems

The first use case that I present is writing poems. I'm very impressed by the capacity of the tool to write poems. So, for example, I ask here the tool to write a poem to criticize the madness in the certificate of appreciation we see every day on LinkedIn.

But when I asked the tool to propose improvements to a poem I wrote, the tool returned some nice proposals, but some of the given suggestions were against the idea I wanted to give in my poem. Thus, someone has to be careful about the results returned by the tool, more than copying and pasting the results. Nevertheless, the tool may help a poet to get some ideas. A poet can read these suggestions that inspire him to develop other ideas. However, there will always be a debate about whether a poem is written entirely by AI, HI, or HI inspired by AI.

• Solving mathematical puzzles and equations 

We can ask ChatGPT to do various mathematical equations and explain different mathematical theories. For example, I asked the tool to explain Goldbach Conjecture to me. The tool gave me a good explanation and let me know that my counterexample is not valid.

However, the tool capacity is only good in explaining mathematical theorems but not in reasoning. For example, I asked here the tool to give me a prime number that starts with 2 and finishes by 9. The number should have 8 digits. The tool returned a number that claimed to be prime. The tool insisted on the correctness of the result, but when I gave it the factors of the returned answer, it acknowledged the error.

• Writing and debugging a code 

Here I asked the tool to write a python code to cryptanalysis a text encrypted using Vigenere Cipher. The tool returns the code with an example.

The tool can also be used to debug codes. It is very helpful feature that can help developers to identify quickly the errors in their codes.

• Writing essays and summarising articles

I asked the tool to write a teaching statement for me. I even asked it to write a more extended version because I was unsatisfied with the first returned version. 

• Giving opinions

I asked ChatGPT about its opinion regarding one of my ideas in this article. The tool confirms my analysis regarding Human Intelligence.

 

• Generating exams

I can give some texts to the ChatGPT and ask it to propose exam questions. Here I give it an article written in French. The tool returned me a bunch of possible exam questions.

• Ethical issues

ChatGPT has an ethical component that allows filtering out some requests. For example, here, the tool refuses to help me extract a secret key from the memory of Ransomware. Most likely, the tool misunderstood my request and interpreted it as a secret key extraction from the normal process. The management of the ethical part is complicated because what is allowed in one country might not be allowed in another. The future will tell us how OpenAI will manage this complex part. 

So the number of possible ways we can use this tool is unlimited. It is now important to see how this tool will change our society.

The impact of ChatGPT on Education

I want to handle now the question that many professors ask: should we allow students to use ChatGPT, or should we consider it a cheating tool? 

Before I answer this question, I want to describe this tool based on the use cases I presented earlier. If we accept to assimilate this tool into a wise man who always has an answer to every question we ask. The answer can be good, wrong, or inaccurate. The answer can be inaccurate because we need to ask the correct question or because the tool can not handle the question (e.e., reasoning about logical puzzles). 

So using the tool is not a simple click button, we need to build some skills to know how to use it. First, we need to ask the correct right question, and then we need to have the ability to know whether the returned answer is correct or not. Asking the right question is a skill, and we need, as teachers, to develop this skill with our students. For example, professors of programming languages have to adapt their teaching methods from asking specific questions (e.g., creating a login PHP page) to open questions (e.g., building a warehouse management application). They should evaluate students' capacity to ask good questions rather than their problem-solving ability. Georg Cantor once said, "In mathematics the art of asking questions is more valuable than solving problems. ". We need to do the same with AI.

The Math and Logic professors are not affected that much by the ChatGPT tool, due to its limitations in reasoning about mathematical equations and logic puzzles. However, all the teaching courses that require students to write articles or essays are largely affected by this tool.

We should allow our students to use ChatGPT. We can develop many pedagogical approaches to incorporate the tool into our teaching materials. For example, we ask our students to write an essay about some topic and then ask the tool to generate a report on the same topic and compare it with their work. This will help students understand their errors and improve themselves without the help of the professors. It can also help students see that not all the returned answers are profound. In addition, we should ask students to discuss some returned answers that look good but contain many errors. 

One week after the release of ChatGPT, I asked my students to use ChatGPT during their exams, which were designed to be an open book (I might be the first one in the world who did that but not sure about it). While ChatGPT helped students understand the context of the exam questions quickly, ChatGPT didn't help them obtain perfect grades because all the returned answers were general and less profound than the discussion we had in the class.

As professors, we need to understand that ChatGPT is a useful tool for students and themselves. It helps us optimize our time if we ask our students to test their questions with the tool before they send us their questions. I will ask my students next semester to use the tool to answer the questions I ask during my class sessions. I will let them read the returned answers and discuss them with me. In the exam, I will also allow them to use the tool, but I will expect more profound answers based on our class discussions. It will be easy for me to know whether the exam answers are returned by the tool or the students because I expect them to return answers based on the ideas we discuss in class. 

Finally, I decided to hire a research assistant to help me define a customized AI assistant that would be trained based on emails that I sent to my students. OpenAI provides complete documentation (https://openai.com/blog/customized-gpt-3/) that helps customize the GPT-3 model. The personal AI assistant will allow me to use my time more efficiently as I don't need to repeat the same answers to different students.

Can we tell whether professors jobs will disappear one day?

This question is valid only when we believe that machines will outweigh human thinking. We must understand that the machines are applying statistical-based processes to return answers to our queries. The term thinking doesn't exist with machines. Machines need to be fuelled by data generated by humans (i.e., use cases) while trying to solve their problems. The uncertainty of our world makes infinite the number of problems humans have to handle. So machines must wait for humans to perceive the problems, generate data, train the machines, and finally solve problems. So machines can only address issues humans perceive; they can never be conscious of our problems and what use cases we define to these problems. Therefore, HI (Human Intelligence) will consistently outperform AI. In addition, humans can always develop more profound thinking than machines. 

It is essential to differentiate Global Human Intelligence (GHI) from Individual Human Intelligence (IHI). The GHI represents the total sum of IHIs studied, proved, and analyzed in different fields. The GHI represents the ultimate knowledge that humanity has built till our present time. The IHI can be, however, partial and not a correct knowledge that many humans have in different fields based on their own experience. It is for sure that AI outweighs the IHI of every individual, but it is also very sure that AI will never outweigh the GHI. Thus, we need to see AI machines as knowledge improvement tool that helps every IHI to develop so that the GHI of humanity will always get better.

Our role as professors is to consistently check these tools so we can discuss more profound ideas and concepts in class. Students can then use these tools to understand the context of our questions and correlate the results of the tool with the concepts discussed in our classes. 

Conclusions

I believe ChatGPT is great tool universities and schools can use to improve education. It pushes professors and teachers to constantly improve and update their courses because they have to ensure that their teaching materials are always more profounds than AI chatbots. It also helps students to develop the capacity to analyze answers and ask good questions rather than focusing on technical details.

Before the advent of computers and the Internet, students had to cross multiple miles to reach their university libraries, spend hours finding the needed books, and other hours finding the required information. Before the release of ChatGPT, university libraries had lost much of their value because students tended to Google everything, and all information became practically accessible from their homes. However, students have had to spend quite a long time to come up with the right good question and find the correct answers. With ChatGPT, students' research time will be largely reduced because they now have a virtual teacher who can answer their questions whenever they want. So from a knowledge point of view, everything has stayed the same. It is only easier for students to find the correct information if they ask the right questions and can understand and analyse the answers.

I have always indicated to my students that understanding the WHY is more important than the HOW. Now we have the tool to let everyone find the WHY he wants. The deeper he goes, the better IHI he will have. If everyone makes the same effort, the GHI of humanity will be the real winner. This tool will help us focus on the concepts rather than the technicalities. Ultimately, it will be the reason for the renaissance of Philosophy and Math teaching, where abstraction is more important than the techniques and the HOW. It is improbable that AI machines would master the WHY more than humans because AI machines are not conscious entities. AI machines can perfectly tell us HOW to solve problems, but they don't know WHY humans want to solve these problems. 

So all depends on us as professors and teachers. Our role as professors is not only defined by what we teach; it is about helping students to develop their reasoning capacities using the available tools that we have in our current time.  The ChatGPT tool will be a source of cheating if we, as professors, refrain from developing new pedagogical approaches that incorporate AI chatbots in our teaching materials. Preventing students from using those tools will slow and hinder the GHI construction process that humanity needs to understand and solve the problems of our world.

I strongly believe that ChatGPT is a great tool that allows creators and cheaters to improve their performance, whether to create or cheat better.  So, it is up to every person to decide how to use ChatGPT, to create (by analysing and learning from its answers) or to cheat (by copy-pasting its answers without any analysis). 

Presentation Slides

https://docs.google.com/presentation/d/1FuBNN8FyaTX4z6NJpom3pHRLBfwyo0s56mBwed5DIp8/edit?usp=sharing