ChatGPT and education: some notes and reflections
(IMPORTANT NOTE: this is a translated version of the original article published in Spanish.)

ChatGPT and education: some notes and reflections

(IMPORTANT NOTE: this is a translated version of the original article published in Spanish).

ChatGPT (short for Generative Pre-trained Transformer) has become one of the year's big stars.

News, articles, and innovations are flooding social media, newspapers, and blogs. That's why, instead of making a technical description of the technology supporting it, I want to share my personal experience with it and how I see education being affected. For deeper information and tech background, watch this video or read this excellent summary of potentialities.

So, I invite you to:

  • Know the basics of ChatGPT, and see some real examples,
  • Reflect on the dimension and power of this technology,
  • Review my experience as a user,
  • Reflect together on the uses that we can give in the educational field


Hello, how can I help you?

ChatGPT is an advanced chatbot or conversational assistant that allows us to interact as if we were talking to a human being, answering different problems or queries.?

Tip #1:?If you have yet to experience using ChatGPT personally, I recommend that you do. You can do it from a computer or mobile device. And it's as simple as visiting https://chat.openai.com/, registering an account, and starting to "talk" with ChatGPT.?Later, please don't say I didn't warn you.

However, I know it's normal to have little time to experiment, so I share four brief video samples of ChatGPT in action below.

(sorry, they are in Spanish, from the original version of this article)

1) Modifying the style and recipients of an invitation text

2) Creating functions in Excel and code for us

3) Analyzing a story and explaining the moral

4) Creating a math exercise guide, in seconds


Tip #2: As I mentioned before, personally testing ChatGPT is recommended and the only way to perceive its true power. As it is still a testing environment, please be careful with the information you share and expose.


We are facing significant change

Since I experienced the internet for the first time at 18 years old, I don't remember anything that caused the same sensation when I first used it. I can highlight two fundamental feelings:

  • Firstly, a degree of amazement as "ah... this is a total and huge change".
  • And then, a sense of immeasurability. The internet was so big that I needed help understanding where to start.

I still remember the person who introduced me to the internet telling me: "Ok, we are connected now. Now search for whatever you want". It was a proposal too broad for me to process. It took me time to understand how the world was changing, how to speak its language, and the most challenging thing: how to tell the world what this was going to do.

We are aware at this point (and if we are not, it would be good to start being so) that today our interactions with different forms of AI are frequent and systematic in our daily activity. They can be obvious, like Siri or Alexa, and less evident in web search engines like Google. But increasingly, they are also more frequent in a hidden way on platforms like Netflix, Gmail, social media, and almost any application we use today in the cloud. The often so-called "algorithm."

For those of us who have been closely following developments in AI in its broad field, we know that advances in conversational AI systems are taking rapid and steady steps.

But unlike other cases, ChatGPT has made a significant turn: making an artificial intelligence tool available to all with an easy-to-use interface, as if talking to another expert human being.

And ChatGPT gives us results just by asking it simple questions like:

  • I need the Python code that allows me to access database X
  • I was hoping you could write a text about topic Y, oriented to young technologists.
  • I need to find a pattern in this dataset that I list below. Can you help me?


What ChatGPT is and is not

ChatGPT is an advanced language model that has been trained on a large amount of text and can answer questions, generate text, solve problems, and have natural conversations with people.

The concept of training is perhaps the most difficult to interpret if we are not familiar with the subject. The words?training,?comprehension, and even?intelligence?do not precisely refer to the usual meaning.

But for now, as a starting point, you need to understand that:

  • The model is trained through an intentional and supervised process.
  • The responses it gives are based on that training, as it has no innate knowledge or skills. Its ability to understand and perform tasks depends on the information it has been trained on.

To better understand the functioning and dynamics behind this technology, I recommend reading "ChatGPT and The Professional's Guide to Using AI" by Allie K. Miller, an exciting introduction to the subject.

Tip #3: distinguish this kind of tool from a search engine like Google. The latter is a search service that collects existing information in the (almost) infinite and disordered sea of the internet and offers users an index of search results in an organized manner. It also uses (and increasingly more) AI to order the data according to the user's context and other variables.

In fact, for the moment, ChatGPT does not access the internet to give its answers and is trained chiefly with information before 2021."


My experience as a user

I am curious about these topics, so I have been interacting and testing with ChatGPT for several weeks. I want to avoid staying with what others relate to.

In addition to the short and simple examples at the beginning of this article, I am systematically testing ChatGPT: I have asked it to demonstrate theorems, write poems in the style of an author, give me suggestions for writing a letter to a customer, write code in Logo and Basic (the first programming languages I learned), analyze data, develop a WordPress plugin, and a wide variety of things that have occurred to me. Always with mixed but surprising results.

Interestingly, the more you interact with the tool, the better you learn to communicate what you need, thus obtaining better results.

Here are other examples:

  • Some obvious ideas include writing thank you, greeting, or first contact emails.
  • Defining a sales and follow-up plan via email, and the associated texts.
  • Generating a list of diverse ideas for a corporate event.
  • Performing data analysis and patterns to propose conclusions in a research.
  • In the legal field, I have been told about tests for producing writing that is currently developed by junior profiles and then adapted by more advanced profiles.
  • In programming, ChatGPT is increasingly reliable in code development (although it still with many errors, but increasingly reliable).
  • ChatGPT can develop simple scripts or macros to be used in office applications, even for those who are not programmers.

I assure you that when you start using it, it is challenging to stop thinking about tangible uses.


Can technology like ChatGPT bring about change in education?

It is the first question that comes to mind as someone who works in this field. And it's in this direction that I've been conducting my testing.

Initially, public reactions were disbelief and fear, especially regarding the possibility of students using ChatGPT for academic cheating. There are a lot of articles on this topic, such as "AI bot ChatGPT stuns academics with essay-writing skills and usability" and others with headlines as strong as "The College Essay is dead", which also spawned countless Twitter threads, like this one. Proposals have even been made for "How to spot AI-generated text."

In my tests, I have found that ChatGPT can understand the logic of different types of exams and give accurate answers. Over the past few weeks, with the collaboration of enthusiastic teachers, we conducted several experiments on this topic using different formats and proposals, with more than surprising results. Simply by copying and pasting whole "papers" from different types of exams, ChatGPT was able to produce accurate answers. And the most notable thing is that these answers were always distinct (they are particular elaborations in each case).

One of the issues that will force us to review urgently is the imperfect way in which we assess learning (and why we are doing it). Discussing this is good news.

However, while it is true that this situation can generate headlines, I believe we cannot stop there. There is a new world ahead with a power that we can still not grasp.

In general, most people tend to think of artificial intelligence narrowly in relation to ourselves and the risks we face: Will it take my job away? Is it better or worse than me at such a skill or task?

And this limits us in understanding how AI is evolving and what that might mean for our future. I recommend reading Kevin Roose's article "We Need to Talk About How Good AI Is Getting" in the New York Times.

Tip #4: I am again calling to avoid the paralyzing syndrome of the robot invasion. This frequently happens whenever a genuinely disruptive technology appears in education. Getting out of this dichotomous paradigm between humans and dehumanizing technology is essential. I invite you from now on to open your minds and let ideas fly (always with a critical spirit, of course).


Some specific proposals for education

As I mentioned a few paragraphs above, I have been testing ChatGPT for various activities for several weeks. With the collaboration of some friends who are educators and interested in learning, I am promoting and participating in different tests and uses applied to teaching and learning.

Of course, these are still minor, initial experiences and not scientific research, but I hope they can be the starting point for many ideas and proposals.

And I repeat something regarding collaboration: as the days go by, the improvements in the results are substantial because we learn to use the tool better and better, and by sharing the cases with other users, new and better ideas arise.

For now, I have classified my experiences with ChatGPT related to education into three broad areas: content generation, assessment and feedback, and communication support.

Here they are:

Content generation

  • Creation of math exercise guides on particular topics. We have been able to create guides that take hours in seconds.
  • Explanations of a topic, with different approaches or adaptations according to the learner's profile: for example, to explain the concept of Uniformly Varied Motion, we ask ChatGPT for different explanations according to particular learning styles, ages, or prior knowledge.
  • Examples of sentences that include a characteristic: with diphthongs, or with an implicit subject, or even with more complex resources such as dialogues or anaphoras. We have asked for several examples on a particular topic.
  • Explanations of topics oriented to special needs, or simplified with certain characteristics: we are exploring how to transform a text to orient it to certain learning characteristics that require different formats.
  • Creation of Escape Room-type activities, with the definition of stations, difficulty and recipients. It speeds up the process from hours to seconds, we have been able to create creative scenarios and never before thought of in seconds. It can also adapt existing ones.
  • Explaining concepts in different ways: based on different ways of understanding, a concept can be re-explained, emphasizing a particular aspect, or longer or shorter length, etc.
  • Summarizing or simplifying: summarizing texts, simplifying them, or even identifying keywords or main ideas.

Assessment and feedback:

  • We are testing the evaluation of written productions against a rubric or text model and asking for suggestions and feedback.
  • Correction of exam-style question/answer. Comparing with a base model, obtaining the first feedback based on different parameters (length, word type, complexity, etc.)
  • Creation of games or riddles based on math equations, which allow to approach the application of concepts to real life.
  • Analysis of patterns and evolution: according to two or more productions over time by a student (it can be a text, composition, etc.), analyzing the variations and providing an evolutionary feedback. This can be done freely, or with certain control points or expected rubric.

Communication support:

  • Finding the right words to communicate a case, a situation, or a suggestion is often complex and time-consuming. ChatGPT can help share feedback more accurately.

Tip #5:?ChatGPT is a technology under development and experimentation and is by no means a resource to base teaching and learning processes without supervision or verification of results (not yet...).


Final thoughts

To conclude, I'd like to share some personal notes that I made after my experiences with ChatGPT:

  • It's understandable the enthusiasm and reactions that it causes in different sectors. We must keep sight of the fact that it is still a testing and development technology, but it is advancing quickly. We must keep sight of it.
  • ChatGPT is not the only one of its kind. Still, it is the first available openly, freely, and disseminated to the common subject. Others will appear soon with similar models. We should see it as one of many solutions, not even the best. But it enables us to start discussing the topic and to understand a logic very different from what we are used to.
  • Opinions and debates. Like any disruptive applied technology, it generates various discussions. Before expressing an opinion or debating, try it out for yourself. Don't rely solely on articles, networks, or newspapers. Even if it is minimal, experiencing the use will help create greater empathy and awareness about what is being talked about. It is still free and easy.
  • It is still a costly technology. It is available for testing because it pays for itself with the millions of tests that are done every day. But these models will probably have a subscription or monetization model that has yet to be defined. These variables should be evaluated for severe application strategies.
  • Should I stop working because AI will do everything? No, at least not now and soon. But it is a topic to pay close attention to, as it will improve significantly quickly, and it is definitely a game changer. Something is going to happen.
  • I believe the correct focus today is to consider it an enhancer or amplifier of human capabilities, not a replacement for them.
  • It definitely puts us in a place that challenges what we know, even the ethics of using and applying technologies. In this regard, I think it's essential to simultaneously develop teams, organizations, and groups that help with guidelines, guidelines, and horizons. We must be careful and review these steps.
  • Would students use it to produce effortless work? Definitely, well, with a different effort because they will have to learn to use these tools. But more importantly, it is challenging for the educational system to find the most creative ways to evaluate and, even better, to rethink the assessment process.
  • Sharing experiences is the best way to learn more about something so new and unknown, so I invite you to share them here or through networks and let's continue navigating these new opportunities.

And finally, I confess that in my case I did it... I couldn't resist using ChatGPT to correct, search, and write some paragraphs of this article.

Francisco


About Francisco Lehmann :

For more than twenty years worked and helped organizations and ventures lining learning, UX and IT, to create new approaches and learning environments.

He participates, leads, and advises on projects that involve Virtuality, Artificial Intelligence, Machine Learning, Blockchain, and UX. All realities are rapidly transforming the ways of learning, and its scope and drastically modifying society, employment, and the economy.?

He is currently the Deputy Overall Director at Belgrano Day School, Buenos Aires, Argentina, managing an ambitious educational innovation project for this renowned K12 institution.?


Sources, references and articles of interest:

It is fantastic to see educators going beyond the knee-jerk fear of plagiarism and discussing some of the potential benefits of using AI in the classroom.

Dr. Karina J. Baum

Director of Global Education Buckingham Browne & Nichols School

1 年

Gracias por escribir acerca de esto y gracias por los tips, Francisco Lehmann!

要查看或添加评论,请登录

社区洞察

其他会员也浏览了