top of page
Writer's pictureBenedict Chia

Let’s talk about (and with) ChatGPT!

Happy new year everyone! Literally, sharing this post on the eve of Chinese New Year. I am pleased that my first blog post of 2023 is a collaborative effort with a former ONL222 coursemate, Jacqueline Hoppenreijs (@JHTHoppenreijs)!


2023 just started and the world is thoroughly captivated by ChatGPT as seen in its meteoric rise in popularity (within days of launch in late Nov 2022) leading to its current 29 billion dollar valuation. Crazy numbers. Fittingly, my first blog entry in 2023 will start things off with a chat about (and with) ChatGPT.


ChatGPT official logo
ChatGPT official logo

"ChatGPT logo" is licensed under CC BY-SA 4.0


ChatGPT finding its way into our daily routine


Slowly but surely, ChatGPT has invaded our daily lives, whether you like it or not. Some cool things it can do now include: write a twitter thread, write a novel, prep questions for guest speakers, create games, help with dating, deal with loneliness, name things, translate on the go, outline a fitness plan, write and debug code, create lesson plans, handle social media comments and reviews, crime fighting, check for bias in writing, design furniture, write novel, prep for interview, gift ideas, explain complex concepts, editing for APA style, etc. (Schaefer, 2022).



“What would you like ChatGPT to do for you, as a learner, educator, leader, friend, family member? What would you NOT want it to do?”

Dr Kay Oddone already gifted the world, and ONL enthusiasts specifically, with a great post on how #ChatGPT can impact educators and Teacher librarians. It has several great leads and is definitely a good starting point if you are keen to deep dive further. If you are hard pressed for time, go straight to the bottom of her post to check out her curated list of resources relating to how chatbots, chatGPT and AI are used in teaching and learning.


If you are somewhat new to ChatGPT (hey, all of us have to start somewhere right?), and wish to learn about ChatGPT in a light hearted approach, read on!



*What is ChatGPT?


So, what is ChatGPT, when did it really take off, who is impacted, why it's such a big deal and how can we (educators) leverage it to improve student learning experience?


ChatGPT, also known as the Generative Pre-trained Transformer, is a large language model developed by OpenAI. It uses deep learning techniques to generate text that is similar to human-written text. It was first released in 2019 and has quickly become one of the most powerful and widely used language models.

The impact of ChatGPT is wide-reaching, as it has the potential to improve many natural language processing tasks such as text generation, language translation, and question answering. It has been used in a variety of applications, including chatbots, virtual assistants, and content generation.

The significance of ChatGPT lies in its ability to generate text that is difficult to distinguish from text written by humans. This has led to concerns about its potential use in generating fake or misleading information, but it also opens up many possibilities for new and more sophisticated applications.

For educators, ChatGPT could be leveraged to improve student learning experience in several ways such as creating personalized and engaging learning content, facilitating formative assessments, providing real-time feedback and support, and helping students with language difficulties. It could be used to generate practice problems and quizzes, or to create customized study guides for students. Additionally, it could be used to generate summaries of articles or lectures for students who have difficulty with reading comprehension.

It is important to note that although the technology has huge potential to improve learning experience, it is also important for educators to be aware of its limitations and potential biases in order to use it responsibly and effectively.



ChatGPT vs Google


Let’s address the elephant in the room. Will ChatGPT replace Google? It’s easy to think of the similarities: ChatGPT can help you find recipes, translations and answers to all sorts of questions that you might usually ask Google. However, instead of a slew of 50 websites, ChatGPT directly gives you an answer in a very human-like response. ChatGPT 1: Google 0


Should you want to know by whom, or when that information was published, you may not get an accurate answer from ChatGPT. This links directly to another limitation, i.e. ChatGPT trained itself using (relatively) old, and offline information. At the point of this writing, ChatGPT’s knowledge is limited to events up to May 2021. So, if you are looking for latest information relating to vaccines, anti-COVID measures, travel restrictions, chess cheating saga, the 2022 world cup, or key events that happened in 2022, you are much better off relying on good old Google. Score tied at one a piece.


Google is also biased in that it has its own, not-super-transparent, ways of deciding on the relevance of a certain result, and of course it makes money from people and firms that search or want to be found through it. It’s not said that that won’t be the case for ChatGPT as it has been hinted that it will not remain free forever. For a fuller analysis into this mano a mano face off, check out this comparison article.



*ChatGPT vs ONL!


Jac and I attempted to have some fun with ChatGPT by capturing our reactions while discussing how to go about our little experiment. They say time flies when you are having fun and indeed, the process of consolidating our ideas and throwing it back and forth went by so fast, we overran the meeting by more than an hour!

Cost of using Google doc, zoom, chatGPT = free. The fun we had... priceless.
Cost of using Google doc, zoom, chatGPT = free. The fun we had... priceless.

Check out our reaction video. Warning, ONL administrators with a weak heart, please proceed with caution. https://youtu.be/xjaXvqxwiDY

Time Stamps also under the description of the YouTube video.


TIME STAMPS

0:00 Introduction (mostly chitchat)

02:17 Why we are doing this experiment

03:34 Tweaking the questions for ChatGPT

19:37 Setting the context (as an educator)

20:58 Context on how we are going to do this experiment

21:30 Topic 1, Digital Literacies

24:33 What do I do if my students want to add me as a facebook friend?

26:52 Topic 2, Openness in Education

34:06 Topic 3, Collaboration & Group Work

49:48 Topic 4, Emergency Remote Teaching and Learning

56:16 Cliff Hanger!



*ChatGPT bane or boon?

ChatGPT, like any technology, can be both a boon and a bane depending on how it is used.

As a boon, ChatGPT can be used for a wide range of natural language processing tasks, such as text generation, language translation, and question answering. It can also be used in chatbots and virtual assistants to improve the user experience and automate certain tasks.

On the other hand, ChatGPT can also be a bane if it's not used responsibly. For example, if the model is trained on biased data, it can perpetuate and amplify that bias. Additionally, if the model is used to generate fake or misleading information, it can be a tool for disinformation.

It's important to use caution and consider the potential consequences of using this technology, and put in place appropriate safeguards to mitigate any negative effects.



Is ChatGPT a panacea?


As powerful as ChatGPT is shown to be, it still has limitations. Yes, it hallucinates and makes up stuff when it does not know the correct answer. Problem is that it fakes this so well that it's difficult to catch by an untrained eye. An AI text agent can make an argument from a personal blog sound like a well-written argument published in an academic journal. Only an expert of the context would be able to see that the argument is of low quality and that the citations are wrong (Illia et al., 2022).


While this article focuses on Ethical implications of AI text agents, it also highlights several current limitations like “The lowest denominator problem” which basically means that AI text agents (when training) treat all large data sets as equally important. It cannot identify the quality of articles between a New York Times presenting an anti-vax argument with scientific evidence vs an anonymous blog post giving the same argument but with non-scientific and biased statements. Basically, AI training prioritises quantity over quality.


Other ChatGPT failures are captured in all its glory via this Github archive. Technically, it's limited to what it has been taught and how it's been taught. Hence, it's incapable of original thought / creation... for now.



*ChatGPT has different personalities & a sense of humour

Due to its popularity, you may find that ChatGPT occasionally posts the downtime “error” message.


Is ChatGPT showing off his sense of humour? Alright, challenge accepted!

Thank you ChatGPT... bigly.
Thank you ChatGPT... bigly.


Trivia time!


Q1: Which of the paragraphs above were written by ChatGPT?

A1: Those responses marked with asterisks were all generated entirely by ChatGPT! Gotcha!!


I can see you shaking your head in disbelief. Why not just try it out yourselves?

  1. Go to URL https://chat.openai.com/ and create an account.

  2. Pray that it’s not down due to heavy traffic and that it's still free to use.

  3. Type in your query and let us know your reactions!


Q2: Is education destroyed now that chatGPT is here? Is chatGPT a friend or foe? Do you see this as an opportunity or disaster to higher ed? What other pitfalls can you think of?

A2: Jac & I don’t have an answer to the questions above. But we would love to have you join us and figure it out together as we dive deeper into this seemingly bottomless rabbit hole.



Conclusion


This has been a fun post to create. Linking up with a former ONLer and putting into practice a little from last year’s ONL222 teachings on open networked learning is a great way to resume our ONL journey.


If any of you are keen to also co-write / co-experiment something, drop us a message / email / Tweet or leave a comment below. #ONL4ever!


@chiabinghowe | bhchia@nus.edu.sg

21 Jan 2023


References

82 views0 comments

Recent Posts

See All

Comments


bottom of page