By now, most of us have heard of ChatGPT. Some of us may be hesitant to use it, while others are testing its boundaries and seeing how far it can go. In collaboration with Erasmus Tech Community and Turing Students, S4S organized a panel discussion and event centered around ChatGPT and its role in education. They tackled this topic with professionals from different perspectives giving their opinion.
The cold and rain didn’t stop a packed Sanders room full of students showing up to GPTalk. This event promised a wide range of panelists presenting and discussing ChatGPT, and the role of generative AI in education, for both students and staff. The panel included Eduard van Valkenburg, a senior cloud solution architect at Microsoft, Kim van Broekhoven, assistant professor at ESSB, Bas Smit, Strategic Information Manager on Education at the Erasmus Digitization & Information Services (EDIS), and Willem van der Maden, Ph.D. candidate at TU Delft who is researching designing well-being aligned AI.
What is artificial intelligence?
To start the event, Eduard van Valkenberg introduced the topic of ChatGPT and discussed what AI is, as well as its uses and limitations from a professional perspective. Considering the massive investment that Microsoft has just made into ChatGPT, Eduard was able to bring some very insightful information to the table. Eduard discussed what ChatGPT is by starting with a small history lesson on artificial intelligence, otherwise known as AI, and how ChatGPT is a form of generative AI, which can produce new output- as opposed to other types of AI. He explained how this new output is the product of the AI being trained with trillions of texts.
What makes ChatGPT different?
What it can do from this training, Eduard highlighted, is predict the next most likely word to appear, based on what your input is. That is all generative AI like ChatGPT is- predicting what is most likely to be said based on what has already been said and the AI has learnt from. This gives ChatGPT the enormous power it has- the more it is used, the more it learns and is able to use more information to come up with its answers. However, the main limitation that Eduard discussed was that generative AI such as ChatGPT lack interpretability and explainability. This is because the AI cannot describe the steps it took to reach the answer that it has given- and engineers that develop AI such as these also cannot precisely say the reasoning of the AI in the steps it took to generate its answer.
"AI can help with the product, but not the process, in academia"
ChatGPT and education: friend or foe?
Following Eduard, Kim van Broekhoven presented the educational implications of ChatGPT and generative AI in an academic context. Kim highlighted the fact that, although it gives the impression that it understands as a human could understand, ChatGPT and other AI can only replicate and generate output- but not understand it as a human would. Kim mentioned that AI provides the opportunity of moving away from assembly-line style working, as students can use it as a tool to expand their capabilities and solve wicked problems. She also sees opportunities for students to learn using ChatGPT, such as learning by teaching- using ChatGPT to generate output that then students have to critically analyze and assess how correct it is. However, she also sees a lot of limitations and potential problems that come with students using and relying on ChatGPT, such as damage to academic writing skills. Overall, Kim concluded with a point to sum up an academic perspective on ChatGPT: AI can help with the product, but not the process, in academia.
Student perspectives on ChatGPT
The event then concluded with a panel discussion, where the students and the panelists discussed their views and questions about ChatGPT. The discussion focused a lot on whether the use of ChatGPT is something that can be prevented, or is it now part of our lives and here to stay. The consensus from both the professional and academic side is that no matter how hard someone works to gatekeep its use, it will keep developing to be able to overtake software that is supposed to prevent it. Students also asked if to get good results, ChatGPT is needed. Kim mentioned a research paper in which students who did use ChatGPT on average scored better, however for high performing students ChatGPT didn’t enhance their performance by much- but it did for low performing students. This shows that ChatGPT has the potential to narrow down the range of scores in academic performance.
All in all, the event was a successful and insightful look into ChatGPT and generative AI in education. While no one has a clear answer about what ChatGPT should and shouldn’t be used for, everyone has an opinion on it.
Read the questions the students asked through Mentimeter.
- More information
This event was organised by S4S, Erasmus Tech Community and Turing Students.
- Related links
- AI tools in education