It has been a bit less than a year since I wrote a blog post on how generative AI can be used in higher education, with a focus on applications for teaching, research, and admin. I was a relatively early adopter of ChatGPT and have had my subscription to the service for quite a while. As I am someone who likes tinkering with New Shiny Toys, I tried all the applications in the beginning. Now, more than two years later, and in the light of my daily work practice (in which classes need to be taught and papers need to be written, and not all my time can go into diddling around with new shiny toys), I have adopted generative AI for some of my day-to-day activities, but also understand the limitations of the tool. I, for one, am not a fan of all AI-generated blog posts or AI-generated search summaries. I find them just boring and bland and lacking sparkle and personality.
Recent changes to the AI landscape
When I planned to write this post, the seas seemed relatively calm – no major upheaval at OpenAI, no major scandals in the AI world, just a steady stream of bloggers and internet famous people trying to give advice on how to use generative AI best. And then DeepSeek showed up: a product of a Chinese startup that developed a cost-effective AI model that manages to replicate the performance of ChatGPT – and they made it for way less money: $6 million using approximately 2,000 GPUs.
The arrival of DeepSeek not only sent shockwaves through the AI world, but also the stock market. Nvidia’s stock price took a dive as investors feared that more efficient AI models would reduce the demand for Nvidia’s high-end chips. At the time of writing, OpenAI has already launched their answer to DeepSeek with o-3 mini, a new model. So, in the midst of all these ongoing changes, I am trying to wrap my head around the current situation without getting too distracted.
Ways I implement GenAI in my academic activities
I use generative AI on a nearly daily basis. My usage has shifted over the past months, and here is how I currently use ChatGTP:
- Part of my teaching: All my students take the Prompt Engineering and Trustworthy Generative AI courses as part of my class this semester, and in their writing assignments, I ask them to share with me the prompts they used and how they fact-checked the information in their report.
- Interpret call documents: For complicated calls for proposals, I use ChatGPT to read the documents and help me understand if a certain cost (for example) can be expensed or not on the project.
- Check compliance with call documents: While I prefer doing my own writing (as, for me, writing is a way to organize my thoughts), I use ChatGPT to verify if I have answered all the questions and aspects to address in the respective sections, identify which parts could be sharpened, and give me suggestions on how to improve.
- Clean up my writing: As a speaker of English as a second language (and Spanish as a second language), I have asked ChatGPT to identify grammatical errors in my writing. I often ask it explicitly not to change wording unless it is blatantly wrong, so that I can keep my own voice in my writing.
- HTML formatting: For the archive page on my blog, I let ChatGPT generate the HTML code of my recent posts, so that I can more quickly add them (instead of having to edit every link manually).
- Assigning tasks: For the journal of which I am editor-in-chief, I keep an Excel file with the status of each article ID. For the weekly update I send to all associate editors, I feed the Excel file into ChatGPT and ask it (with a lengthy prompt), to make a list of pending action items for each editor.
- Develop prompts for think days: I organize Think Days for myself: sessions in which I take a deep dive into a certain topic that I want to reflect on and journal about. I develop the outline and journaling prompts with ChatGPT.
- Analyse my gym stats: I put in my PRs and weights I lift, and let it analyse my level of strength, the weaknesses in my strength profile, and which exercises I should add to my routine.
What has changed?
Personally, I can say that my use of generative AI has changed from playing around and trying out all the options, to a fixed set of activities for which I know that I can use ChatGPT in a solid way that does not interfere with my thinking processes.
At the same time, I also experience a certain AI-hype-fatigue. I still read updates on recent research and insights via LinkedIn and take a course here and there on Coursera on the topic, but I am not constantly looking for news and updates anymore. I guess this change reflects how the novelty is wearing off.
Conclusion
To conclude this post, I wanted to give one message. If you have not used generative AI tools yet, I would like to invite you to at least explore the tools, see what they can and cannot do, and think about how your students could be and may be using the tools already in class for their learning (or, to get their homeworks done faster). Once you have a feeling for the range of applications, you can think of a few repetitive standard tasks that you can outsource to ChatGPT.
Leave a Reply