During Summer 2023, Teaching and Learning Innovations hosted a Summer AI Residency to interrogate, tinker, and explore generative artificial intelligence (generative AI) and large language model (LLM) diffusion in higher education. Our Summer Residents included Jerilee Petralba (Staff in Residence) and Dan Lenz (Faculty in Residence).
This week, we are thrilled to share this post from TLi’s Summer AI Staff-in-Residence, Jerilee Petralba.
When I heard the buzz about OpenAI/Chat GPT, a large language model chatbot, my initial reaction was unease. However, the lure of innovative technology that could streamline my work outweighed my somewhat irrational fear about the potential creation of a Skynet-like scenario. Reservedly, I created my account and started my problem-solving conversations with ChatGPT. I am still very cautious about trusting the mind-blowing changes AI will surely usher in.
I recall the excitement of making my Twitter, now called X, account in 2009. I remember how social media felt revolutionary. The instantaneous updates were addicting yet foreign. Although some benefits still outweigh the toxic discourse prolific online, I never fathomed the harmful aspects. I remain actively engaged yet highly conscious of the potential for misuse and abuse, lessons learned after Twitter. Below are some key takeaways from my exploration of AI this summer.
Posing a question may appear uncomplicated, but the outcomes can vary significantly. It’s crucial to establish clear parameters. The more detailed and specific your prompt, the more accurate the results. If you veer away from your initial question, you can easily reference past prompts to get back on track. ChatGPT also learns from your previous interactions and provides better responses to follow-up questions related to prior queries, a feature I like the most.
Usefulness and Productivity
Before chatGPT, I would scour online forums, look for YouTube video tutorials, and conduct Google searches because there’s always someone who already asked the same question online. ChatGPT has significantly narrowed down my search window. When asking questions for Software as a Service (SaaS) products like Airtable and Power BI, it’s especially beneficial. The solutions for questions regarding formulas are less open to interpretation, so the AI is more decisive in providing scripts, code snippets, or equations as potential solutions to your query. The solutions offered frequently fail to deliver the code I want. I typically respond with phrases like “That didn’t work” or “I got this error (insert error message.”). This can result in the instant generation of another set of codes or formulas.
The most impactful AI is the task-specific products since they deliver the most everyday efficiencies. When I discovered Scribe during TLi’s Summer AI Residency, it was a game changer. Much of my workload requires a lot of process mapping and the creation of guides & tutorials. I spent significant time writing instructions, capturing, cropping, and redacting images. With Scribe, the task is significantly easier and cuts down many tedious and repetitive steps. It doesn’t exactly do the job, but it provides a document I can edit to fit my needs.
Fair Use versus Copyrights
As an artist, I’ve used generative AI to help build concept images for reference. It’s much faster and helps visualize and fine-tune color schemes, themes, and styles for inspiration in under a minute. However, it’s hard to consider a design solely mine if I only use AI prompts to generate images. I’ve also used ChatGPT for Marketing copywriter-type queries. If I need to rewrite a phrase or marketing blurb, I’ll enter the phrase in ChatGPT and enter prompts like “more suggestions,” “keep it concise,” or “add/remove (insert word here)” to yield more optimal suggestions. Since English is my second language, I’ve found the grammar part of ChatGPT as an asset. English sentence structures are intimidating, and using AI to review my writing and check for grammar errors is such a boon.
With AI help, I grapple with the idea of fair use versus copyright infringement. Are the results mine since the original phrasing was mine, or is this someone else’s work? Recent lawsuits are winding their way through national and international courts regarding copyright infringement of creations used as data points to train generative AI. As I write this article, the US government is unveiling oversight via executive order. We may also have to wait for copyright laws to help shape the use and future of AI. I am still trying to decide where I stand.
Privacy, Policies, and Paywalls
Although creating an account is free, nothing is truly free. I am fully aware that the data I provide and share with the different AI systems is a form of payment. For the system to learn, it needs input and interactions from humans. As a campus employee, I have to be mindful of the campus guidelines that may limit how much information I can ask or share when interacting with AIs. I must ensure that I share no personally identifiable information (PII) and safeguard the privacy of students and the campus. Lastly, the paywalls may also be a barrier since some AI products require monthly or annual subscriptions to access all features.
AI provides many options and directions you may have yet to consider. I’ve spent hours asking ChatGPT for Airtable coding for a project and could not get it to work. But I found a solution when I talked with my colleague. It needed no coding and was solved using common sense applications. It’s so easy to overcomplicate projects with AI when they may not even require scripting or complicated coding to solve. There are considerable time savings with the instant responses and solutions AI delivers. But it can be an endless loop of back-and-forth interaction with a machine, so be mindful of your time investment. AI is a tool. It helps us finish our jobs faster, but it’s not a replacement.