Truthfully, I’m kind of tired of talking about AI. Many faculty in my school (reasonably!) have a lot of anxiety about AI use by their students. Our school trains the journalists and communications professionals of tomorrow, I would HOPE they have some concerns.

It’s not fun to explain to a panicked, exhausted instructor that they are the AI detector they seek, despite what GPTZero promised. That the answer might be going back to the drawing board on some of the practices they’ve used for years with great results. That sometimes, even when you do everything “right” for your students to your best ability, some will choose to cheat or violate your policies. That yes, you can construct a set of requirements and surveillance tools so strict that most students won’t be able to cheat, but it may cost you your time and the trust of your students. That yes, you can “go back to paper” if you please, but it will have the same effect. That maybe too much effort is being invested in policing students, to the point that you’re becoming suspicious of everyone. 

That’s not to say all faculty are AI-averse. A select few live on the opposite of the spectrum, already experienced in using AI for things as complex as writing python scripts for their research. They are frustrated they are not allowed to use certain university data with the most cutting edge tools, and that we are so slow to approve and integrate new technologies as an institution. My job is to help them integrate these new, unvetted tools into their classes in a way that is compliant with FERPA and university data policies, a task that sounds like silly red tape to the instructor. I am tired of these conversations too.

Oh, right, this was a discussion post. You see, I’m so tired of talking about AI, that I can’t shut up about AI. 

What is generative AI?

• Generative AI is an umbrella term used to refer to a number of artificial intelligence models capable of generating unique outputs.
• These tools are trained on massive datasets, use natural language processing, pattern recognition, and are capable of producing unique outputs that can convincingly mimic human outputs.
• GenAI tools are “unsupervised” models capable of self-improvement in a way that traditional AI tools were not, with the ability to learn beyond the datasets they were originally trained on, and train on unstructured data.

How can it be used to aid learning and education?

• Intelligent search for research and learning: Traditional search engines are keyword oriented, and are hard to use as an early learner who lacks vocabulary in a subject. AI-based search tools can help predict what the user is searching for when they can’t fully articulate their question. Tools like Elicit and Perplexity are very compelling to me as they use AI to connect users with published sources. A big shout out to Angel for recommending these in one of the live class sessions!
• Task initiation and blank slate syndrome: I’ve observed in myself and other students that getting started on a project is often the hardest part, and this is the source of a lot of procrastination. With GenAI, learners can get past this stage more quickly and with less executive function needed. One strategy I use is to write a “brain dump” prompt where I first free-write, with no attention to structure. Then, I ask an AI to help me structure my writing into main ideas. When you truly have no ideas, you can ask AI to help you get started, though this should be used with caution if you’re not familiar with the subject as you won’t be able to verify accuracy.
• Alt text for screen readers: ChatGPT (and likely others) can already produce excellent alt text for most images. Tip: tell ChatGPT not to include “image of” when you request alt text and it will remember! I think this is my most consistent, reliable use case for GenAI as of right now in my personal workflows.
• For teachers, perhaps fewer papers to grade? The essay is not dead yet, but AI has certainly encouraged teachers to consider assessment formats other than writing. More often, students are asked to prepare presentations, visuals, videos, podcasts, and other multimedia format assignments. I am not a teacher, but I imagine these must be more pleasant to grade, yes? …joking but not

What are its pitfalls and limitations?

• Disinformation, authenticity and hallucinations: Most GenAI models are not concerned with the truth, only giving the user a response they will be happy with. Students are not equipped with the knowledge to judge the accuracy of responses, and risk learning incorrect or confusing information. This will have a compounding effect as AI models increasingly train on AI-created content. The Content Authenticity Initiative offers an interesting credentialing concept that could help if tech companies adopt it, but it is a long ways off, and doesn’t solve for text-based disinformation.
• Shrinking teacher presence and over-standardization of learning: Recommendations that teachers and instructional designers use AI to develop entire modules, especially for online courses, concern me. I am a big believer in teachers, and I think their unique personalities and idiosyncrasies are important. AI-generated content often adopts a formal, detached tone that is not engaging. AI lacks the experiences, stories, humor, and care for students that teachers provide.
• Capitalism and the inevitable enshittification of AI: For now, many AI tools offer free versions that are fairly high quality. There are no ads, at least explicitly in the way we see on social media apps or elsewhere. Yes, I’m paying with my data… but I gave up on fighting that battle long ago. This is a temporary golden era to gain users and future customers. Eventually, these companies will need to be profitable, and the most useful features will be locked behind ever-increasing paywalls. The models we see today are mainly trained on human-generated content, and the quality will degrade as the ratio of AI-generated content becomes the majority of what we see online. 

Here we are, at the end of the post, and I can’t help but feel like that was long and kind of a bummer? Maybe ChatGPT can help. I asked ChatGPT to create a more concise version of this post with a positive tone, and keep the prompt headings intact. Below is the unedited response.

ChatGPT Edition:

I’m aware that AI discussions are everywhere, especially among my colleagues who train future journalists and communications professionals. While some faculty are anxious about AI’s role in education, I see opportunities. Rather than focusing on policing AI use, there’s value in rethinking our teaching methods to stay ahead of the curve. AI detection isn’t flawless, and overly restrictive policies can erode trust with students. But we can embrace tools like generative AI to enhance learning while upholding our standards.

Not all faculty are wary of AI—some are eager to integrate the latest technologies, even for advanced tasks like coding. My role is to ensure these tools comply with university policies, which can sometimes feel like red tape, but it’s important to protect student data.

What is generative AI?

Generative AI refers to models that generate unique outputs using large datasets. These tools use natural language processing and pattern recognition, and they can even improve beyond their original training data.

How can it be used to aid learning and education?

AI can assist with intelligent search, making it easier for students to find sources even if they’re unsure how to articulate their question. It can also help overcome writer’s block by providing structure to ideas. Tools like ChatGPT even generate high-quality alt text for accessibility. For educators, AI is opening up new assessment formats like multimedia projects, moving beyond the traditional essay.

What are its pitfalls and limitations?

AI still struggles with accuracy and may produce misleading information. It’s also essential to maintain the human touch in education, as AI-generated content lacks the personal connections and care that teachers bring. Additionally, while many AI tools are free now, there’s concern that access will become restricted as companies seek profit.