As educators, we’re already living in the future. Artificial Intelligence (AI) has reshaped not only how our learners learn, but also how we teach.
Previous generations of teachers could only dream of a machine making a worksheet for them. As easily as students can use AI to write a paragraph response to “What is a central theme in Macbeth?” I can utilize it to create instructional sheets, quick blog posts (not this one) and even replicate the human voice.
When I think of AI, I think of ChatGPT.
What I didn’t know was that there were two types of generative AI. ChatGPT is akin to a massive dragnet scooping up the Internet’s collected open resources, which it mashes up into its algorithm. Like an icing cone, ChatGPT then channels the wealth of knowledge stored in it to filter out a response.
However, ChatGPT does not distinguish between mistaken information, bias, hate speech or other non-truthful content. Thereby, whichever response is generated by ChatGPT to a question is influenced by the totality of the Internet’s biases, fears, hopes and truths.
It’s like biting into a hotdog- sometimes you don’t want to know what’s inside.
Another approach is a small model AI. Notebook LM, a Google product, takes the opposite approach to ChatGPT.
Instead of filtering a wide ‘base’ of information into your response, LM works only off the sources you provide it.
For example, I could ask a question of it and upload 3 academic articles for it to interpret from. With LM, the AI only draws from the provided articles, helping you to curate the foundation of your AI product.
As an educator, this is incredibly valuable for preventing some of the less-than-classroom appropriate content that ChatGPT can raise. I am more in control of what sort of content is created.
The Podcast
For someone who’s never made a podcast before, LM made the job easy for me.
I only had to clicked a button (or two).
I wanted a podcast on how suspense is created within “The Monkey’s Paw” by W.W. Jacobs and “The Lottery” by Shirley Jackson. I uploaded these two stories in .pdf format as well as this TED Ed video on the same topic (linked here).
I also prompted LM to focus the podcast for a Grade 10 audience who need an introduction to what suspense is.
The podcast is linked here.
I was surprised at how natural the ‘podcasters’ voices sounded. Aside from a few “Right’s” that were awkwardly placed and too fast, they sounded like normal, if slightly ‘bland’ human voices. If I didn’t know an AI made this, I could easily have been convinced otherwise.
Overall, the podcast was accurate to the material. It accurately communicated what suspense is, and even had some applicable examples for a high school student (the feeling of waiting for a test to be handed back). It gave a serviceable analysis of how suspense operates within “The Monkey’s Paw,” and even applied the relevant portions of the TED Ed video to it.
Its analysis of “The Monkey’s Paw” had some holes, however. In one instance, the ‘podcasters’ analyse a detail of a strange man hanging around White’s door. The way the ‘podcasters’ frame this character implies that he is a clue in a series of events towards a much larger payoff. While this is- on a surface level- accurate, the stranger in Jacob’s story is there to inform the White’s of their son’s death at the factory. Instead, the podcast frames it in murder mystery-esque terms, misrepresenting the story.
Little errors like this make it important for the instructor to periodically pause the recording and correct any misunderstandings before students internalize them.
Moreover, there was no mention of “The Lottery.”
I was left in suspense for the podcast, wondering when Jackson’s story would appear. Alas, it did not.
If I was to use an AI podcast in my teaching, I would absolutely take the time to listen to it first. Twenty minutes of focused instruction is one thing for students, it’s another to assign a podcast on a story that is never mentioned.
While LM glazed over a third of it’s assigned sources, the AI podcast feature can be an effective tool in a TOC’s toolkit. In about 10 minutes (considering loading time), I was able to compile three sources to create a relatively focused and somewhat engaging teaching resource.