WN #2

Nope, AI Isn’t Perfect. And That’s OK.

Nope, AI Isn’t Perfect. And That’s OK.

The recent AI boom has been met with plenty of skepticism. Among the biggest criticisms of generative AI models are their propensity to “hallucinate” when faced with limitations in training data, and to propagate the biases that are embedded in that data. This is enough to scare some educators away from using these new tools. But the very flaws in generative AI that critics highlight can actually serve as the most valuable teaching moments, equipping students with the critical thinking skills they'll need to navigate the complex world they are, and will be, living in.

“Hallucinations” refer to things that AI makes up – false, nonsensical, or irrelevant outputs that result from inaccurate predictions about what a plausible response might be. They are a function of how AI works. AI models are constantly predicting what should come next given the patterns they have analyzed in their training data. They often get it right, but sometimes they get it wrong. And when that training data contains biases, that influences AI’s predictions as well.

However, when we dismiss AI because of the potential for hallucinations and bias, we are ignoring the fact that “hallucinations” and bias run rampant in the information we consume every day anyway. Just open up a social media feed, or listen to anyone…I don’t know…talk.

It could be argued that the ability to verify information and be aware of embedded biases are among the most important skills we can teach our students today. And that’s where AI’s perceived shortcomings also provide tremendous opportunity. We need to teach our students not to blindly trust AI just as we shouldn’t blindly trust anything we see or hear. And when AI produces an output that is biased, the astute educator sees a teachable moment. What’s more, the ongoing dialogue that shapes our interactions with generative AI provides an opportunity for us to call out, modify, and rectify the output in real-time, the same way that filling out a feedback form, or leaving the 500th comment on a misleading social media post, doesn’t. So instead of shying away from AI's limitations, let's embrace them as classroom catalysts for fostering essential life skills in an ever-evolving digital age.

Districts are Changing their Tune on AI

A trio of articles published in the New York Times last week focus on how schools are coming around to the benefits of using AI. This one takes an in-depth look at how and why some schools have flipped from banning AI to embracing it. This one collects sentiments from teachers and students. And this one gives tips to help educators “survive” and “thrive” in the new AI-infused landscape.

LAUSD Creates Its Own Chatbot

Los Angeles Unified School District has created its own AI chatbot, “Ed”, to keep parents up to date on grades, test results, and attendance. It is part of a wider initiative to enhance support systems for students. While “Ed” isn’t a game-changer (schools have always had systems to communicate this data via reports, phone calls, and online messaging systems), the chatbot gives a glimpse into a future where schools and districts customize their own AI tools to serve their specific needs.

Amazon Provides Glimpse into AI Use Case for Educators

Amazon is playing with generative AI for product reviews. AI will provide an overall summary of a product’s features and customer sentiment collected from reviews, and make it easier to search reviews for specific types of information. What does this have to do with education? AI’s ability to quickly process and generate insights from a large number of separate submissions provides a window into how it could be used when evaluating student work.

📌 Listen in Your Own Language

YouTube is partnering with Aloud to add AI-powered dubbing to its videos. The goal is to break down language barriers for its users, democratizing content and information for a global audience. Check out some examples on Aloud’s website.

📌 AI-integration for Google Workspace

Google is working to integrate AI into its Workspace apps. The AI features are currently availble to try out on Docs, Sheets, Slides, and Gmail. If you want to give them a spin, you can sign up here to get free access.

📌 A Shortcut to Creating Engaging Videos

Want to create short videos for social media, YouTube, or your LMS (Canvas, Google Classroom, etc.), but hate recording voiceovers? Aug X uses AI to make it much easier for you, even letting you use your own voice by simply recording a snippet. They also have several other tools to help you create engaging videos even with no prior experience.

ChatGPT’s Custom Instructions

Last week I mentioned that ChatGPT’s custom instructions feature was recently opened up to free users. Above is an example of how you can leverage this feature as an educator. With custom instructions (turn this feature on in your settings) you can tell ChatGPT about your class, your students, your teaching preferences, or any other helpful context, and even how you would like it to respond. These instructions are pre-loaded into all of your future chats, making for a more streamlined way to customize your experience.

As of right now, you can only have one set of custom instructions, which you can toggle on and off as needed. I use ChatGPT for different purposes, so I like to keep a Google Doc with a list of custom instructions for different scenarios that I can copy and paste into ChatGPT depending on the task at hand.

Narrative Hooks to Increase Engagement

Now it’s easier than ever to kick off a unit or lesson with an engaging hook to increase student engagement. For the example above, I used the following prompt:

I am teacher starting a unit on European colonization of the Americas. Write a gripping narrative hook for this unit. Include the reader in the story in a way that inspires them to engage with the unit and sparks curiosity and wonder. Give me two options, one from the perspective of the colonizers and one from the perspective of the colonized.

This can be easily customized to any topic, subject, or grade level.

Note: The first iteration of this prompt didn’t ask for two options, and I only got a Euro-centric version. So I edited the prompt to include an alternate perspective, helping mitigate the inherent bias of the model. Going back to my opening thoughts in today’s newsletter, see what I mean?

  • With the proliferation of AI-generated music, Google and Universal Music Group are discussing a partnership to license artists’ voices, allowing for their legitimate use by creators and providing a revenue stream for copyright holders.

  • If you haven’t jumped into the world of AI image creation yet, get inspired by the fascinating works of creators like @benmyhre or the collections of @ai_art_community

  • Even though AI can be explained with math and science, that doesn’t mean it’s without its own mysteries

That’s all for this week! If you appreciate the content in this newsletter, consider subscribing for free, or sharing with people in your network who may find value in it.