Saving 1-3 Hours Per Lecture with Copilot

I preview PowerPoint's new functionality, and I discuss whether AI skill-leveling is always good.

[image created with Dall-E 3 via ChatGPT Plus]

Welcome to AutomatED: the newsletter on how to teach better with tech.

Each week, I share what I have learned — and am learning — about AI and tech in the university classroom. What works, what doesn't, and why.

In this week’s piece, I discuss Microsoft’s Copilot options and how they can save us time, and I reflect on when AI-fueled improvements in student outcomes are good and when they are not.

💡 Idea of the Week:
Is Higher Performance Always Good?

Back in October, Caleb reported on several studies that showed that most workers who used AI saved time using it, but “less skilled workers gained more from AI assistance than the more skilled ones.” This means that teams of workers using AI had more “skill equality” than those teams who did not use it.

These business results are being corroborated in educational settings. A recent study on law school students, “Lawyering in the Age of Artificial Intelligence” in Minnesota Legal Studies Research by Jonathan H. Choi, Amy Monahan, and Daniel Schwarcz, found that

access to GPT-4 only slightly and inconsistently improved the quality of participants’ legal analysis but induced large and consistent increases in speed. AI assistance improved the quality of output unevenly—where it was useful at all, the lowest-skilled participants saw the largest improvements. On the other hand, AI assistance saved participants roughly the same amount of time regardless of their baseline speed.

Choi, Monahan, and Schwarcz

This result fits with what I have seen in my own experience teaching Philosophy at the undergraduate level. Students can analyze and create arguments — sequences of claims that conjoin to support a given conclusion — faster and more effectively with AI, and differences between students are reduced through AI access (especially if students consult AI after attempting a problem or exercise on their own). The custom GPTs that I have built for my students have only amplified this effect, like the one I built for my “Philosophy of AI” class.

A custom GPT I built that my students use in one of my classes.

My idea of the week is this: while normally we would be pleased if our teaching were to lead to our students being homogeneously high performing, it is not clear if we should feel the same way about students who arrive there via AI-infused teaching.

To see the worry, consider if your students met high standards and received high scores by relying on their friends or an overly helpful tutor to complete their assignments. It would normally be a good thing for them to be performing so well across the board, but not if they get there via the wrong sort of shortcut.

My current thinking is that the lesson here is one we have already discussed at length, especially in the context of assignment design: namely, that the power of AI requires a reappraisal of our objectives and goals.

The law school students in the aforementioned study are completing the same assignment that they would have been tasked with, had they taken the class before the advent of AI. (Sameness of assignment is a necessary prerequisite for the authors to compare pre- and post-AI outcomes.) Now, if this sort of assignment mirrors a task that they will be expected to complete as lawyers, then their ability to complete it more consistently and effectively with AI would seem to be a good thing. How would it be bad if, with the help of AI, more people can lawyer better?

However, problems arise if their professors aim to have them complete an assignment that doesn’t align with their future careers. Perhaps the assignment does not incentivize them to be sufficiently sensitive to ways that AI can make mistakes. While they can generally complete similar tasks with AI, they would sometimes fail — unbenownst to them — with disastrous results for their hypothetical clients. Or perhaps the assignment was originally intended to build fundamental skills that would be, in turn, used for the tasks of their careers. Lacking these fundamental skills because they passed the assignment to the AI, they would be lost in their careers when asked to complete tasks neither they nor AI can complete well.

In these cases, AI can help students complete an assignment at a homogeneously high level, but there is reason to worry that this is the wrong sort of shortcut. Yet, whether this is a legitimate worry in this situation depends entirely on countless facts about the assignments, the practice of law in the future, current AI tools, the similarities between those tools and future ones, etc. And that is precisely my point.

👀 What to Keep an Eye on:
Our Microsoft 365 Copilot How-To

This coming Wednesday (February 7th), we will be releasing our first Premium piece on how to best leverage Copilot for Microsoft 365 as a college/university educator. I have been pressure testing all of its features and experimenting with creative ways to use it to save you time and improve your pedagogy, whether with PowerPoint, Word, Excel, or Outlook.

I am not quite done with my analysis, but I currently estimate that if you receive lots of emails, have many students, and/or teach two or more classes a semester, you would save 5-10 hours a week if you used Copilot and followed all of our Premium piece’s tips and tricks.

In today’s piece, I will preview one of the more straightfoward use-cases of Copilot for Microsoft 365. But first, I want to explain a bit about what Copilot is. Copilot has four main manifestations or versions:

  1. the free web version — “Microsoft Copilot” — which used to be called “Bing Chat” and which is an AI-powered search engine analogous to Google’s Bard;

  2. the Windows version — “Copilot in Windows” — which can “provide centralized AI assistance to you” in your operating system;

  3. the version for Microsoft 365 Enterprise, Business, and Education — “Copilot for Microsoft 365” — that we are most impressed by; and

  4. the version for Microsoft 365 Personal and Family — “Copilot Pro” — that we have found to be less useful (in fact, some of the promised features do not work at the moment, like file referencing).

In all of these cases, Copilot is a large language model (LLM) running Microsoft’s “Prometheus” custom deployment of OpenAI’s ChatGPT4 that merges it with Bing and other Microsoft software.

The main differences between the first three versions are what they have been tuned to do and what information they have access to. For instance, Copilot for Microsoft 365 has access to your emails and the files in your OneDrive, while Microsoft Copilot (the free web version) does not. The fourth version is an attempt to extend the power of Copilot for Microsoft 365 to the rest of the 365 plans.

In terms of access, most everyone globally can access the free Microsoft Copilot via their web browser; many users of Windows can download Copilot in Windows or will be able to do so in the coming months (click here to read how to get it); your organization needs to purchase Copilot for Microsoft 365 and give you access to it; and you can buy Copilot Pro for $20/month in English markets if you have a Microsoft 365 Personal or Family plan.

Two weeks ago, Microsoft released some tips for using the free Microsoft Copilot as an educator. These fit with our prior advice, like that which we gave 5 months ago for prompting to generate lesson plans with ChatGPT. This builds on their December release of a version of Microsoft Copilot with “commercial data protection” (the tool formerly known as “Bing Chat Enterprise”). My university has released a useful discussion of this enhanced version.

More interesting is the announcement from December that Microsoft has now made Copilot for Microsoft 365 available to “education faculty and staff.” What this means is that it is now available for your institution to purchase and enable for your use, if they choose to. So far, most have not chosen to do so.

Since we expect many institutions to purchase and enable Copilot for Microsoft 365 this year, I am going to focus on it in this piece and in our Premium piece for Wednesday.

Two weeks later, we will release a parallel piece on the effective use of AI in Google Workspace.

🧰 An AI Use-Case for Your Toolbox:
Microsoft 365 Copilot for Lecture Slides

I hate PowerPoint. Still, I use PowerPoint as a visual aid in my presentations and lectures. In another piece, we can unpack this tension, but for now I want to briefly highlight one way that Microsoft 365 Copilot can be used to create slides.

To put it bluntly, I think that Copilot saves me 1-3 hours per presentation, at least for those presentations where I need to build the PowerPoint from scratch.

It takes me from blank page — or old theme I have been meaning to update — to the skeleton of a decent presentation in 5-10 minutes.

Step One: Locate/Format Relevant Documents

There are two differences between ChatGPT4 and Copilot + PowerPoint:

  1. ChatGPT4 stinks at producing PowerPoints while Copilot does not; and

  2. Copilot has access to all of the files in your OneDrive.

Supposing you want to create a presentation based on a document — like a lesson plan or a course reading — the first step is to locate that document and place it in your OneDrive.

Next, ideally you would structure the document in a way that is easy for Copilot to evaluate. Like ChatGPT and other LLMs, Copilot works best with documents that have some internal structure, like tiered headers and other tags. However, it is not necessary to modify the structure of the document, and it is worth a shot to see if Copilot can perform well in producing a PowerPoint from it without any modifications.

The PowerPoint I will use as an illustration here will be lecture slides about a course reading that my students would read in advance of a lecture in my political Philosophy class. This reading is an excerpt of a journal article from Jason Brennan (“Polluting the Polls: When Citizens Should Not Vote” in the Australasian Journal of Philosophy) that concerns why citizens have an obligation to not vote badly. Since this reading is in pdf form, I copy-pasted its content into a Word document and saved it on my drive without any further effort, so as to enable Copilot to parse it effectively.

Step Two: Prompt

The second step is to tell Copilot what you want to do. Click the new icon in the upper right corner of the PowerPoint Ribbon:

The Copilot prompt window will open on the right side of your screen. There, you have 2000 characters per prompt — you can chain prompts — to tell it to create a presentation from the file you located and/or formatted. You can reference the file by typing “/” followed by the file name, if you have opened it recently and your OneDrive has indexed it, or you can reference it via a direct link to the file in your OneDrive (right click on the file in your File Explorer and click “Copy Link”).

Here is what it looked like for me:

Without any editing or further prompting (you can always respond in the prompt window with desired adjustments), the result was sensitive to the content of the document. Having taught this content recently, I was impressed by the result.

Here are three of the slides from the slide preview pane, as well as one slide in its fullness, without any edits from me:

Not bad for a few minutes of work and a completely blank prompt! I am off to the races…

In our Premium piece Wednesday, I will discuss ways to improve the outputs of Copilot in PowerPoint, including some prompts that I have found to be very effective. I am very excited about the present and future of Microsoft 365 Copilot to save time and improve our pedagogy.

Two weeks ago, we updated our college/university course design GPT so that it can produce better assignment rubrics. Give it a try, if you have ChatGPT Plus! It can produce assignments, assignment sequences, rubrics, and AI course policies. We have designed it to be especially effective when it comes to pedagogical issues related to AI.

You can even get it involved in any other GPT conversation you are having by @’ing it!

📝 How to Sign Up for Our
“Build Your Own GPT” Webinar

Our 2-hour "Build Your Own GPT" webinar is rapidly approaching! It is scheduled for Saturday, February 17th at 12pm New York time. We are excited to see the interest in this event!

In short, it is a Zoom-based learning experience for anyone who wants to incorporate custom GPTs into their pedagogy this year. With evidence growing of the effectiveness of this sort of intervention, now is the time to jump on the opportunity afforded by recent advances from OpenAI.

The case for leveling up your skills creating custom GPTs is only made stronger by Microsoft’s release of Copilot Studio, which brings the power of custom GPTs to the Microsoft 365 ecosystem.

We have only 6 slots left in the first cohort. We want to keep each webinar small to maximize value for all attendees.

Now, you might be thinking: “wait, Graham, what are custom GPTs?” In brief, they are AI chatbots fueled by ChatGPT4 that you supply with instructions and files, thereby empowering them to effectively play a unique role. I have discussed using GPTs for in-class activities, as they can enable you to coach your students at scale. I use two in my current classes, and my students report that they find them incredibly useful. In fact, they often ask me whether my GPTs get anything wrong, given their impressive capabilities (the answer: “yes, they are fallible” and a reminder of why a critical AI literacy stance is crucial).

The webinar will occur on Zoom on Saturday, February 17th, from 12pm to 2pm Eastern Standard Time. By the end of the webinar, you can expect to have learned how to develop your own GPTs — but you will also walk away with one already developed!

The price is $99. All Premium subscribers ($5/month or $50/year, with prices going up soon) get a 10% discount code, included immediately below.

To sign up, click this link or the below button:

🔜 What’s Next for Premium Subscribers

Late in the fall of 2023, we started posting Premium pieces every two weeks, consisting of comprehensive guides, releases of exclusive AI tools like AutomatED-built GPTs, Q&As with the AutomatED team, in-depth explanations of AI use-cases, and other deep dives.

Our next two Premium pieces will be released on the following dates and will cover these topics:

  1. February 7th - an AI use-case deep dive into how professors and others in higher ed can best leverage Copilot for Microsoft 365.

  2. February 21st - an AI use-case deep dive similar to the above but focused on Google’s Duet AI for Workspace.

If your college or university uses the Microsoft 365 suite or Google Workspace, you won’t want to miss these deep dives.

So far, we have four Premium pieces:

To get access to Premium, you can either upgrade for $5/month, $50/year, or get one free month for every two (non-Premium) subscribers that you refer to AutomatED (note: we expect to raise prices this spring, so now is the time to lock in a lower rate).

To get credit for referring subscribers to AutomatED, you need to click on the button below or copy/paste the included link in an email to them.

(They need to subscribe after clicking your link, or otherwise their subscription won’t count for you. If you cannot see the referral section immediately below, you need to subscribe first and/or log in.)