What We Risk By Automating Tasks We Loathe
I don't think I've ever met someone who loves PowerPoint. Most students find it a rote mechanism to space out to during lectures. Some slide deck adherents would argue that this is due to lazy presentation skills, a lack of engaging visuals, or, as our former president opined, no "killer graphics."
Several software programs can now generate an end-to-end slide deck using generative text, images, and even audio and video. In fact, the phrase "end-to-end" is one to keep an eye out for because copilots aren't what most of the venture capital flowing into AI wants—autonomous agents that do tasks from start to finish.
I'll introduce you to both shortly, but before I do, I want you to question a premise that most AI marketing extols: Why not let AI automate the boring, repetitive tasks away for you? Doing so will give you time to pursue the work that is truly meaningful to you. On the surface, it is a great pitch, but what is lost when we automate tasks we loathe?
Meet Devin
The team at Cognition Labs produced one such agent called Devin. It is being marketed as an AI agent that can do parts of the job of a software engineer. This includes planning, debugging code, testing code, etc. Last week, I wrote about how autonomous agents might impact education by influencing and interacting with students. In the video below, you can see how Devin works to code, and think about the implications beyond software engineering.
Suppose we see the deployment of autonomous agents that can reason between complex problems and conduct long-term planning. In that case, we are moving beyond generative AI's stochastic arguments, simply repeating parts of its training data and entering frightening new territory. Of course, we're nowhere near a point where something like Devin is ready to be deployed meaningfully. It only completes end-to-end tasks accurately about 14% of the time. But don't let that unimpressive number fool you. When ChatGPT was launched in November of 2022, it couldn't strategically reason.
If various AI models continue to show advances in strategic reasoning, we are in for quite a ride. While we aren't there yet, I think taking the time to imagine what this might mean through something as low-stakes as PowerPoint is a helpful exercise in framing some of the potential implications agents might have for how we teach and how students learn.
Thinking Like An Agent-Generating Slides
Let's role-play as an agent and use some of the newly available tools to see how difficult it would be to create an end-to-end slide show using generative tools. Below are some options:
Gamma
Gamma is free to sign up and use. It recently upgraded to allow users to generate text, images, and slides. It is fully online and cloud-based. The speed it produces a slide deck is stupefyingly quick. I think Gamma is one of the best tools for this exercise.
Google Duet Workspace
Bard is now Gemini, and Google's Duet AI for workspace is likewise transitioning to Gemini for their workspace. You can sign up for a 14-day trial to play with Gemini in Docs, Sheets, Gmail, and Slides through Google Labs. According to Google's recent launch announcement, a user can generate a complete slide deck with Gemini—text, images, and video.
Beautiful.ai
There are several impressive AI presentation tools, some of which predate our current generative market. Beautiful.ai has established roots before GPT and does a pretty seamless job of creating a presentation. Though it isn't quite an end-to-end experience. It is, by design, a tool that invites a user to interact with it. Think of it like Copilot for slides. Unfortunately, there is no free trial.
Generating a Slide Deck Using Gamma
I used Gamma to generate a slide deck based on The Enduring Role of Writing in Our AI Era post from a few weeks ago. It took me less than a minute to prompt it to create a complete slide deck. The product isn't great, but I can certainly begin to see how the end-to-end experience of using various generative AI tools to build something feels like. If I were an AI agent, I would use a web browser to search for more sources, along with its strategic reasoning to hone, shape, and edit the output. The image is linked to the slide deck and you can click on it to access it.
We Shouldn't Fully Automate the Rote Tasks
I truly despise presentation software. For the past 18 months, I've had to build and rebuild slide decks just to use as a visual background in talk after talk about how quickly generative AI is moving. Updating slides is a pain, and building one from the ground up is even more so, so why would I rather create a slide deck than cede even a task I loathe deeply to generative AI?
Each presentation is, at its core, a performance. Current etiquette requires that I provide attendees with some type of visual text or image signposts to guide my talk. Yes, it is boring. Yes, it is stale. Yes, hitting a generate button unburdens my mind from the task of creating "killer graphics" to accompany the information I am trying to convey. But I feel like there is a significant danger in automating away some part of this dynamic.
How would you feel if you sat through an hour long presentation, reading and seeing nothing but generative material? What would this do to your experience of being an audience member? Would you view the speaker with trust and find what they said of interest or would you instead be disinclined and skeptical because the speaker used AI to supplement their talk? Would you wonder if the words the speaker chose were their own?
As much as presenting material with presentation software has become a rote exercise, it still carries with it a substantial weight. If we find ourselves ceding the boring and mundane tasks to AI, we risk alienating the very people we are attempting to engage. This isn't to say that a presentation is engaging by default in the first place or that there isn't a place to use AI in the production of one. Rather, it's a reminder that the decisions we make have a greater impact downstream than we realize.
What's the Right Attitude Toward AI?
I've been called both an early adopter and a too-critical opponent of AI. The truth is, I don't neatly fit into either of those boxes. I've been using generative AI tools since May 2022, and my position then remains the same now — I'm a curious skeptic.
I'm fascinated by the potential of these tools to augment and enhance our work and creativity. There's no denying the impressive capabilities we're already seeing with text generation, image creation, coding assistance, and more. Used thoughtfully, AI can be a powerful productivity multiplier.
At the same time, I have significant concerns about the broader implications of this accelerating technology, especially for education and society at large. We're traversing new ground at a breakneck pace, and it's crucial that we don't blindly embrace AI without considering the potential risks.
My worry is that by automating away too many tasks, even seemingly rote ones like creating slide decks, we risk losing something vital—humanity at the heart of knowledge work. Presentations are indeed often boring affairs, but they represent the culmination of a person's work, ideas, and thought process. They are not thoughtless tasks to be offloaded. If we simply generate a mundane task through AI without any personal investment, what does that do to the speaker-audience connection? The feeling of authenticity and trust?
I don't have all the answers, but I believe the right approach is one of prudent exploration as opposed to ostriching our way through this. We should absolutely leverage AI to augment our abilities but be wary not to entirely replace the human in the loop. I believe that there is immense value in maintaining personal investment, even in mundane tasks, but I fear the existing expertise we've established about applied AI is already settled in comfortable siloed camps with firm attitudes about adopting or resting AI while the rest of the world simply ignores it until they run across a new button in an existing app that magically completes the task before their eyes.
When I'm confronting a talk and thinking about having to put together yet another slide deck I do groan inwardly, but once in the process I realize how important it is to think through what I'll be sharing with the audience and why. It's a great exercise in clarifying my message, particularly as I'm going to do it in real-time in front of an audience, unlike writing where I may never know the audience's reaction in a definitive way. Yes, it can feel like drudge work, but it's necessary drudge work.
Interesting discussion. One thing you don’t mention about Gamma (and I imagine the others) is that the user can edit the slides AI spits out. This way, producing the slides feels, to me at least, more like a collaboration. I also think that audience matters as does the subject of the talk. I use Gamma in my Digital Storytelling class. I am transparent with my students when I use it and allow them limited use with transparency a requirement for them, too. This way we talk about why we use AI for some tasks and not others and interrogate those assumptions together. I am, like you, a curious skeptic. I am going to share this piece with them :)