7 Tasks To Do Before Spring

How to make the most of a pause in instruction.

Author

Graham Clay
December 20, 2024 • Estimated Reading Time: 22 minutes

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

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

In each edition, 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 second-to-last piece for 2024, I…

  • outline 7 tasks you might consider doing in the next few weeks, including one that probably shouldn’t be called a ‘task’

  • share the latest AI news, including some surprising developments from the big dogs

  • and reflect on my free webinar from December 6th (and ask for your opinion about the next one)

Let’s dive in!

💡 7 Tasks to Do Before Spring

When I give talks, trainings, and webinars, many professors report to me that they appreciate it when I reveal more of my own views and predictions on the “state of play” of the AI-education space, especially during Q&A periods (I guess that’s when I loosen up a bit).

I figure now is a good time to do something analogous here, with a focus on what tasks I personally think you should be working on now.

Here they are, in rough order of operations…

1. Experiment With the Latest AI

The AI landscape is constantly evolving — the amount of money, effort, and intellectual firepower involved in developing and improving AI tools is absolutely massive right now — and it has shifted dramatically in just the last few weeks. As I detail in my “Quick Hits” news section below, we're seeing groundbreaking developments from both Google and OpenAI, from autonomous research assistants to more powerful language models to sophisticated video generation tools.

If you haven't hands-on tested these tools recently, your mental model of AI's capabilities is outdated. The rate of development makes staying up to date hard, but being outdated is a matter of degree, and you don’t want to be too far out of date, whether as a teacher or as a researcher.

In other words, this isn't just about staying informed in an abstract sense — it's about understanding how these advances impact your field.

For instance, Google's new Gemini 2.0 (which I explain below) isn't just another incremental update; it's being implemented now to deploy supervised "agentic" capabilities that could transform how we approach complex academic tasks. Google launched "Deep Research," an AI research assistant in Gemini Advanced that can autonomously explore complex topics across the web and compile comprehensive reports with source links. Meanwhile, OpenAI's new premium LLM “o1 pro mode” is showing unprecedented reliability on advanced academic problems. (Below, I ask you to send me any tough problems to feed to the beast; I bought it for a month, for science.)

Take advantage of this time to experiment with these tools in your specific domain.

Sure, you can also read about what they can do, but I think you can’t just read about them — you need to test them with real academic challenges from your field and your own classes, especially given you have unique expertise and knowledge. Your firsthand experience with these tools will help you make informed decisions about how to integrate them into your teaching, how to modify your methods to avoid their pitfalls, and how to guide your students in using them effectively.

Relevant Resources:

2. Review How Last Semester Went

Now that grades are submitted, it's worth taking a systematic look at how your courses went. Tools like ChatGPT's Advanced Data Analysis can help analyze your (anonymized) gradebook data in ways that weren't easily accessible before. Beyond simple grade averages, you can explore correlations between different assignments, identify where students consistently struggled, and spot patterns across different types of assessments.

The data might surprise you. Perhaps those early writing assignments weren't actually predictive of final essay performance, or maybe quiz scores in certain units highlighted conceptual gaps that persisted throughout the term. Understanding these patterns can help you adjust your approach for spring — whether that means restructuring assignment sequences, rethinking assessment types, or identifying where students need additional support.

Just as importantly, reflect on how AI impacted your teaching this past term. What worked about your AI policies and what needed clarification? Were students asking for more guidance on appropriate AI use in specific contexts? Did you notice any patterns in how students incorporated (or misused) AI tools? Taking stock now, while the semester is fresh in your mind, will help you make thoughtful adjustments for spring (a topic I cover in the next item on this list).

Relevant Resources:

3. Engage in AI-Sensitive Course Design

The break between terms is your chance to fundamentally rethink your courses through an AI lens. With new AI tools emerging every other month and students becoming increasingly sophisticated in their use, last semester's course design might already be outdated. This isn't just about tweaking a few assignments — it's about reimagining your entire course structure, assessment methods, etc.

Start by evaluating each assignment and assessment against current AI capabilities. Which ones are still achieving their intended learning objectives, and which have become vulnerable to AI shortcuts? Consider whether traditionally written assignments might work better as oral presentations, or if take-home essays should shift to in-class writings. Multimedia is still a big challenge for AI. But remember that format changes alone aren't enough — you need to think about content too. Tasks requiring specialized knowledge application or complex, multi-step reasoning often remain more resistant to AI than those asking for simple synthesis or summary, although this gap is shrinking (see above).

But don't just focus on preventing AI misuse. Think about where AI tools could actually enhance learning in your course. Could a custom GPT help students brainstorm research topics? Might ChatGPT serve as a practice partner for language learners? The key is designing a course that thoughtfully incorporates AI where it adds value while maintaining the integrity of core learning experiences that need to remain AI-free.

Finally, reflect on whether you want to fundamentally change the paradigm around your use of student data, with AI or otherwise. If, for instance, there are tons of ways to use student data that you wish you had student consent to carry out, now’s the time to implement a way to acquire and manage informed student consent.

Relevant Resources:

4. Build Custom GPTs, Gems, Projects, … For Each Use Case

Now's also the time to create AI tools tailored to your specific teaching, research, and administrative needs. With custom GPTs, Google Gems, and Claude Projects, you can build AI assistants that understand your context, are encapsulated for a specific use case, and perhaps even shareable. The key is to develop these tools before the semester starts, when you have time to test and refine them.

Start by identifying specific use cases where AI could enhance your teaching, say. Maybe you need a feedback assistant to help grade more efficiently, a course design wizard to help plan assignments, or a tutor to guide students through complex problem-solving.

Building these tools effectively requires careful planning and testing. Your instructions need to be clear and comprehensive, your knowledge base needs to be well-organized and relevant, and you need to experiment with different approaches to find what works best.

Good news: I'm excited to announce that I have been collaborating with Charlie Fuller — an e-learning expert and learning experience wizard — on an interactive and engaging course that walks you through building custom GPTs. By the end, you have at least one custom GPT, fully ready to deploy.

This course is like giving steroids to my ✨Tutorial on the topic. We supplemented and updated its content, and we also created a detailed template to help you craft your GPT’s instructions. We even built a GPT tutor to guide you through the process.

The course itself consists of a sequence of short, engaging videos that are bookended by illustrative text and images. It is much more manageable way to approach the challenge of custom GPT design, breaking down the process into more actionable units that are illustrated clearly so you know exactly what to do.

The course will be released in the next 3 weeks.

If this interests you, express interest via the poll below so we can keep you in the loop.

Should I notify you when our custom GPT course is available?

Login or Subscribe to participate in polls.

Relevant Resources:

5. Get Your “Dumb” Tech Sorted Out, Too

While we're all excited about AI (or maybe it’s just me…), don't forget about the basic tech tools that can make your life easier. Here are four quick wins to implement during your break:

First, set up automated shortcuts for your daily routines. Whether using Apple Shortcuts, Windows Power Automate, or Google App Script, create simple automations for repetitive tasks: launch your morning browser tabs in one click, display your daily meeting schedule at a glance, or create a pipeline from voice notes to email drafts. These small time-savers add up.

Second, declutter your digital workspace. Clean up your inbox with rules that automatically sort incoming emails (e.g., student emails by course section, department announcements, or committee work). Archive old files, organize your cloud storage, and create a logical folder structure for the upcoming semester.

Third, optimize your notifications. Most of us have our devices interrupting us hundreds of times daily. Take 30 minutes to audit your notification settings: silence non-urgent apps during teaching hours, set up VIP contacts for important messages, and create focus modes for different types of work (grading, research, meetings).

Fourth, configure your browser for efficiency. Set up bookmark folders for each course, install useful extensions like Zotero for research management or Grammarly for typo catching, and organize your tabs into groups by project or course.

Relevant Resources:

6. Reflect on AI’s Role in Your Research

The AI capabilities that are transforming teaching are equally revolutionary for research. With tools like Google's Deep Research assistant (see above), ChatGPT Pro's enhanced academic reliability, and specialized research models being deployed at major institutions, now is the crucial time to consider how AI fits into your scholarly work.

We’re getting to the point that AI can help with almost any text-related tasks, including highly advanced ones like your research. You don’t want to look up in 2026 or 2030 and be woefully behind a paradigm shift in methodologies.

Consider setting up a controlled experiment with your own research: take a small piece of your current work and try processing it through different AI tools. Test Perplexity for literature review, see how ChatGPT Pro handles complex theoretical analysis in your field (or send me a sample challenge for me to do it), or explore how these tools might help with data analysis or writing.

Relevant Resources:

7. Take a Nap

No, seriously! You deserve it.

1. Google launched "Deep Research," an AI research assistant in Gemini Advanced that can autonomously explore complex topics across the web and compile comprehensive reports with source links. The feature, available now on desktop and coming to mobile in 2025, represents Google's first "agentic" feature that can perform multi-step tasks under user supervision.

This is a development educators should take a close look at — and keep an eye on through 2025. I will do a deep dive on it early in the new year.

2. The launch of Deep Research is part of Google’s announcement of a new frontier model, Gemini 2.0 that is especially well-suited for "agentic" capabilities that allow the AI to take actions on users' behalf (with supervision). Key features include native image/audio output and tool integration. They're also showcasing research prototypes like Project Astra (for phones/glasses) and Project Mariner (for web browsing), though access is currently limited to “trusted testers.” 2.0 Flash, an experimental version, is available now in the Gemini app.

3. Finally, Google updated NotebookLM with three major changes: a new three-panel interface for better content management, interactive voice conversations with AI hosts during Audio Overviews, and a premium "Plus" subscription for power users that offers 5x higher usage limits. The Plus version will be included in Google One AI Premium in early 2025.

4. OpenAI announced ChatGPT Pro, a $200/month tier giving access to their most powerful models including a new more compute-intensive "o1 pro mode" (as well as higher usage limits on "o1," "o1-mini," and "GPT-4o"). The pro mode showed significantly better reliability on PhD-level science questions and coding challenges, with OpenAI announcing plans to award initial grants to medical researchers at leading institutions.

I went ahead and bought it for one month (for science). Please email me with any hard problems you want me to give to it from your field and I will report back on the results, unedited:

(Also, at the start of the new year, I will be discussing the growing disparities in model access that I predict are coming — as well as why there are still a lot of bright sides of the coming year of AI.)

OpenAI also released Projects, which help you organize your chats, and launched Sora Turbo, an improved version of their text-to-video model, available now on sora.com to ChatGPT Plus and Pro subscribers. The new version is significantly faster and includes features like 1080p resolution, 20-second videos, and a storyboard tool. Notably, all generated videos will include C2PA metadata for transparency and verification, with strict safeguards against misuse like deepfakes. (Be sure to check out Google’s Veo 2, too.)

5. A small study in Computers & Education found that students using ChatGPT for help with essay writing showed markedly different patterns than those working with human teachers. Students using ChatGPT tended to skip evaluating the help received and jumped straight to implementation, while also showing more "executive help-seeking" behaviors (asking for direct answers rather than guidance). The researchers suggest this indicates a need for better scaffolding to ensure students engage metacognitively when using AI tools.

6. Microsoft released Phi-4, a small LM that “excels at complex reasoning in areas such as math, in addition to conventional language processing” and like, Meta Llama 3.3, a new free open-source LLM you can run on your own computer, is comparable in performance to GPT-4. Remember: GPT-4 was released in March 2023 and was far ahead of all others for 6+ months; it’s the model that powered ChatGPT to popular awareness. (This is related to the “bright sides” I mentioned above.)

7. Google DeepMind's new AI weather forecasting model "GenCast" has achieved unprecedented 15-day forecasts, outperforming the European Center for Medium-Range Weather Forecasts (the current world leader) 97.2% of the time. Unlike traditional supercomputer forecasting, GenCast learns weather patterns from 40 years of historical data and can generate probabilistic predictions much faster. (Original paper in Nature here. Discussion in the New York Times here.)

✉️ What You, My Subscribers, Are Saying

My December 6th Webinar
on Feedback & Assessment

Two weeks ago, I hosted my last webinar of 2024 on "How to Use AI to Improve & Accelerate Student Feedback." With nearly 100 educators signed up, we explored concrete strategies for using AI to provide better feedback faster while maintaining student data privacy. The 90-minute session included live demonstrations of ChatGPT, Claude, and custom GPTs for feedback generation, along with detailed discussions of ethical considerations and implementation strategies.

The response was overwhelmingly positive, with attendees especially appreciating the practical, hands-on approach and ready-to-use prompts for different teaching contexts. As with previous webinars, all attendees received the full recording, slides, AI-generated summary, and several Premium resources focused on AI-assisted feedback and student data privacy.

Thanks to all who attended or engaged with me after viewing the recording.

Your Comments

Here are some of your comments:

“I like that you provided actual steps and prompts that can be used to summarize the notes into useful feedback. I find some of these tutorials tell you all the great things Gen AI can do and don't really give you the how to. I like that we can see your prompts which we can use or they can provide us what a good prompt should look like so we can create our own prompts.”

Anonymous Subscriber

“I think that the seminar was great for people with a very good understanding of, and experience with, generative AI. I liked how you provided broad explanations followed by more concrete applications. I also think that all the materials you provided after the seminar (recording, guides, etc.) is super helpful!”

Anonymous Subscriber

“You opened my eyes to different thinking/approaches to incorporating AI into feedback! Much appreciated.”

Anonymous Subscriber

“I was so glad that this wasn't a 101 webinar—sometimes they are hard to find. ”

Anonymous Subscriber

“You gave a good overview for someone (like me) who is a neophyte and wants to learn more.”

Anonymous Subscriber

Thanks for the feedback!

What Should the Next One Cover?

Now the question is:

Please tell me your answer, if you didn’t get a chance to do so last week (note: after you make a selection, you will be redirected to the website where you can also offer typed comments, if you want):

What would you like the next (paid) AutomatED webinar to cover?

Only select what you'd be willing to pay for...

Login or Subscribe to participate in polls.

I did get a good comment/question from one attendee of the webinar who is a Premium subscriber. If I were to have a webinar on how to build custom GPTs, the person asked me whether there be a lot of overlap with my pre-existing ✨Tutorial: How to Build (Better) Custom GPTs:

“I have build AI Tutor GPTs for two of my tax classes along with other GPTs using some of the resources you have provided in your premium content and from Ethan Mollick. I am not sure how much more I would learn from the webinar. If it was a repeat of your premium content then I would not find it overly useful I suppose, but if it was other information, then I would be interested.”

In short: I am going to turn to more advanced topics more generally in 2025, in my Premium content and in my paid webinars. Free newsletters and free webinars will stay more focused on the basics and keeping everyone up to speed, but the webinar at issue here would focus primarily on topics about custom GPTs, Gems, Projects, etc. that are not covered in my pre-existing Tutorial. Topics like: API connections, workflow optimization, when to pivot to Zapier and LLMs’ APIs, etc.

This is especially true since Charlie Fuller and I are about to release a course on basic- and intermediate-level aspects of custom GPT creation (see above).

What'd you think of today's newsletter?

Login or Subscribe to participate in polls.

Graham

Let's transform learning together.

If you would like to consult with me or have me present to your team, discussing options is the first step:

Feel free to connect on LinkedIN, too!