Teacher AI Tools
Lots of benefits, without learning AI prompts
When I think back on all the crap I learned in high school
It’s a wonder I can think at all
And though my lack of education hasn’t hurt me, none
I can read the writing on the wall - Paul Simon
Okay, maybe crap is a strong word to use, but it emphasizes the point I want to make: AI is poised to make significant changes in education. Educators may be able to kick this can down the road a year or two, but it’s not going away. Already at the end of June 2024, over 3/4 of students and teachers were familiar with ChatGPT. And half of teachers and students are using AI tools at least weekly. For schools that ban AI tools in classrooms, that means students are probably using them at home. We are entering the (relatively short) era of prescriptive and proscriptive AI use in education - you may/must use AI for this part, but NOT FOR THAT. Unless we revert to the “in class only” model of education, I think schools will lose the battle if they adopt a “no AI” framework, rather than the “AI as helper” approach.
Over the past few weeks I have experimented with a number of AI tools, and I thought it would be helpful to share some of what I’ve found. My focus here is on using the AI LLMs (Gemini, ChatGPT, Claude, and to a lesser extent - at least right now - Grok, Llama, and Mistral) as opposed to using specific “AI platforms” that simplify interaction with the AI LLMs. However, it’s important to understand that even when using an AI platform (like MagicSchool), you are still interacting with an AI LLM behind the scenes. I look at this as two sides of the same coin, since you’re still relying on NLP and “best next word” to get results; the difference is the amount of effort you put in and the satisfaction you have with the results. But I’m getting ahead of myself.
So WHY would you want to use a set of teacher tools, rather than harness the raw power of the industry-leading AIs from Google, OpenAI, and Anthropic? There are several reasons why you might choose an AI platform rather than direct interaction with an LLM, plus a few reasons why you might opt to work directly with the LLM.
Reasons to choose an AI platform:
built-in guardrails for language, topics, cheating, privacy
point-and-click functionality for common use cases
levels the playing field for technical vs non-technical users
consistency of approach for schools
LLMs were not designed with schools in mind
Reasons to engage directly with LLMs:
deeper control of output via prompt engineering
ability to choose an LLM based on the task
immediate access to the newer models
Why might you or your sstaff choose to work with teacher tools? As I’ve mentioned in earlier posts, one of the strengths of LLMs is that they are designed to do all the heavy lifting for you, unless you specifically tell them NOT TO. For teaching and learning, I see this as a weakness. Rather than stimulate curiosity and guide users to a best solution, LLMs tend to jump to the answer and skip the hard work of brainstorming, problem solving, analogy making, and analysis - unless you give specific instructions to guide the process. Teachers know that students need to wrestle with ideas in order to learn; students need support, reinforcement, and practice, rather than JUST the right answer. And, while some teachers are more able and willing to get into the weeds with the LLM models, many want help with very specific topics - rubric making, turning a video into a presentation (or generating questions based on a video), lesson or unit plans matched to curriculum standards, brainstorming ideas, rewriting text for students below or above grade level, and more. One plug for using a standardized platform for teachers is that the user experience and the ability of “teachers helping teachers” is increased. Plus students will have a much more uniform experience, with AI still being a new thing in many classrooms.
The most popular AI platform right now is MagicSchool, which has many point-and-click tools available - from lesson and unit planning to communication and reporting to content creation. Most functionality is available on the free version, and I have yet to come across a use case that requires a subscription. The other platform that I experimented with is Khanmigo, which also has a wide range of tools available in its free version. While there are other platforms like Eduaide and SchoolAI, their free versions are much more limited, which will likely deter teacher use. I’ll take a deeper dive into both MagicSchool and Khanmigo later this month, since they represent the current best, free, and comprehensive AI platforms for teaching. And both have student tools as well, which is one of the goals for AI in education, right?
Okay, time for the other perspective - why might you bypass the ease of teacher tools in a curated AI platform and deal directly with the AI LLMs? Perhaps you like a challenge, or you’re an early adopter, or you have a more technical background, or you like to fine-tune the output that you get. I know that after looking at LLM output I sometimes say “This looks like a first draft. Do you think you can make this better?” Or “That’s not really what I was hoping for. How about focusing on <thing you want the LLM to address>?” Two other important considerations for using an LLM directly are: even with a free account, you still have (occasional) access to the newest models, which have significant advantages (in reasoning, context, processing, output) over the earlier models; plus each LLM is “best at” different things - Gemini (various types of input), ChatGPT (general purpose conversation), Claude (ethical considerations and nuanced instruction), Llama (best for research and development), and Grok (processing real-time information). The AI platforms like MagicSchool and Khanmigo provide good teacher tools, but not necessarily the best tool for the job. And sometimes that best tool for the job is not the one that is easiest to use.
If your goal is to use AI in your teaching, or to develop a consistent, school-wide approach to AI use, I don’t think you can go wrong with either of the AI platforms (or with a paid version to another platform, if you like). Teachers will help each other learn to use the tools, and students will get a uniform view of how AI enhances education. At the end of the day, teachers will be more efficient in their work, and students will see how AI-powered tools have the potential to make education better.
There is a high probability that “prompt engineering” - phrasing your input to get optimal results - is a temporary step in AI development; that future models will not rely on users asking questions in a certain way. And if that’s the case, maybe direct interaction with LLMs, rather than interaction via third-party tools, is just a phase. The idea is that as AI becomes more of a “general” intelligence, the models may ask for additional detail, or make assumptions based on past interactions, rather than asking you to explicitly state what you want. There certainly is a healthy debate over whether or not we are getting closer to AGI - Sam Altman vs Gary Marcus, for example.
So, do you use a platform of tools on a free or paid subscription, or do you interact directly with the AI LLMs? Or maybe you pick and choose, depending on the question of the day? If you are new to using AI as a teacher, MagicSchool or Khanmigo definitely give you the biggest bang for the buck; however, if you have been using AI for some time, you might benefit from working directly with the LLMs rather than with the specific use-cases (make a slide deck, summarize a video, match a unit plan to curriculum objectives) in the AI platforms. I don’t think there is a wrong answer here, but will be good to discuss with your colleagues as you move forward on AI-enhanced teaching and learning.
Stay tuned for a deeper dive on AI platforms with their point-and-click tools!
That’s all for now,
Cheers,
-Rick