AI Researcher

When it comes to recent major influences on education, there are few things as prominent as generative AI chatbots. Comprising tools like ChatGPT and Google’s Bard, generative AI has swept through schools across the country at an unprecedented rate. Their ability to find solutions to problems and create written works that can seemingly bypass plagiarism checkers (N.B. there are now AI-powered tools that can detect AI-generated work) has forced curriculums to rapidly evolve with this new “AI revolution.” It’s understandable, then, that the terms “ChatGPT” and “AI helper” have had a suboptimal reputation among teachers and instructors. But ironically enough, many educators have themselves begun to use AI tools, and for good reason as well! Many tasks required by instructors are quite tedious. From grading multiple choice answers to writing trivial emails, these duties are often interpreted as time-consuming and detract from the time that could be spent communicating with students and designing quality lesson plans. The former can be especially helpful, as much of the time, grading can be little more than a brain-numbing tedium. And given the customizable nature of AI tools, they serve as the perfect solution, especially for groups like world language instructors.

Before jumping into applications, it is first important to understand the basics of why AI tools in particular are so powerful. Unlike traditional modes of technological tools, modern AI models implement what is known as a neural network, a complex, multilayer set of connections between nodes (or “neurons”) whose nonlinearity can theoretically help identify almost any type of pattern or connection. These networks offer a sense of malleability to AI models, one which can enable the model to fit any mold required by the user. This allows for customization. For instance, if teachers properly query ChatGPT for particular grading help (using something called a “prompt”), the AI chatbot could, in theory, perform whatever grading task was asked.

So why is this particularly useful for language teachers? Well, unlike for other subjects, automatically grading language assignments often requires multiple tools for multiple levels of processing: first would have to translate the text into something the particular program would understand (e.g., English for programs designed around the English language), and only then could it perform the grading task at hand. Generative AI condenses this into one step. Chatbots like ChatGPT are trained on information scraped from all over the internet. As a result, users can expect these models to understand any language well documented online. So, the chatbot can effectively interpret the language in its native language, and depending on the prompt, output something in the same language or an entirely different one.

Michael Cai

AI as a Foreign Language Assignment Grader

Creator of LowTech AI

If you’ve encountered text written by a computer in the past year (you definitely have), it was probably generated by a large language model, AKA an LLM.

LLMs are the AI technology behind Gmail autocomplete, ChatGPT, and 

, but what does that term actually mean? As these systems get more integrated with our work and lives, it’s worthwhile to understand the basics. Here’s the absolute bare necessities you should know.

First off, let’s go through the meaning of each word in “Large Language Models.”


Jack O'Brien

If you do work that’s on a computer, LLMs will be a part of your life. I think it's important to learn the basics.

What’s a Large Language Model?

Using ChatGPT effectively is an acquired skill. Without prior experience and learned intuition, your AI generated responses will probably be improperly formatted, way too long, a bit cold in tone, or simply not what you were hoping for.

I have a good deal of experience working with opaque software systems, so I was lucky to pick up the basics of AI prompting fairly quickly (though I’m still far from an expert prompter). You need to be extremely specific. You need to be concise. You need to explain your thought process and your desired output. Lots of people have said that AI prompting today is like learning how to use Google in 2001, and I think there’s truth to that idea. We’re still learning how to correctly poke the machine.

 people have used ChatGPT, and I’m fairly certain the vast majority of them don’t know how to use it to the best of its ability (enter 1000+ Twitter threads and LinkedIn posts on how “you’re not using ChatGPT correctly”). AI chatbots open up infinite possibilities to explore virtually all documented human knowledge and experience, but a small uninteresting subset of that potential is what most people see today.

AI users don’t really care about the prompt, they care about the value generated from the prompt and the AI model. Signaling that value is more important during the discovery and use of a prompt than understanding the underlying template.

Prompt Engineering

AI as a Foreign Language Assignment Grader

What’s a Large Language Model?

Hide Your Prompt: A Better UX for Sharing Simple AI Tools