3 Generative AI Applications in EdTech

Original post on Medium: 3 Generative AI Applications in EdTech | by Sam Bobo | Medium

In my previous post about?Generative AI and Learning Modalities , I explored the applicability of Generative Artificial Intelligence’s ability to apply medium transformation of curricula to satisfy various different learning types: auditory, visual, kinesthetic, and reading/writing. Upon further reflection and experimentation, there exists a larger portfolio of applicable use cases for Generative AI within the Education space I seek to explore.

The below examples use?OpenAI’s ChatGPT

“Explain As”?— As an educator, whether of formal curricula or extracurricular activities, one constantly seeks the “Aha!” moment there understanding has been solidified and the learning complete. Often times, however, finding that explanation within a particular learning modality might be difficult, whether the concept lacks the interest of the student or the subject matter might be too complex and requires further breakdown.

Many who have experimented with ChatGPT prompts understand that crafting prompts is critical for the success of the outputted content. Typically, when crafting a prompt, the user furnishes a specific role to the system to assume, whether be a Product Manager, Loan Officer, Paralegal, or other, tuning the generative model to write in a specific prose and subject matter vocabulary. Furthermore, this prompt engineering can apply to manner of the specified output, ranging from writing a poem, formal document, and more… the options are endless and completely open-ended. The underlying Large Language Models and broad training allow GPT-3 to build those metaphors, a capability previously unheard of at scale.

Welcome the concept of “explain as.” As an educator, one of your responsibilities is understanding the interests of the student. Take, for example, a student who is fascinated with sports. One could assume that explaining complex and/or “boring” concepts to said student through the lens of the subject matter she/he are interested could yield a higher probability of success that she/he would achieve that “Aha!” moment.

I crafted a ChatGPT prompt to do just that: explain eigenvectors to a student interested in sports. Here are the results:

As you can observe, the system provides a simplistic example in a basketball metaphor using a 3-point shot. This example can be repeated for any/all student interests. For illustrative purposes, I asked ChatGPT to generate another explanation, however, for a student interested in space.

Inductive Learning?— students who are highly motivated have a high degree of intellectual curiosity. While the internet and connected world have vastly accelerated the proliferation of information and web indexers such as Google make filtering for that information quicker, selecting the “blue links” and finding information within a web page could take longer than desired. While certainly tradeoffs exist between computer generated responses and self-indulged searching (for example, the ability to see references and augment knowledge further), students and interested parties alike can simply prompt a Generative AI system with a question they are curious about and get a response:

What is fascinating in the above example is that ChatGPT also extracted source references for further reading, allowing for a hybrid blend of the aforementioned tradeoffs. Furthermore, since LLMs maintain context throughout the duration of the chat, students can ask further questions to delve deeper into the subject matter, a method of inductive learning.

Learning Trees?— while standard curricula exists for primary and secondary education, the path towards a desired outcome can be unique, especially with the presence of micro-learning, massively open online courses, and industry created learning materials — one example includes micro-degrees and digital badging, whoes acceptance is growing larger among organizations daily. Furthermore, the ability to take on independent projects, either as a hobbies or consultant are more widely available through the ability to open source, showcase work, or take part within a competition. The fundamental problem, exists however, about what is the simplest path to obtaining such a desired outcome. Learning tress, similar to a role play game can detail subject material and associated material thereafter in a graph-like manner that build to a desired outcome. EdTech organizations have been striving to build these types of graphs, however, GPT-3 and other Large Language Models have the capability to generate these given its massive and broad training.

Again, as illistrative purposes, I prompted ChatGPT to generate a learning tree on Linear Algebra (our subject de jour)

One can recursively then prompt on each of the subjects within the learning taxonomy to build a tree-like structure and map a subject-matters learning objectives which is absolutely fascinating.

Access to education and high quality information existed, large language models democratized access! I am exited to watch the development of the EdTech space using Generative AI and hope to see some of the predictions above included in solutions that will shape the next generation of learners!